nushell/crates/nu-parser/src/parser.rs

6282 lines
224 KiB
Rust
Raw Normal View History

2021-09-02 10:25:22 +02:00
use crate::{
allow lists to have type annotations (#8529) this pr refines #8270 and closes #8109 # description examples: the original syntax is okay ```nu def okay [nums: list] {} # the type of list will be list<any> ``` empty annotations are allowed in any variation the last two may be caught by a future formatter, but do not affect `nu` code currently ```nu def okay [nums: list<>] {} # okay def okay [nums: list< >] {} # weird but also okay def okay [nums: list< >] {} # also weird but okay ``` types are allowed (See [notes](#notes) below) ```nu def okay [nums: list<int>] {} # `test [a b c]` will throw an error def okay [nums: list< int > {} # any amount of space within the angle brackets is okay def err [nums: list <int>] {} # this is not okay, `nums` and `<int>` will be parsed as # two separate params, ``` nested annotations are allowed in many variations ```nu def okay [items: list<list<int>>] {} def okay [items: list<list>] {} ``` any unterminated annotation is caught ```nu Error: nu::parser::unexpected_eof × Unexpected end of code. ╭─[source:1:1] 1 │ def err [nums: list<int] {} · ▲ · ╰── expected closing > ╰──── ``` unknown types are flagged ```nu Error: nu::parser::unknown_type × Unknown type. ╭─[source:1:1] 1 │ def err [nums: list<str>] {} · ─┬─ · ╰── unknown type ╰──── Error: nu::parser::unknown_type × Unknown type. ╭─[source:1:1] 1 │ def err [nums: list<int, string>] {} · ─────┬───── · ╰── unknown type ╰──── ``` # notes the error message for mismatched types in not as intuitive ```nu Error: nu::parser::parse_mismatch × Parse mismatch during operation. ╭─[source:1:1] 1 │ def err [nums: list<int>] {}; err [a b c] · ┬ · ╰── expected int ╰──── ``` it should be something like this ```nu Error: nu::parser::parse_mismatch × Parse mismatch during operation. ╭─[source:1:1] 1 │ def err [nums: list<int>] {}; err [a b c] · ──┬── · ╰── expected list<int> ╰──── ``` this is currently not implemented
2023-03-24 12:54:06 +01:00
lex::{lex, lex_signature},
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
lite_parser::{lite_parse, LiteCommand, LitePipeline, LiteRedirection, LiteRedirectionTarget},
parse_keywords::*,
parse_patterns::parse_pattern,
parse_shape_specs::{parse_shape_name, parse_type, ShapeDescriptorUse},
Input output checking (#9680) # Description This PR tights input/output type-checking a bit more. There are a lot of commands that don't have correct input/output types, so part of the effort is updating them. This PR now contains updates to commands that had wrong input/output signatures. It doesn't add examples for these new signatures, but that can be follow-up work. # User-Facing Changes BREAKING CHANGE BREAKING CHANGE This work enforces many more checks on pipeline type correctness than previous nushell versions. This strictness may uncover incompatibilities in existing scripts or shortcomings in the type information for internal commands. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-14 05:20:35 +02:00
type_check::{self, math_result_type, type_compatible},
Token, TokenContents,
2021-08-17 01:00:00 +02:00
};
use itertools::Itertools;
use log::trace;
use nu_engine::DIR_VAR_PARSER_INFO;
2021-09-02 03:29:43 +02:00
use nu_protocol::{
ast::*, engine::StateWorkingSet, eval_const::eval_constant, BlockId, DidYouMean, Flag,
ParseError, PositionalArg, Signature, Span, Spanned, SyntaxShape, Type, VarId, ENV_VARIABLE_ID,
IN_VARIABLE_ID,
2021-09-02 03:29:43 +02:00
};
Make EngineState clone cheaper with Arc on all of the heavy objects (#12229) # Description This makes many of the larger objects in `EngineState` into `Arc`, and uses `Arc::make_mut` to do clone-on-write if the reference is not unique. This is generally very cheap, giving us the best of both worlds - allowing us to mutate without cloning if we have an exclusive reference, and cloning if we don't. This started as more of a curiosity for me after remembering that `Arc::make_mut` exists and can make using `Arc` for mostly immutable data that sometimes needs to be changed very convenient, and also after hearing someone complain about memory usage on Discord - this is a somewhat significant win for that. The exact objects that were wrapped in `Arc`: - `files`, `file_contents` - the strings and byte buffers - `decls` - the whole `Vec`, but mostly to avoid lots of individual `malloc()` calls on Clone rather than for memory usage - `blocks` - the blocks themselves, rather than the outer Vec - `modules` - the modules themselves, rather than the outer Vec - `env_vars`, `previous_env_vars` - the entire maps - `config` The changes required were relatively minimal, but this is a breaking API change. In particular, blocks are added as Arcs, to allow the parser cache functionality to work. With my normal nu config, running on Linux, this saves me about 15 MiB of process memory usage when running interactively (65 MiB → 50 MiB). This also makes quick command executions cheaper, particularly since every REPL loop now involves a clone of the engine state so that we can recover from a panic. It also reduces memory usage where engine state needs to be cloned and sent to another thread or kept within an iterator. # User-Facing Changes Shouldn't be any, since it's all internal stuff, but it does change some public interfaces so it's a breaking change
2024-03-19 19:07:00 +01:00
use std::{
borrow::Cow,
Make EngineState clone cheaper with Arc on all of the heavy objects (#12229) # Description This makes many of the larger objects in `EngineState` into `Arc`, and uses `Arc::make_mut` to do clone-on-write if the reference is not unique. This is generally very cheap, giving us the best of both worlds - allowing us to mutate without cloning if we have an exclusive reference, and cloning if we don't. This started as more of a curiosity for me after remembering that `Arc::make_mut` exists and can make using `Arc` for mostly immutable data that sometimes needs to be changed very convenient, and also after hearing someone complain about memory usage on Discord - this is a somewhat significant win for that. The exact objects that were wrapped in `Arc`: - `files`, `file_contents` - the strings and byte buffers - `decls` - the whole `Vec`, but mostly to avoid lots of individual `malloc()` calls on Clone rather than for memory usage - `blocks` - the blocks themselves, rather than the outer Vec - `modules` - the modules themselves, rather than the outer Vec - `env_vars`, `previous_env_vars` - the entire maps - `config` The changes required were relatively minimal, but this is a breaking API change. In particular, blocks are added as Arcs, to allow the parser cache functionality to work. With my normal nu config, running on Linux, this saves me about 15 MiB of process memory usage when running interactively (65 MiB → 50 MiB). This also makes quick command executions cheaper, particularly since every REPL loop now involves a clone of the engine state so that we can recover from a panic. It also reduces memory usage where engine state needs to be cloned and sent to another thread or kept within an iterator. # User-Facing Changes Shouldn't be any, since it's all internal stuff, but it does change some public interfaces so it's a breaking change
2024-03-19 19:07:00 +01:00
collections::{HashMap, HashSet},
num::ParseIntError,
str,
Make EngineState clone cheaper with Arc on all of the heavy objects (#12229) # Description This makes many of the larger objects in `EngineState` into `Arc`, and uses `Arc::make_mut` to do clone-on-write if the reference is not unique. This is generally very cheap, giving us the best of both worlds - allowing us to mutate without cloning if we have an exclusive reference, and cloning if we don't. This started as more of a curiosity for me after remembering that `Arc::make_mut` exists and can make using `Arc` for mostly immutable data that sometimes needs to be changed very convenient, and also after hearing someone complain about memory usage on Discord - this is a somewhat significant win for that. The exact objects that were wrapped in `Arc`: - `files`, `file_contents` - the strings and byte buffers - `decls` - the whole `Vec`, but mostly to avoid lots of individual `malloc()` calls on Clone rather than for memory usage - `blocks` - the blocks themselves, rather than the outer Vec - `modules` - the modules themselves, rather than the outer Vec - `env_vars`, `previous_env_vars` - the entire maps - `config` The changes required were relatively minimal, but this is a breaking API change. In particular, blocks are added as Arcs, to allow the parser cache functionality to work. With my normal nu config, running on Linux, this saves me about 15 MiB of process memory usage when running interactively (65 MiB → 50 MiB). This also makes quick command executions cheaper, particularly since every REPL loop now involves a clone of the engine state so that we can recover from a panic. It also reduces memory usage where engine state needs to be cloned and sent to another thread or kept within an iterator. # User-Facing Changes Shouldn't be any, since it's all internal stuff, but it does change some public interfaces so it's a breaking change
2024-03-19 19:07:00 +01:00
sync::Arc,
};
pub fn garbage(working_set: &mut StateWorkingSet, span: Span) -> Expression {
Expression::garbage(working_set, span)
2021-07-01 02:01:04 +02:00
}
pub fn garbage_pipeline(working_set: &mut StateWorkingSet, spans: &[Span]) -> Pipeline {
Pipeline::from_vec(vec![garbage(working_set, Span::concat(spans))])
2021-09-10 09:28:43 +02:00
}
2021-07-01 03:31:02 +02:00
fn is_identifier_byte(b: u8) -> bool {
b != b'.'
&& b != b'['
&& b != b'('
&& b != b'{'
&& b != b'+'
&& b != b'-'
&& b != b'*'
&& b != b'^'
&& b != b'/'
&& b != b'='
&& b != b'!'
&& b != b'<'
&& b != b'>'
&& b != b'&'
&& b != b'|'
2021-07-01 03:31:02 +02:00
}
pub fn is_math_expression_like(working_set: &mut StateWorkingSet, span: Span) -> bool {
let bytes = working_set.get_span_contents(span);
if bytes.is_empty() {
return false;
}
if bytes == b"true"
|| bytes == b"false"
|| bytes == b"null"
|| bytes == b"not"
|| bytes == b"if"
Add pattern matching (#8590) # Description This adds `match` and basic pattern matching. An example: ``` match $x { 1..10 => { print "Value is between 1 and 10" } { foo: $bar } => { print $"Value has a 'foo' field with value ($bar)" } [$a, $b] => { print $"Value is a list with two items: ($a) and ($b)" } _ => { print "Value is none of the above" } } ``` Like the recent changes to `if` to allow it to be used as an expression, `match` can also be used as an expression. This allows you to assign the result to a variable, eg) `let xyz = match ...` I've also included a short-hand pattern for matching records, as I think it might help when doing a lot of record patterns: `{$foo}` which is equivalent to `{foo: $foo}`. There are still missing components, so consider this the first step in full pattern matching support. Currently missing: * Patterns for strings * Or-patterns (like the `|` in Rust) * Patterns for tables (unclear how we want to match a table, so it'll need some design) * Patterns for binary values * And much more # User-Facing Changes [see above] # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-03-24 02:52:01 +01:00
|| bytes == b"match"
{
return true;
}
let b = bytes[0];
add raw-string literal support (#9956) # Description This PR adds raw string support by using `r#` at the beginning of single quoted strings and `#` at the end. Notice that escapes do not process, even within single quotes, parentheses don't mean anything, $variables don't mean anything. It's just a string. ```nushell ❯ echo r#'one\ntwo (blah) ($var)'# one\ntwo (blah) ($var) ``` Notice how they work without `echo` or `print` and how they work without carriage returns. ```nushell ❯ r#'adsfa'# adsfa ❯ r##"asdfa'@qpejq'## asdfa'@qpejq ❯ r#'asdfasdfasf ∙ foqwejfqo@'23rfjqf'# ``` They also have a special configurable color in the repl. (use single quotes though) ![image](https://github.com/nushell/nushell/assets/343840/8780e21d-de4c-45b3-9880-2425f5fe10ef) They should work like rust raw literals and allow `r##`, `r###`, `r####`, etc, to help with having one or many `#`'s in the middle of your raw-string. They should work with `let` as well. ```nushell r#'some\nraw\nstring'# | str upcase ``` closes https://github.com/nushell/nushell/issues/5091 # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: WindSoilder <WindSoilder@outlook.com> Co-authored-by: Ian Manske <ian.manske@pm.me>
2024-05-02 15:36:37 +02:00
// check for raw string
if bytes.starts_with(b"r#") {
return true;
}
if b == b'(' || b == b'{' || b == b'[' || b == b'$' || b == b'"' || b == b'\'' || b == b'-' {
return true;
}
let starting_error_count = working_set.parse_errors.len();
// Number
parse_number(working_set, span);
if working_set.parse_errors.len() == starting_error_count {
return true;
}
working_set.parse_errors.truncate(starting_error_count);
// Filesize
parse_filesize(working_set, span);
if working_set.parse_errors.len() == starting_error_count {
return true;
}
working_set.parse_errors.truncate(starting_error_count);
parse_duration(working_set, span);
if working_set.parse_errors.len() == starting_error_count {
return true;
}
working_set.parse_errors.truncate(starting_error_count);
parse_datetime(working_set, span);
if working_set.parse_errors.len() == starting_error_count {
return true;
}
working_set.parse_errors.truncate(starting_error_count);
parse_binary(working_set, span);
if working_set.parse_errors.len() == starting_error_count {
return true;
}
working_set.parse_errors.truncate(starting_error_count);
parse_range(working_set, span);
if working_set.parse_errors.len() == starting_error_count {
return true;
}
working_set.parse_errors.truncate(starting_error_count);
false
}
2021-07-01 03:31:02 +02:00
fn is_identifier(bytes: &[u8]) -> bool {
bytes.iter().all(|x| is_identifier_byte(*x))
}
Add pattern matching (#8590) # Description This adds `match` and basic pattern matching. An example: ``` match $x { 1..10 => { print "Value is between 1 and 10" } { foo: $bar } => { print $"Value has a 'foo' field with value ($bar)" } [$a, $b] => { print $"Value is a list with two items: ($a) and ($b)" } _ => { print "Value is none of the above" } } ``` Like the recent changes to `if` to allow it to be used as an expression, `match` can also be used as an expression. This allows you to assign the result to a variable, eg) `let xyz = match ...` I've also included a short-hand pattern for matching records, as I think it might help when doing a lot of record patterns: `{$foo}` which is equivalent to `{foo: $foo}`. There are still missing components, so consider this the first step in full pattern matching support. Currently missing: * Patterns for strings * Or-patterns (like the `|` in Rust) * Patterns for tables (unclear how we want to match a table, so it'll need some design) * Patterns for binary values * And much more # User-Facing Changes [see above] # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-03-24 02:52:01 +01:00
pub fn is_variable(bytes: &[u8]) -> bool {
2021-07-01 03:31:02 +02:00
if bytes.len() > 1 && bytes[0] == b'$' {
is_identifier(&bytes[1..])
} else {
is_identifier(bytes)
}
}
2021-11-14 20:40:26 +01:00
pub fn trim_quotes(bytes: &[u8]) -> &[u8] {
if (bytes.starts_with(b"\"") && bytes.ends_with(b"\"") && bytes.len() > 1)
|| (bytes.starts_with(b"\'") && bytes.ends_with(b"\'") && bytes.len() > 1)
|| (bytes.starts_with(b"`") && bytes.ends_with(b"`") && bytes.len() > 1)
2021-11-14 20:40:26 +01:00
{
&bytes[1..(bytes.len() - 1)]
} else {
bytes
}
}
pub fn trim_quotes_str(s: &str) -> &str {
if (s.starts_with('"') && s.ends_with('"') && s.len() > 1)
|| (s.starts_with('\'') && s.ends_with('\'') && s.len() > 1)
|| (s.starts_with('`') && s.ends_with('`') && s.len() > 1)
{
&s[1..(s.len() - 1)]
} else {
s
}
}
pub(crate) fn check_call(
working_set: &mut StateWorkingSet,
command: Span,
sig: &Signature,
call: &Call,
) {
2021-10-13 19:53:27 +02:00
// Allow the call to pass if they pass in the help flag
if call.named_iter().any(|(n, _, _)| n.item == "help") {
return;
2021-10-13 19:53:27 +02:00
}
if call.positional_len() < sig.required_positional.len() {
// Comparing the types of all signature positional arguments against the parsed
// expressions found in the call. If one type is not found then it could be assumed
// that that positional argument is missing from the parsed call
for argument in &sig.required_positional {
let found = call.positional_iter().fold(false, |ac, expr| {
if argument.shape.to_type() == expr.ty || argument.shape == SyntaxShape::Any {
true
} else {
ac
}
});
if !found {
if let Some(last) = call.positional_iter().last() {
working_set.error(ParseError::MissingPositional(
2022-01-04 00:14:33 +01:00
argument.name.clone(),
Protocol: debug_assert!() Span to reflect a valid slice (#6806) Also enforce this by #[non_exhaustive] span such that going forward we cannot, in debug builds (1), construct invalid spans. The motivation for this stems from #6431 where I've seen crashes due to invalid slice indexing. My hope is this will mitigate such senarios 1. https://github.com/nushell/nushell/pull/6431#issuecomment-1278147241 # Description (description of your pull request here) # Tests Make sure you've done the following: - [ ] Add tests that cover your changes, either in the command examples, the crate/tests folder, or in the /tests folder. - [ ] Try to think about corner cases and various ways how your changes could break. Cover them with tests. - [ ] If adding tests is not possible, please document in the PR body a minimal example with steps on how to reproduce so one can verify your change works. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [ ] `cargo clippy --workspace --features=extra -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [ ] `cargo test --workspace --features=extra` to check that all the tests pass # Documentation - [ ] If your PR touches a user-facing nushell feature then make sure that there is an entry in the documentation (https://github.com/nushell/nushell.github.io) for the feature, and update it if necessary.
2022-12-03 10:44:12 +01:00
Span::new(last.span.end, last.span.end),
2022-01-04 00:14:33 +01:00
sig.call_signature(),
));
return;
2022-01-04 00:14:33 +01:00
} else {
working_set.error(ParseError::MissingPositional(
2022-01-04 00:14:33 +01:00
argument.name.clone(),
Protocol: debug_assert!() Span to reflect a valid slice (#6806) Also enforce this by #[non_exhaustive] span such that going forward we cannot, in debug builds (1), construct invalid spans. The motivation for this stems from #6431 where I've seen crashes due to invalid slice indexing. My hope is this will mitigate such senarios 1. https://github.com/nushell/nushell/pull/6431#issuecomment-1278147241 # Description (description of your pull request here) # Tests Make sure you've done the following: - [ ] Add tests that cover your changes, either in the command examples, the crate/tests folder, or in the /tests folder. - [ ] Try to think about corner cases and various ways how your changes could break. Cover them with tests. - [ ] If adding tests is not possible, please document in the PR body a minimal example with steps on how to reproduce so one can verify your change works. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [ ] `cargo clippy --workspace --features=extra -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [ ] `cargo test --workspace --features=extra` to check that all the tests pass # Documentation - [ ] If your PR touches a user-facing nushell feature then make sure that there is an entry in the documentation (https://github.com/nushell/nushell.github.io) for the feature, and update it if necessary.
2022-12-03 10:44:12 +01:00
Span::new(command.end, command.end),
2022-01-04 00:14:33 +01:00
sig.call_signature(),
));
return;
2022-01-04 00:14:33 +01:00
}
}
}
let missing = &sig.required_positional[call.positional_len()];
if let Some(last) = call.positional_iter().last() {
working_set.error(ParseError::MissingPositional(
2022-01-04 00:14:33 +01:00
missing.name.clone(),
Protocol: debug_assert!() Span to reflect a valid slice (#6806) Also enforce this by #[non_exhaustive] span such that going forward we cannot, in debug builds (1), construct invalid spans. The motivation for this stems from #6431 where I've seen crashes due to invalid slice indexing. My hope is this will mitigate such senarios 1. https://github.com/nushell/nushell/pull/6431#issuecomment-1278147241 # Description (description of your pull request here) # Tests Make sure you've done the following: - [ ] Add tests that cover your changes, either in the command examples, the crate/tests folder, or in the /tests folder. - [ ] Try to think about corner cases and various ways how your changes could break. Cover them with tests. - [ ] If adding tests is not possible, please document in the PR body a minimal example with steps on how to reproduce so one can verify your change works. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [ ] `cargo clippy --workspace --features=extra -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [ ] `cargo test --workspace --features=extra` to check that all the tests pass # Documentation - [ ] If your PR touches a user-facing nushell feature then make sure that there is an entry in the documentation (https://github.com/nushell/nushell.github.io) for the feature, and update it if necessary.
2022-12-03 10:44:12 +01:00
Span::new(last.span.end, last.span.end),
2022-01-04 00:14:33 +01:00
sig.call_signature(),
))
} else {
working_set.error(ParseError::MissingPositional(
2022-01-04 00:14:33 +01:00
missing.name.clone(),
Protocol: debug_assert!() Span to reflect a valid slice (#6806) Also enforce this by #[non_exhaustive] span such that going forward we cannot, in debug builds (1), construct invalid spans. The motivation for this stems from #6431 where I've seen crashes due to invalid slice indexing. My hope is this will mitigate such senarios 1. https://github.com/nushell/nushell/pull/6431#issuecomment-1278147241 # Description (description of your pull request here) # Tests Make sure you've done the following: - [ ] Add tests that cover your changes, either in the command examples, the crate/tests folder, or in the /tests folder. - [ ] Try to think about corner cases and various ways how your changes could break. Cover them with tests. - [ ] If adding tests is not possible, please document in the PR body a minimal example with steps on how to reproduce so one can verify your change works. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [ ] `cargo clippy --workspace --features=extra -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [ ] `cargo test --workspace --features=extra` to check that all the tests pass # Documentation - [ ] If your PR touches a user-facing nushell feature then make sure that there is an entry in the documentation (https://github.com/nushell/nushell.github.io) for the feature, and update it if necessary.
2022-12-03 10:44:12 +01:00
Span::new(command.end, command.end),
2022-01-04 00:14:33 +01:00
sig.call_signature(),
))
}
2021-07-02 04:22:54 +02:00
} else {
for req_flag in sig.named.iter().filter(|x| x.required) {
if call.named_iter().all(|(n, _, _)| n.item != req_flag.long) {
working_set.error(ParseError::MissingRequiredFlag(
2021-07-02 04:22:54 +02:00
req_flag.long.clone(),
command,
));
}
}
}
}
Allow spreading arguments to commands (#11289) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> Finishes implementing https://github.com/nushell/nushell/issues/10598, which asks for a spread operator in lists, in records, and when calling commands. # Description <!-- Thank you for improving Nushell. Please, check our [contributing guide](../CONTRIBUTING.md) and talk to the core team before making major changes. Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience. --> This PR will allow spreading arguments to commands (both internal and external). It will also deprecate spreading arguments automatically when passing to external commands. # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> - Users will be able to use `...` to spread arguments to custom/builtin commands that have rest parameters or allow unknown arguments, or to any external command - If a custom command doesn't have a rest parameter and it doesn't allow unknown arguments either, the spread operator will not be allowed - Passing lists to external commands without `...` will work for now but will cause a deprecation warning saying that it'll stop working in 0.91 (is 2 versions enough time?) Here's a function to help with demonstrating some behavior: ```nushell > def foo [ a, b, c?, d?, ...rest ] { [$a $b $c $d $rest] | to nuon } ``` You can pass a list of arguments to fill in the `rest` parameter using `...`: ```nushell > foo 1 2 3 4 ...[5 6] [1, 2, 3, 4, [5, 6]] ``` If you don't use `...`, the list `[5 6]` will be treated as a single argument: ```nushell > foo 1 2 3 4 [5 6] # Note the double [[]] [1, 2, 3, 4, [[5, 6]]] ``` You can omit optional parameters before the spread arguments: ```nushell > foo 1 2 3 ...[4 5] # d is omitted here [1, 2, 3, null, [4, 5]] ``` If you have multiple lists, you can spread them all: ```nushell > foo 1 2 3 ...[4 5] 6 7 ...[8] ...[] [1, 2, 3, null, [4, 5, 6, 7, 8]] ``` Here's the kind of error you get when you try to spread arguments to a command with no rest parameter: ![image](https://github.com/nushell/nushell/assets/45539777/93faceae-00eb-4e59-ac3f-17f98436e6e4) And this is the warning you get when you pass a list to an external now (without `...`): ![image](https://github.com/nushell/nushell/assets/45539777/d368f590-201e-49fb-8b20-68476ced415e) # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass (on Windows make sure to [enable developer mode](https://learn.microsoft.com/en-us/windows/apps/get-started/developer-mode-features-and-debugging)) - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> Added tests to cover the following cases: - Spreading arguments to a command that doesn't have a rest parameter (unexpected spread argument error) - Spreading arguments to a command that doesn't have a rest parameter *but* there's also a missing positional argument (missing positional error) - Spreading arguments to a command that doesn't have a rest parameter but does allow unknown arguments, such as `exec` (allowed) - Spreading a list literal containing arguments of the wrong type (parse error) - Spreading a non-list value, both to internal and external commands - Having named arguments in the middle of rest arguments - `explain`ing a command call that spreads its arguments # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> # Examples Suppose you have multiple tables: ```nushell let people = [[id name age]; [0 alice 100] [1 bob 200] [2 eve 300]] let evil_twins = [[id name age]; [0 ecila 100] [-1 bob 200] [-2 eve 300]] ``` Maybe you often find yourself needing to merge multiple tables and want a utility to do that. You could write a function like this: ```nushell def merge_all [ ...tables ] { $tables | reduce { |it, acc| $acc | merge $it } } ``` Then you can use it like this: ```nushell > merge_all ...([$people $evil_twins] | each { |$it| $it | select name age }) ╭───┬───────┬─────╮ │ # │ name │ age │ ├───┼───────┼─────┤ │ 0 │ ecila │ 100 │ │ 1 │ bob │ 200 │ │ 2 │ eve │ 300 │ ╰───┴───────┴─────╯ ``` Except they had duplicate columns, so now you first want to suffix every column with a number to tell you which table the column came from. You can make a command for that: ```nushell def select_and_merge [ --cols: list<string>, ...tables ] { let renamed_tables = $tables | enumerate | each { |it| $it.item | select $cols | rename ...($cols | each { |col| $col + ($it.index | into string) }) }; merge_all ...$renamed_tables } ``` And call it like this: ```nushell > select_and_merge --cols [name age] $people $evil_twins ╭───┬───────┬──────┬───────┬──────╮ │ # │ name0 │ age0 │ name1 │ age1 │ ├───┼───────┼──────┼───────┼──────┤ │ 0 │ alice │ 100 │ ecila │ 100 │ │ 1 │ bob │ 200 │ bob │ 200 │ │ 2 │ eve │ 300 │ eve │ 300 │ ╰───┴───────┴──────┴───────┴──────╯ ``` --- Suppose someone's made a command to search for APT packages: ```nushell # The main command def search-pkgs [ --install # Whether to install any packages it finds log_level: int # Pretend it's a good idea to make this a required positional parameter exclude?: list<string> # Packages to exclude repositories?: list<string> # Which repositories to look in (searches in all if not given) ...pkgs # Package names to search for ] { { install: $install, log_level: $log_level, exclude: ($exclude | to nuon), repositories: ($repositories | to nuon), pkgs: ($pkgs | to nuon) } } ``` It has a lot of parameters to configure it, so you might make your own helper commands to wrap around it for specific cases. Here's one example: ```nushell # Only look for packages locally def search-pkgs-local [ --install # Whether to install any packages it finds log_level: int exclude?: list<string> # Packages to exclude ...pkgs # Package names to search for ] { # All required and optional positional parameters are given search-pkgs --install=$install $log_level [] ["<local URI or something>"] ...$pkgs } ``` And you can run it like this: ```nushell > search-pkgs-local --install=false 5 ...["python2.7" "vim"] ╭──────────────┬──────────────────────────────╮ │ install │ false │ │ log_level │ 5 │ │ exclude │ [] │ │ repositories │ ["<local URI or something>"] │ │ pkgs │ ["python2.7", vim] │ ╰──────────────┴──────────────────────────────╯ ``` One thing I realized when writing this was that if we decide to not allow passing optional arguments using the spread operator, then you can (mis?)use the spread operator to skip optional parameters. Here, I didn't want to give `exclude` explicitly, so I used a spread operator to pass the packages to install. Without it, I would've needed to do `search-pkgs-local --install=false 5 [] "python2.7" "vim"` (explicitly pass `[]` (or `null`, in the general case) to `exclude`). There are probably more idiomatic ways to do this, but I just thought it was something interesting. If you're a virologist of the [xkcd](https://xkcd.com/350/) kind, another helper command you might make is this: ```nushell # Install any packages it finds def live-dangerously [ ...pkgs ] { # One optional argument was given (exclude), while another was not (repositories) search-pkgs 0 [] ...$pkgs --install # Flags can go after spread arguments } ``` Running it: ```nushell > live-dangerously "git" "*vi*" # *vi* because I don't feel like typing out vim and neovim ╭──────────────┬─────────────╮ │ install │ true │ │ log_level │ 0 │ │ exclude │ [] │ │ repositories │ null │ │ pkgs │ [git, *vi*] │ ╰──────────────┴─────────────╯ ``` Here's an example that uses the spread operator more than once within the same command call: ```nushell let extras = [ chrome firefox python java git ] def search-pkgs-curated [ ...pkgs ] { (search-pkgs 1 [emacs] ["example.com", "foo.com"] vim # A must for everyone! ...($pkgs | filter { |p| not ($p | str contains "*") }) # Remove packages with globs python # Good tool to have ...$extras --install=false python3) # I forget, did I already put Python in extras? } ``` Running it: ```nushell > search-pkgs-curated "git" "*vi*" ╭──────────────┬───────────────────────────────────────────────────────────────────╮ │ install │ false │ │ log_level │ 1 │ │ exclude │ [emacs] │ │ repositories │ [example.com, foo.com] │ │ pkgs │ [vim, git, python, chrome, firefox, python, java, git, "python3"] │ ╰──────────────┴───────────────────────────────────────────────────────────────────╯ ```
2023-12-28 08:43:20 +01:00
fn parse_external_arg(working_set: &mut StateWorkingSet, span: Span) -> ExternalArgument {
Re-implement aliases (#8123) # Description This PR adds an alternative alias implementation. Old aliases still work but you need to use `old-alias` instead of `alias`. Instead of replacing spans in the original code and re-parsing, which proved to be extremely error-prone and a constant source of panics, the new implementation creates a new command that references the old command. Consider the new alias defined as `alias ll = ls -l`. The parser creates a new command called `ll` and remembers that it is actually a `ls` command called with the `-l` flag. Then, when the parser sees the `ll` command, it will translate it to `ls -l` and passes to it any parameters that were passed to the call to `ll`. It works quite similar to how known externals defined with `extern` are implemented. The new alias implementation should work the same way as the old aliases, including exporting from modules, referencing both known and unknown externals. It seems to preserve custom completions and pipeline metadata. It is quite robust in most cases but there are some rough edges (see later). Fixes https://github.com/nushell/nushell/issues/7648, https://github.com/nushell/nushell/issues/8026, https://github.com/nushell/nushell/issues/7512, https://github.com/nushell/nushell/issues/5780, https://github.com/nushell/nushell/issues/7754 No effect: https://github.com/nushell/nushell/issues/8122 (we might revisit the completions code after this PR) Should use custom command instead: https://github.com/nushell/nushell/issues/6048 # User-Facing Changes Since aliases are now basically commands, it has some new implications: 1. `alias spam = "spam"` (requires command call) * **workaround**: use `alias spam = echo "spam"` 2. `def foo [] { 'foo' }; alias foo = ls -l` (foo defined more than once) * **workaround**: use different name (commands also have this limitation) 4. `alias ls = (ls | sort-by type name -i)` * **workaround**: Use custom command. _The common issue with this is that it is currently not easy to pass flags through custom commands and command referencing itself will lead to stack overflow. Both of these issues are meant to be addressed._ 5. TODO: Help messages, `which` command, `$nu.scope.aliases`, etc. * Should we treat the aliases as commands or should they be separated from regular commands? 6. Needs better error message and syntax highlight for recursed alias (`alias f = f`) 7. Can't create alias with the same name as existing command (`alias ls = ls -a`) * Might be possible to add support for it (not 100% sure) 8. Standalone `alias` doesn't list aliases anymore 9. Can't alias parser keywords (e.g., stuff like `alias ou = overlay use` won't work) * TODO: Needs a better error message when attempting to do so # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-02-27 08:44:05 +01:00
let contents = working_set.get_span_contents(span);
if contents.starts_with(b"$") || contents.starts_with(b"(") {
Allow spreading arguments to commands (#11289) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> Finishes implementing https://github.com/nushell/nushell/issues/10598, which asks for a spread operator in lists, in records, and when calling commands. # Description <!-- Thank you for improving Nushell. Please, check our [contributing guide](../CONTRIBUTING.md) and talk to the core team before making major changes. Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience. --> This PR will allow spreading arguments to commands (both internal and external). It will also deprecate spreading arguments automatically when passing to external commands. # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> - Users will be able to use `...` to spread arguments to custom/builtin commands that have rest parameters or allow unknown arguments, or to any external command - If a custom command doesn't have a rest parameter and it doesn't allow unknown arguments either, the spread operator will not be allowed - Passing lists to external commands without `...` will work for now but will cause a deprecation warning saying that it'll stop working in 0.91 (is 2 versions enough time?) Here's a function to help with demonstrating some behavior: ```nushell > def foo [ a, b, c?, d?, ...rest ] { [$a $b $c $d $rest] | to nuon } ``` You can pass a list of arguments to fill in the `rest` parameter using `...`: ```nushell > foo 1 2 3 4 ...[5 6] [1, 2, 3, 4, [5, 6]] ``` If you don't use `...`, the list `[5 6]` will be treated as a single argument: ```nushell > foo 1 2 3 4 [5 6] # Note the double [[]] [1, 2, 3, 4, [[5, 6]]] ``` You can omit optional parameters before the spread arguments: ```nushell > foo 1 2 3 ...[4 5] # d is omitted here [1, 2, 3, null, [4, 5]] ``` If you have multiple lists, you can spread them all: ```nushell > foo 1 2 3 ...[4 5] 6 7 ...[8] ...[] [1, 2, 3, null, [4, 5, 6, 7, 8]] ``` Here's the kind of error you get when you try to spread arguments to a command with no rest parameter: ![image](https://github.com/nushell/nushell/assets/45539777/93faceae-00eb-4e59-ac3f-17f98436e6e4) And this is the warning you get when you pass a list to an external now (without `...`): ![image](https://github.com/nushell/nushell/assets/45539777/d368f590-201e-49fb-8b20-68476ced415e) # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass (on Windows make sure to [enable developer mode](https://learn.microsoft.com/en-us/windows/apps/get-started/developer-mode-features-and-debugging)) - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> Added tests to cover the following cases: - Spreading arguments to a command that doesn't have a rest parameter (unexpected spread argument error) - Spreading arguments to a command that doesn't have a rest parameter *but* there's also a missing positional argument (missing positional error) - Spreading arguments to a command that doesn't have a rest parameter but does allow unknown arguments, such as `exec` (allowed) - Spreading a list literal containing arguments of the wrong type (parse error) - Spreading a non-list value, both to internal and external commands - Having named arguments in the middle of rest arguments - `explain`ing a command call that spreads its arguments # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> # Examples Suppose you have multiple tables: ```nushell let people = [[id name age]; [0 alice 100] [1 bob 200] [2 eve 300]] let evil_twins = [[id name age]; [0 ecila 100] [-1 bob 200] [-2 eve 300]] ``` Maybe you often find yourself needing to merge multiple tables and want a utility to do that. You could write a function like this: ```nushell def merge_all [ ...tables ] { $tables | reduce { |it, acc| $acc | merge $it } } ``` Then you can use it like this: ```nushell > merge_all ...([$people $evil_twins] | each { |$it| $it | select name age }) ╭───┬───────┬─────╮ │ # │ name │ age │ ├───┼───────┼─────┤ │ 0 │ ecila │ 100 │ │ 1 │ bob │ 200 │ │ 2 │ eve │ 300 │ ╰───┴───────┴─────╯ ``` Except they had duplicate columns, so now you first want to suffix every column with a number to tell you which table the column came from. You can make a command for that: ```nushell def select_and_merge [ --cols: list<string>, ...tables ] { let renamed_tables = $tables | enumerate | each { |it| $it.item | select $cols | rename ...($cols | each { |col| $col + ($it.index | into string) }) }; merge_all ...$renamed_tables } ``` And call it like this: ```nushell > select_and_merge --cols [name age] $people $evil_twins ╭───┬───────┬──────┬───────┬──────╮ │ # │ name0 │ age0 │ name1 │ age1 │ ├───┼───────┼──────┼───────┼──────┤ │ 0 │ alice │ 100 │ ecila │ 100 │ │ 1 │ bob │ 200 │ bob │ 200 │ │ 2 │ eve │ 300 │ eve │ 300 │ ╰───┴───────┴──────┴───────┴──────╯ ``` --- Suppose someone's made a command to search for APT packages: ```nushell # The main command def search-pkgs [ --install # Whether to install any packages it finds log_level: int # Pretend it's a good idea to make this a required positional parameter exclude?: list<string> # Packages to exclude repositories?: list<string> # Which repositories to look in (searches in all if not given) ...pkgs # Package names to search for ] { { install: $install, log_level: $log_level, exclude: ($exclude | to nuon), repositories: ($repositories | to nuon), pkgs: ($pkgs | to nuon) } } ``` It has a lot of parameters to configure it, so you might make your own helper commands to wrap around it for specific cases. Here's one example: ```nushell # Only look for packages locally def search-pkgs-local [ --install # Whether to install any packages it finds log_level: int exclude?: list<string> # Packages to exclude ...pkgs # Package names to search for ] { # All required and optional positional parameters are given search-pkgs --install=$install $log_level [] ["<local URI or something>"] ...$pkgs } ``` And you can run it like this: ```nushell > search-pkgs-local --install=false 5 ...["python2.7" "vim"] ╭──────────────┬──────────────────────────────╮ │ install │ false │ │ log_level │ 5 │ │ exclude │ [] │ │ repositories │ ["<local URI or something>"] │ │ pkgs │ ["python2.7", vim] │ ╰──────────────┴──────────────────────────────╯ ``` One thing I realized when writing this was that if we decide to not allow passing optional arguments using the spread operator, then you can (mis?)use the spread operator to skip optional parameters. Here, I didn't want to give `exclude` explicitly, so I used a spread operator to pass the packages to install. Without it, I would've needed to do `search-pkgs-local --install=false 5 [] "python2.7" "vim"` (explicitly pass `[]` (or `null`, in the general case) to `exclude`). There are probably more idiomatic ways to do this, but I just thought it was something interesting. If you're a virologist of the [xkcd](https://xkcd.com/350/) kind, another helper command you might make is this: ```nushell # Install any packages it finds def live-dangerously [ ...pkgs ] { # One optional argument was given (exclude), while another was not (repositories) search-pkgs 0 [] ...$pkgs --install # Flags can go after spread arguments } ``` Running it: ```nushell > live-dangerously "git" "*vi*" # *vi* because I don't feel like typing out vim and neovim ╭──────────────┬─────────────╮ │ install │ true │ │ log_level │ 0 │ │ exclude │ [] │ │ repositories │ null │ │ pkgs │ [git, *vi*] │ ╰──────────────┴─────────────╯ ``` Here's an example that uses the spread operator more than once within the same command call: ```nushell let extras = [ chrome firefox python java git ] def search-pkgs-curated [ ...pkgs ] { (search-pkgs 1 [emacs] ["example.com", "foo.com"] vim # A must for everyone! ...($pkgs | filter { |p| not ($p | str contains "*") }) # Remove packages with globs python # Good tool to have ...$extras --install=false python3) # I forget, did I already put Python in extras? } ``` Running it: ```nushell > search-pkgs-curated "git" "*vi*" ╭──────────────┬───────────────────────────────────────────────────────────────────╮ │ install │ false │ │ log_level │ 1 │ │ exclude │ [emacs] │ │ repositories │ [example.com, foo.com] │ │ pkgs │ [vim, git, python, chrome, firefox, python, java, git, "python3"] │ ╰──────────────┴───────────────────────────────────────────────────────────────────╯ ```
2023-12-28 08:43:20 +01:00
ExternalArgument::Regular(parse_dollar_expr(working_set, span))
Re-implement aliases (#8123) # Description This PR adds an alternative alias implementation. Old aliases still work but you need to use `old-alias` instead of `alias`. Instead of replacing spans in the original code and re-parsing, which proved to be extremely error-prone and a constant source of panics, the new implementation creates a new command that references the old command. Consider the new alias defined as `alias ll = ls -l`. The parser creates a new command called `ll` and remembers that it is actually a `ls` command called with the `-l` flag. Then, when the parser sees the `ll` command, it will translate it to `ls -l` and passes to it any parameters that were passed to the call to `ll`. It works quite similar to how known externals defined with `extern` are implemented. The new alias implementation should work the same way as the old aliases, including exporting from modules, referencing both known and unknown externals. It seems to preserve custom completions and pipeline metadata. It is quite robust in most cases but there are some rough edges (see later). Fixes https://github.com/nushell/nushell/issues/7648, https://github.com/nushell/nushell/issues/8026, https://github.com/nushell/nushell/issues/7512, https://github.com/nushell/nushell/issues/5780, https://github.com/nushell/nushell/issues/7754 No effect: https://github.com/nushell/nushell/issues/8122 (we might revisit the completions code after this PR) Should use custom command instead: https://github.com/nushell/nushell/issues/6048 # User-Facing Changes Since aliases are now basically commands, it has some new implications: 1. `alias spam = "spam"` (requires command call) * **workaround**: use `alias spam = echo "spam"` 2. `def foo [] { 'foo' }; alias foo = ls -l` (foo defined more than once) * **workaround**: use different name (commands also have this limitation) 4. `alias ls = (ls | sort-by type name -i)` * **workaround**: Use custom command. _The common issue with this is that it is currently not easy to pass flags through custom commands and command referencing itself will lead to stack overflow. Both of these issues are meant to be addressed._ 5. TODO: Help messages, `which` command, `$nu.scope.aliases`, etc. * Should we treat the aliases as commands or should they be separated from regular commands? 6. Needs better error message and syntax highlight for recursed alias (`alias f = f`) 7. Can't create alias with the same name as existing command (`alias ls = ls -a`) * Might be possible to add support for it (not 100% sure) 8. Standalone `alias` doesn't list aliases anymore 9. Can't alias parser keywords (e.g., stuff like `alias ou = overlay use` won't work) * TODO: Needs a better error message when attempting to do so # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-02-27 08:44:05 +01:00
} else if contents.starts_with(b"[") {
Allow spreading arguments to commands (#11289) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> Finishes implementing https://github.com/nushell/nushell/issues/10598, which asks for a spread operator in lists, in records, and when calling commands. # Description <!-- Thank you for improving Nushell. Please, check our [contributing guide](../CONTRIBUTING.md) and talk to the core team before making major changes. Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience. --> This PR will allow spreading arguments to commands (both internal and external). It will also deprecate spreading arguments automatically when passing to external commands. # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> - Users will be able to use `...` to spread arguments to custom/builtin commands that have rest parameters or allow unknown arguments, or to any external command - If a custom command doesn't have a rest parameter and it doesn't allow unknown arguments either, the spread operator will not be allowed - Passing lists to external commands without `...` will work for now but will cause a deprecation warning saying that it'll stop working in 0.91 (is 2 versions enough time?) Here's a function to help with demonstrating some behavior: ```nushell > def foo [ a, b, c?, d?, ...rest ] { [$a $b $c $d $rest] | to nuon } ``` You can pass a list of arguments to fill in the `rest` parameter using `...`: ```nushell > foo 1 2 3 4 ...[5 6] [1, 2, 3, 4, [5, 6]] ``` If you don't use `...`, the list `[5 6]` will be treated as a single argument: ```nushell > foo 1 2 3 4 [5 6] # Note the double [[]] [1, 2, 3, 4, [[5, 6]]] ``` You can omit optional parameters before the spread arguments: ```nushell > foo 1 2 3 ...[4 5] # d is omitted here [1, 2, 3, null, [4, 5]] ``` If you have multiple lists, you can spread them all: ```nushell > foo 1 2 3 ...[4 5] 6 7 ...[8] ...[] [1, 2, 3, null, [4, 5, 6, 7, 8]] ``` Here's the kind of error you get when you try to spread arguments to a command with no rest parameter: ![image](https://github.com/nushell/nushell/assets/45539777/93faceae-00eb-4e59-ac3f-17f98436e6e4) And this is the warning you get when you pass a list to an external now (without `...`): ![image](https://github.com/nushell/nushell/assets/45539777/d368f590-201e-49fb-8b20-68476ced415e) # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass (on Windows make sure to [enable developer mode](https://learn.microsoft.com/en-us/windows/apps/get-started/developer-mode-features-and-debugging)) - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> Added tests to cover the following cases: - Spreading arguments to a command that doesn't have a rest parameter (unexpected spread argument error) - Spreading arguments to a command that doesn't have a rest parameter *but* there's also a missing positional argument (missing positional error) - Spreading arguments to a command that doesn't have a rest parameter but does allow unknown arguments, such as `exec` (allowed) - Spreading a list literal containing arguments of the wrong type (parse error) - Spreading a non-list value, both to internal and external commands - Having named arguments in the middle of rest arguments - `explain`ing a command call that spreads its arguments # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> # Examples Suppose you have multiple tables: ```nushell let people = [[id name age]; [0 alice 100] [1 bob 200] [2 eve 300]] let evil_twins = [[id name age]; [0 ecila 100] [-1 bob 200] [-2 eve 300]] ``` Maybe you often find yourself needing to merge multiple tables and want a utility to do that. You could write a function like this: ```nushell def merge_all [ ...tables ] { $tables | reduce { |it, acc| $acc | merge $it } } ``` Then you can use it like this: ```nushell > merge_all ...([$people $evil_twins] | each { |$it| $it | select name age }) ╭───┬───────┬─────╮ │ # │ name │ age │ ├───┼───────┼─────┤ │ 0 │ ecila │ 100 │ │ 1 │ bob │ 200 │ │ 2 │ eve │ 300 │ ╰───┴───────┴─────╯ ``` Except they had duplicate columns, so now you first want to suffix every column with a number to tell you which table the column came from. You can make a command for that: ```nushell def select_and_merge [ --cols: list<string>, ...tables ] { let renamed_tables = $tables | enumerate | each { |it| $it.item | select $cols | rename ...($cols | each { |col| $col + ($it.index | into string) }) }; merge_all ...$renamed_tables } ``` And call it like this: ```nushell > select_and_merge --cols [name age] $people $evil_twins ╭───┬───────┬──────┬───────┬──────╮ │ # │ name0 │ age0 │ name1 │ age1 │ ├───┼───────┼──────┼───────┼──────┤ │ 0 │ alice │ 100 │ ecila │ 100 │ │ 1 │ bob │ 200 │ bob │ 200 │ │ 2 │ eve │ 300 │ eve │ 300 │ ╰───┴───────┴──────┴───────┴──────╯ ``` --- Suppose someone's made a command to search for APT packages: ```nushell # The main command def search-pkgs [ --install # Whether to install any packages it finds log_level: int # Pretend it's a good idea to make this a required positional parameter exclude?: list<string> # Packages to exclude repositories?: list<string> # Which repositories to look in (searches in all if not given) ...pkgs # Package names to search for ] { { install: $install, log_level: $log_level, exclude: ($exclude | to nuon), repositories: ($repositories | to nuon), pkgs: ($pkgs | to nuon) } } ``` It has a lot of parameters to configure it, so you might make your own helper commands to wrap around it for specific cases. Here's one example: ```nushell # Only look for packages locally def search-pkgs-local [ --install # Whether to install any packages it finds log_level: int exclude?: list<string> # Packages to exclude ...pkgs # Package names to search for ] { # All required and optional positional parameters are given search-pkgs --install=$install $log_level [] ["<local URI or something>"] ...$pkgs } ``` And you can run it like this: ```nushell > search-pkgs-local --install=false 5 ...["python2.7" "vim"] ╭──────────────┬──────────────────────────────╮ │ install │ false │ │ log_level │ 5 │ │ exclude │ [] │ │ repositories │ ["<local URI or something>"] │ │ pkgs │ ["python2.7", vim] │ ╰──────────────┴──────────────────────────────╯ ``` One thing I realized when writing this was that if we decide to not allow passing optional arguments using the spread operator, then you can (mis?)use the spread operator to skip optional parameters. Here, I didn't want to give `exclude` explicitly, so I used a spread operator to pass the packages to install. Without it, I would've needed to do `search-pkgs-local --install=false 5 [] "python2.7" "vim"` (explicitly pass `[]` (or `null`, in the general case) to `exclude`). There are probably more idiomatic ways to do this, but I just thought it was something interesting. If you're a virologist of the [xkcd](https://xkcd.com/350/) kind, another helper command you might make is this: ```nushell # Install any packages it finds def live-dangerously [ ...pkgs ] { # One optional argument was given (exclude), while another was not (repositories) search-pkgs 0 [] ...$pkgs --install # Flags can go after spread arguments } ``` Running it: ```nushell > live-dangerously "git" "*vi*" # *vi* because I don't feel like typing out vim and neovim ╭──────────────┬─────────────╮ │ install │ true │ │ log_level │ 0 │ │ exclude │ [] │ │ repositories │ null │ │ pkgs │ [git, *vi*] │ ╰──────────────┴─────────────╯ ``` Here's an example that uses the spread operator more than once within the same command call: ```nushell let extras = [ chrome firefox python java git ] def search-pkgs-curated [ ...pkgs ] { (search-pkgs 1 [emacs] ["example.com", "foo.com"] vim # A must for everyone! ...($pkgs | filter { |p| not ($p | str contains "*") }) # Remove packages with globs python # Good tool to have ...$extras --install=false python3) # I forget, did I already put Python in extras? } ``` Running it: ```nushell > search-pkgs-curated "git" "*vi*" ╭──────────────┬───────────────────────────────────────────────────────────────────╮ │ install │ false │ │ log_level │ 1 │ │ exclude │ [emacs] │ │ repositories │ [example.com, foo.com] │ │ pkgs │ [vim, git, python, chrome, firefox, python, java, git, "python3"] │ ╰──────────────┴───────────────────────────────────────────────────────────────────╯ ```
2023-12-28 08:43:20 +01:00
ExternalArgument::Regular(parse_list_expression(working_set, span, &SyntaxShape::Any))
} else if contents.starts_with(b"r#") {
ExternalArgument::Regular(parse_raw_string(working_set, span))
Allow spreading arguments to commands (#11289) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> Finishes implementing https://github.com/nushell/nushell/issues/10598, which asks for a spread operator in lists, in records, and when calling commands. # Description <!-- Thank you for improving Nushell. Please, check our [contributing guide](../CONTRIBUTING.md) and talk to the core team before making major changes. Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience. --> This PR will allow spreading arguments to commands (both internal and external). It will also deprecate spreading arguments automatically when passing to external commands. # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> - Users will be able to use `...` to spread arguments to custom/builtin commands that have rest parameters or allow unknown arguments, or to any external command - If a custom command doesn't have a rest parameter and it doesn't allow unknown arguments either, the spread operator will not be allowed - Passing lists to external commands without `...` will work for now but will cause a deprecation warning saying that it'll stop working in 0.91 (is 2 versions enough time?) Here's a function to help with demonstrating some behavior: ```nushell > def foo [ a, b, c?, d?, ...rest ] { [$a $b $c $d $rest] | to nuon } ``` You can pass a list of arguments to fill in the `rest` parameter using `...`: ```nushell > foo 1 2 3 4 ...[5 6] [1, 2, 3, 4, [5, 6]] ``` If you don't use `...`, the list `[5 6]` will be treated as a single argument: ```nushell > foo 1 2 3 4 [5 6] # Note the double [[]] [1, 2, 3, 4, [[5, 6]]] ``` You can omit optional parameters before the spread arguments: ```nushell > foo 1 2 3 ...[4 5] # d is omitted here [1, 2, 3, null, [4, 5]] ``` If you have multiple lists, you can spread them all: ```nushell > foo 1 2 3 ...[4 5] 6 7 ...[8] ...[] [1, 2, 3, null, [4, 5, 6, 7, 8]] ``` Here's the kind of error you get when you try to spread arguments to a command with no rest parameter: ![image](https://github.com/nushell/nushell/assets/45539777/93faceae-00eb-4e59-ac3f-17f98436e6e4) And this is the warning you get when you pass a list to an external now (without `...`): ![image](https://github.com/nushell/nushell/assets/45539777/d368f590-201e-49fb-8b20-68476ced415e) # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass (on Windows make sure to [enable developer mode](https://learn.microsoft.com/en-us/windows/apps/get-started/developer-mode-features-and-debugging)) - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> Added tests to cover the following cases: - Spreading arguments to a command that doesn't have a rest parameter (unexpected spread argument error) - Spreading arguments to a command that doesn't have a rest parameter *but* there's also a missing positional argument (missing positional error) - Spreading arguments to a command that doesn't have a rest parameter but does allow unknown arguments, such as `exec` (allowed) - Spreading a list literal containing arguments of the wrong type (parse error) - Spreading a non-list value, both to internal and external commands - Having named arguments in the middle of rest arguments - `explain`ing a command call that spreads its arguments # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> # Examples Suppose you have multiple tables: ```nushell let people = [[id name age]; [0 alice 100] [1 bob 200] [2 eve 300]] let evil_twins = [[id name age]; [0 ecila 100] [-1 bob 200] [-2 eve 300]] ``` Maybe you often find yourself needing to merge multiple tables and want a utility to do that. You could write a function like this: ```nushell def merge_all [ ...tables ] { $tables | reduce { |it, acc| $acc | merge $it } } ``` Then you can use it like this: ```nushell > merge_all ...([$people $evil_twins] | each { |$it| $it | select name age }) ╭───┬───────┬─────╮ │ # │ name │ age │ ├───┼───────┼─────┤ │ 0 │ ecila │ 100 │ │ 1 │ bob │ 200 │ │ 2 │ eve │ 300 │ ╰───┴───────┴─────╯ ``` Except they had duplicate columns, so now you first want to suffix every column with a number to tell you which table the column came from. You can make a command for that: ```nushell def select_and_merge [ --cols: list<string>, ...tables ] { let renamed_tables = $tables | enumerate | each { |it| $it.item | select $cols | rename ...($cols | each { |col| $col + ($it.index | into string) }) }; merge_all ...$renamed_tables } ``` And call it like this: ```nushell > select_and_merge --cols [name age] $people $evil_twins ╭───┬───────┬──────┬───────┬──────╮ │ # │ name0 │ age0 │ name1 │ age1 │ ├───┼───────┼──────┼───────┼──────┤ │ 0 │ alice │ 100 │ ecila │ 100 │ │ 1 │ bob │ 200 │ bob │ 200 │ │ 2 │ eve │ 300 │ eve │ 300 │ ╰───┴───────┴──────┴───────┴──────╯ ``` --- Suppose someone's made a command to search for APT packages: ```nushell # The main command def search-pkgs [ --install # Whether to install any packages it finds log_level: int # Pretend it's a good idea to make this a required positional parameter exclude?: list<string> # Packages to exclude repositories?: list<string> # Which repositories to look in (searches in all if not given) ...pkgs # Package names to search for ] { { install: $install, log_level: $log_level, exclude: ($exclude | to nuon), repositories: ($repositories | to nuon), pkgs: ($pkgs | to nuon) } } ``` It has a lot of parameters to configure it, so you might make your own helper commands to wrap around it for specific cases. Here's one example: ```nushell # Only look for packages locally def search-pkgs-local [ --install # Whether to install any packages it finds log_level: int exclude?: list<string> # Packages to exclude ...pkgs # Package names to search for ] { # All required and optional positional parameters are given search-pkgs --install=$install $log_level [] ["<local URI or something>"] ...$pkgs } ``` And you can run it like this: ```nushell > search-pkgs-local --install=false 5 ...["python2.7" "vim"] ╭──────────────┬──────────────────────────────╮ │ install │ false │ │ log_level │ 5 │ │ exclude │ [] │ │ repositories │ ["<local URI or something>"] │ │ pkgs │ ["python2.7", vim] │ ╰──────────────┴──────────────────────────────╯ ``` One thing I realized when writing this was that if we decide to not allow passing optional arguments using the spread operator, then you can (mis?)use the spread operator to skip optional parameters. Here, I didn't want to give `exclude` explicitly, so I used a spread operator to pass the packages to install. Without it, I would've needed to do `search-pkgs-local --install=false 5 [] "python2.7" "vim"` (explicitly pass `[]` (or `null`, in the general case) to `exclude`). There are probably more idiomatic ways to do this, but I just thought it was something interesting. If you're a virologist of the [xkcd](https://xkcd.com/350/) kind, another helper command you might make is this: ```nushell # Install any packages it finds def live-dangerously [ ...pkgs ] { # One optional argument was given (exclude), while another was not (repositories) search-pkgs 0 [] ...$pkgs --install # Flags can go after spread arguments } ``` Running it: ```nushell > live-dangerously "git" "*vi*" # *vi* because I don't feel like typing out vim and neovim ╭──────────────┬─────────────╮ │ install │ true │ │ log_level │ 0 │ │ exclude │ [] │ │ repositories │ null │ │ pkgs │ [git, *vi*] │ ╰──────────────┴─────────────╯ ``` Here's an example that uses the spread operator more than once within the same command call: ```nushell let extras = [ chrome firefox python java git ] def search-pkgs-curated [ ...pkgs ] { (search-pkgs 1 [emacs] ["example.com", "foo.com"] vim # A must for everyone! ...($pkgs | filter { |p| not ($p | str contains "*") }) # Remove packages with globs python # Good tool to have ...$extras --install=false python3) # I forget, did I already put Python in extras? } ``` Running it: ```nushell > search-pkgs-curated "git" "*vi*" ╭──────────────┬───────────────────────────────────────────────────────────────────╮ │ install │ false │ │ log_level │ 1 │ │ exclude │ [emacs] │ │ repositories │ [example.com, foo.com] │ │ pkgs │ [vim, git, python, chrome, firefox, python, java, git, "python3"] │ ╰──────────────┴───────────────────────────────────────────────────────────────────╯ ```
2023-12-28 08:43:20 +01:00
} else if contents.len() > 3
&& contents.starts_with(b"...")
&& (contents[3] == b'$' || contents[3] == b'[' || contents[3] == b'(')
{
ExternalArgument::Spread(parse_value(
working_set,
Span::new(span.start + 3, span.end),
&SyntaxShape::List(Box::new(SyntaxShape::Any)),
))
Re-implement aliases (#8123) # Description This PR adds an alternative alias implementation. Old aliases still work but you need to use `old-alias` instead of `alias`. Instead of replacing spans in the original code and re-parsing, which proved to be extremely error-prone and a constant source of panics, the new implementation creates a new command that references the old command. Consider the new alias defined as `alias ll = ls -l`. The parser creates a new command called `ll` and remembers that it is actually a `ls` command called with the `-l` flag. Then, when the parser sees the `ll` command, it will translate it to `ls -l` and passes to it any parameters that were passed to the call to `ll`. It works quite similar to how known externals defined with `extern` are implemented. The new alias implementation should work the same way as the old aliases, including exporting from modules, referencing both known and unknown externals. It seems to preserve custom completions and pipeline metadata. It is quite robust in most cases but there are some rough edges (see later). Fixes https://github.com/nushell/nushell/issues/7648, https://github.com/nushell/nushell/issues/8026, https://github.com/nushell/nushell/issues/7512, https://github.com/nushell/nushell/issues/5780, https://github.com/nushell/nushell/issues/7754 No effect: https://github.com/nushell/nushell/issues/8122 (we might revisit the completions code after this PR) Should use custom command instead: https://github.com/nushell/nushell/issues/6048 # User-Facing Changes Since aliases are now basically commands, it has some new implications: 1. `alias spam = "spam"` (requires command call) * **workaround**: use `alias spam = echo "spam"` 2. `def foo [] { 'foo' }; alias foo = ls -l` (foo defined more than once) * **workaround**: use different name (commands also have this limitation) 4. `alias ls = (ls | sort-by type name -i)` * **workaround**: Use custom command. _The common issue with this is that it is currently not easy to pass flags through custom commands and command referencing itself will lead to stack overflow. Both of these issues are meant to be addressed._ 5. TODO: Help messages, `which` command, `$nu.scope.aliases`, etc. * Should we treat the aliases as commands or should they be separated from regular commands? 6. Needs better error message and syntax highlight for recursed alias (`alias f = f`) 7. Can't create alias with the same name as existing command (`alias ls = ls -a`) * Might be possible to add support for it (not 100% sure) 8. Standalone `alias` doesn't list aliases anymore 9. Can't alias parser keywords (e.g., stuff like `alias ou = overlay use` won't work) * TODO: Needs a better error message when attempting to do so # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-02-27 08:44:05 +01:00
} else {
// Eval stage trims the quotes, so we don't have to do the same thing when parsing.
let (contents, err) = unescape_string_preserving_quotes(contents, span);
if let Some(err) = err {
working_set.error(err);
}
Re-implement aliases (#8123) # Description This PR adds an alternative alias implementation. Old aliases still work but you need to use `old-alias` instead of `alias`. Instead of replacing spans in the original code and re-parsing, which proved to be extremely error-prone and a constant source of panics, the new implementation creates a new command that references the old command. Consider the new alias defined as `alias ll = ls -l`. The parser creates a new command called `ll` and remembers that it is actually a `ls` command called with the `-l` flag. Then, when the parser sees the `ll` command, it will translate it to `ls -l` and passes to it any parameters that were passed to the call to `ll`. It works quite similar to how known externals defined with `extern` are implemented. The new alias implementation should work the same way as the old aliases, including exporting from modules, referencing both known and unknown externals. It seems to preserve custom completions and pipeline metadata. It is quite robust in most cases but there are some rough edges (see later). Fixes https://github.com/nushell/nushell/issues/7648, https://github.com/nushell/nushell/issues/8026, https://github.com/nushell/nushell/issues/7512, https://github.com/nushell/nushell/issues/5780, https://github.com/nushell/nushell/issues/7754 No effect: https://github.com/nushell/nushell/issues/8122 (we might revisit the completions code after this PR) Should use custom command instead: https://github.com/nushell/nushell/issues/6048 # User-Facing Changes Since aliases are now basically commands, it has some new implications: 1. `alias spam = "spam"` (requires command call) * **workaround**: use `alias spam = echo "spam"` 2. `def foo [] { 'foo' }; alias foo = ls -l` (foo defined more than once) * **workaround**: use different name (commands also have this limitation) 4. `alias ls = (ls | sort-by type name -i)` * **workaround**: Use custom command. _The common issue with this is that it is currently not easy to pass flags through custom commands and command referencing itself will lead to stack overflow. Both of these issues are meant to be addressed._ 5. TODO: Help messages, `which` command, `$nu.scope.aliases`, etc. * Should we treat the aliases as commands or should they be separated from regular commands? 6. Needs better error message and syntax highlight for recursed alias (`alias f = f`) 7. Can't create alias with the same name as existing command (`alias ls = ls -a`) * Might be possible to add support for it (not 100% sure) 8. Standalone `alias` doesn't list aliases anymore 9. Can't alias parser keywords (e.g., stuff like `alias ou = overlay use` won't work) * TODO: Needs a better error message when attempting to do so # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-02-27 08:44:05 +01:00
ExternalArgument::Regular(Expression::new(
working_set,
Expr::String(contents),
span,
Type::String,
))
Re-implement aliases (#8123) # Description This PR adds an alternative alias implementation. Old aliases still work but you need to use `old-alias` instead of `alias`. Instead of replacing spans in the original code and re-parsing, which proved to be extremely error-prone and a constant source of panics, the new implementation creates a new command that references the old command. Consider the new alias defined as `alias ll = ls -l`. The parser creates a new command called `ll` and remembers that it is actually a `ls` command called with the `-l` flag. Then, when the parser sees the `ll` command, it will translate it to `ls -l` and passes to it any parameters that were passed to the call to `ll`. It works quite similar to how known externals defined with `extern` are implemented. The new alias implementation should work the same way as the old aliases, including exporting from modules, referencing both known and unknown externals. It seems to preserve custom completions and pipeline metadata. It is quite robust in most cases but there are some rough edges (see later). Fixes https://github.com/nushell/nushell/issues/7648, https://github.com/nushell/nushell/issues/8026, https://github.com/nushell/nushell/issues/7512, https://github.com/nushell/nushell/issues/5780, https://github.com/nushell/nushell/issues/7754 No effect: https://github.com/nushell/nushell/issues/8122 (we might revisit the completions code after this PR) Should use custom command instead: https://github.com/nushell/nushell/issues/6048 # User-Facing Changes Since aliases are now basically commands, it has some new implications: 1. `alias spam = "spam"` (requires command call) * **workaround**: use `alias spam = echo "spam"` 2. `def foo [] { 'foo' }; alias foo = ls -l` (foo defined more than once) * **workaround**: use different name (commands also have this limitation) 4. `alias ls = (ls | sort-by type name -i)` * **workaround**: Use custom command. _The common issue with this is that it is currently not easy to pass flags through custom commands and command referencing itself will lead to stack overflow. Both of these issues are meant to be addressed._ 5. TODO: Help messages, `which` command, `$nu.scope.aliases`, etc. * Should we treat the aliases as commands or should they be separated from regular commands? 6. Needs better error message and syntax highlight for recursed alias (`alias f = f`) 7. Can't create alias with the same name as existing command (`alias ls = ls -a`) * Might be possible to add support for it (not 100% sure) 8. Standalone `alias` doesn't list aliases anymore 9. Can't alias parser keywords (e.g., stuff like `alias ou = overlay use` won't work) * TODO: Needs a better error message when attempting to do so # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-02-27 08:44:05 +01:00
}
}
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
pub fn parse_external_call(working_set: &mut StateWorkingSet, spans: &[Span]) -> Expression {
trace!("parse external");
let head_contents = working_set.get_span_contents(spans[0]);
let head_span = if head_contents.starts_with(b"^") {
Protocol: debug_assert!() Span to reflect a valid slice (#6806) Also enforce this by #[non_exhaustive] span such that going forward we cannot, in debug builds (1), construct invalid spans. The motivation for this stems from #6431 where I've seen crashes due to invalid slice indexing. My hope is this will mitigate such senarios 1. https://github.com/nushell/nushell/pull/6431#issuecomment-1278147241 # Description (description of your pull request here) # Tests Make sure you've done the following: - [ ] Add tests that cover your changes, either in the command examples, the crate/tests folder, or in the /tests folder. - [ ] Try to think about corner cases and various ways how your changes could break. Cover them with tests. - [ ] If adding tests is not possible, please document in the PR body a minimal example with steps on how to reproduce so one can verify your change works. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [ ] `cargo clippy --workspace --features=extra -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [ ] `cargo test --workspace --features=extra` to check that all the tests pass # Documentation - [ ] If your PR touches a user-facing nushell feature then make sure that there is an entry in the documentation (https://github.com/nushell/nushell.github.io) for the feature, and update it if necessary.
2022-12-03 10:44:12 +01:00
Span::new(spans[0].start + 1, spans[0].end)
} else {
spans[0]
};
2022-03-03 20:05:55 +01:00
let head_contents = working_set.get_span_contents(head_span).to_vec();
let head = if head_contents.starts_with(b"$") || head_contents.starts_with(b"(") {
Make external command substitution works friendly(like fish shell, trailing ending newlines) (#7156) # Description As title, when execute external sub command, auto-trimming end new-lines, like how fish shell does. And if the command is executed directly like: `cat tmp`, the result won't change. Fixes: #6816 Fixes: #3980 Note that although nushell works correctly by directly replace output of external command to variable(or other places like string interpolation), it's not friendly to user, and users almost want to use `str trim` to trim trailing newline, I think that's why fish shell do this automatically. If the pr is ok, as a result, no more `str trim -r` is required when user is writing scripts which using external commands. # User-Facing Changes Before: <img width="523" alt="img" src="https://user-images.githubusercontent.com/22256154/202468810-86b04dbb-c147-459a-96a5-e0095eeaab3d.png"> After: <img width="505" alt="img" src="https://user-images.githubusercontent.com/22256154/202468599-7b537488-3d6b-458e-9d75-d85780826db0.png"> # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace --features=extra -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace --features=extra` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2022-11-23 04:51:57 +01:00
// the expression is inside external_call, so it's a subexpression
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
let arg = parse_expression(working_set, &[head_span]);
Box::new(arg)
} else {
Restore tilde expansion on external command names (#13001) # Description Fix a regression introduced by #12921, where tilde expansion was no longer done on the external command name, breaking things like ```nushell > ~/.cargo/bin/exa ``` This properly handles quoted strings, so they don't expand: ```nushell > ^"~/.cargo/bin/exa" Error: nu::shell::external_command × External command failed ╭─[entry #1:1:2] 1 │ ^"~/.cargo/bin/exa" · ─────────┬──────── · ╰── Command `~/.cargo/bin/exa` not found ╰──── help: `~/.cargo/bin/exa` is neither a Nushell built-in or a known external command ``` This required a change to the parser, so the command name is also parsed in the same way the arguments are - i.e. the quotes on the outside remain in the expression. Hopefully that doesn't break anything else. 🤞 Fixes #13000. Should include in patch release 0.94.1 cc @YizhePKU # User-Facing Changes - Tilde expansion now works again for external commands - The `command` of `run-external` will now have its quotes removed like the other arguments if it is a literal string - The parser is changed to include quotes in the command expression of `ExternalCall` if they were present # Tests + Formatting I would like to add a regression test for this, but it's complicated because we need a well-known binary within the home directory, which just isn't a thing. We could drop one there, but that's kind of a bad behavior for a test to do. I also considered changing the home directory for the test, but that's so platform-specific - potentially could get it working on specific platforms though. Changing `HOME` env on Linux definitely works as far as tilde expansion works. - :green_circle: `toolkit fmt` - :green_circle: `toolkit clippy` - :green_circle: `toolkit test` - :green_circle: `toolkit test stdlib`
2024-05-30 03:48:29 +02:00
// Eval stage will unquote the string, so we don't bother with that here
let (contents, err) = unescape_string_preserving_quotes(&head_contents, head_span);
if let Some(err) = err {
working_set.error(err)
}
Box::new(Expression::new(
working_set,
Expr::String(contents),
head_span,
Type::String,
))
};
let args = spans[1..]
.iter()
.map(|&span| parse_external_arg(working_set, span))
.collect();
Expression::new(
working_set,
Expr::ExternalCall(head, args),
Span::concat(spans),
Type::Any,
)
2021-09-02 10:25:22 +02:00
}
2021-07-01 02:01:04 +02:00
fn ensure_flag_arg_type(
working_set: &mut StateWorkingSet,
arg_name: String,
arg: Expression,
arg_shape: &SyntaxShape,
long_name_span: Span,
) -> (Spanned<String>, Expression) {
if !type_compatible(&arg.ty, &arg_shape.to_type()) {
working_set.error(ParseError::TypeMismatch(
arg_shape.to_type(),
arg.ty,
arg.span,
));
(
Spanned {
item: arg_name,
span: long_name_span,
},
Expression::garbage(working_set, arg.span),
)
} else {
(
Spanned {
item: arg_name,
span: long_name_span,
},
arg,
)
}
}
2021-09-02 10:25:22 +02:00
fn parse_long_flag(
working_set: &mut StateWorkingSet,
spans: &[Span],
spans_idx: &mut usize,
sig: &Signature,
) -> (Option<Spanned<String>>, Option<Expression>) {
2021-09-02 10:25:22 +02:00
let arg_span = spans[*spans_idx];
let arg_contents = working_set.get_span_contents(arg_span);
if arg_contents.starts_with(b"--") {
2021-10-12 19:44:23 +02:00
// FIXME: only use the first flag you find?
2021-09-02 10:25:22 +02:00
let split: Vec<_> = arg_contents.split(|x| *x == b'=').collect();
let long_name = String::from_utf8(split[0].into());
if let Ok(long_name) = long_name {
2021-10-13 19:53:27 +02:00
let long_name = long_name[2..].to_string();
2021-09-02 10:25:22 +02:00
if let Some(flag) = sig.get_long_flag(&long_name) {
if let Some(arg_shape) = &flag.arg {
if split.len() > 1 {
// and we also have the argument
let long_name_len = long_name.len();
2021-09-02 10:25:22 +02:00
let mut span = arg_span;
span.start += long_name_len + 3; //offset by long flag and '='
let arg = parse_value(working_set, span, arg_shape);
let (arg_name, val_expression) = ensure_flag_arg_type(
working_set,
long_name,
arg,
arg_shape,
Span::new(arg_span.start, arg_span.start + long_name_len + 2),
);
(Some(arg_name), Some(val_expression))
2021-09-02 10:25:22 +02:00
} else if let Some(arg) = spans.get(*spans_idx + 1) {
let arg = parse_value(working_set, *arg, arg_shape);
2021-09-02 10:25:22 +02:00
*spans_idx += 1;
let (arg_name, val_expression) =
ensure_flag_arg_type(working_set, long_name, arg, arg_shape, arg_span);
(Some(arg_name), Some(val_expression))
2021-07-08 22:29:00 +02:00
} else {
working_set.error(ParseError::MissingFlagParam(
arg_shape.to_string(),
arg_span,
));
2021-09-02 10:25:22 +02:00
(
Some(Spanned {
item: long_name,
span: arg_span,
}),
2021-09-02 10:25:22 +02:00
None,
)
2021-07-08 22:29:00 +02:00
}
} else {
2021-09-02 10:25:22 +02:00
// A flag with no argument
// It can also takes a boolean value like --x=true
if split.len() > 1 {
// and we also have the argument
let long_name_len = long_name.len();
let mut span = arg_span;
span.start += long_name_len + 3; //offset by long flag and '='
let arg = parse_value(working_set, span, &SyntaxShape::Boolean);
let (arg_name, val_expression) = ensure_flag_arg_type(
working_set,
long_name,
arg,
&SyntaxShape::Boolean,
Span::new(arg_span.start, arg_span.start + long_name_len + 2),
);
(Some(arg_name), Some(val_expression))
} else {
(
Some(Spanned {
item: long_name,
span: arg_span,
}),
None,
)
}
2021-07-08 22:29:00 +02:00
}
} else {
working_set.error(ParseError::UnknownFlag(
sig.name.clone(),
long_name.clone(),
arg_span,
sig.clone().formatted_flags(),
));
2021-09-02 10:25:22 +02:00
(
Some(Spanned {
item: long_name.clone(),
span: arg_span,
}),
2021-09-02 10:25:22 +02:00
None,
)
2021-07-08 22:29:00 +02:00
}
} else {
working_set.error(ParseError::NonUtf8(arg_span));
(
Some(Spanned {
item: "--".into(),
span: arg_span,
}),
None,
)
2021-07-08 22:29:00 +02:00
}
2021-09-02 10:25:22 +02:00
} else {
(None, None)
2021-07-08 22:29:00 +02:00
}
2021-09-02 10:25:22 +02:00
}
2021-07-08 22:29:00 +02:00
2021-09-02 10:25:22 +02:00
fn parse_short_flags(
working_set: &mut StateWorkingSet,
spans: &[Span],
spans_idx: &mut usize,
positional_idx: usize,
sig: &Signature,
) -> Option<Vec<Flag>> {
2021-09-02 10:25:22 +02:00
let arg_span = spans[*spans_idx];
let arg_contents = working_set.get_span_contents(arg_span);
if let Ok(arg_contents_uft8_ref) = str::from_utf8(arg_contents) {
if arg_contents_uft8_ref.starts_with('-') && arg_contents_uft8_ref.len() > 1 {
let short_flags = &arg_contents_uft8_ref[1..];
let num_chars = short_flags.chars().count();
let mut found_short_flags = vec![];
let mut unmatched_short_flags = vec![];
for (offset, short_flag) in short_flags.char_indices() {
let short_flag_span = Span::new(
arg_span.start + 1 + offset,
arg_span.start + 1 + offset + short_flag.len_utf8(),
);
if let Some(flag) = sig.get_short_flag(short_flag) {
// Allow args in short flag batches as long as it is the last flag.
if flag.arg.is_some() && offset < num_chars - 1 {
working_set
.error(ParseError::OnlyLastFlagInBatchCanTakeArg(short_flag_span));
break;
}
found_short_flags.push(flag);
} else {
unmatched_short_flags.push(short_flag_span);
2021-07-08 22:29:00 +02:00
}
}
if found_short_flags.is_empty()
// check to see if we have a negative number
&& matches!(
sig.get_positional(positional_idx),
Some(PositionalArg {
shape: SyntaxShape::Int | SyntaxShape::Number | SyntaxShape::Float,
..
})
)
&& String::from_utf8_lossy(working_set.get_span_contents(arg_span))
.parse::<f64>()
.is_ok()
{
return None;
} else if let Some(first) = unmatched_short_flags.first() {
let contents = working_set.get_span_contents(*first);
working_set.error(ParseError::UnknownFlag(
sig.name.clone(),
format!("-{}", String::from_utf8_lossy(contents)),
*first,
sig.clone().formatted_flags(),
));
2021-07-08 22:29:00 +02:00
}
2021-09-02 10:25:22 +02:00
Some(found_short_flags)
} else {
None
}
2021-09-02 10:25:22 +02:00
} else {
working_set.error(ParseError::NonUtf8(arg_span));
None
2021-07-08 22:29:00 +02:00
}
2021-09-02 10:25:22 +02:00
}
2021-07-08 22:29:00 +02:00
2021-09-02 10:25:22 +02:00
fn first_kw_idx(
working_set: &StateWorkingSet,
signature: &Signature,
spans: &[Span],
spans_idx: usize,
positional_idx: usize,
) -> (Option<usize>, usize) {
for idx in (positional_idx + 1)..signature.num_positionals() {
if let Some(PositionalArg {
shape: SyntaxShape::Keyword(kw, ..),
..
}) = signature.get_positional(idx)
{
for (span_idx, &span) in spans.iter().enumerate().skip(spans_idx) {
let contents = working_set.get_span_contents(span);
2021-08-26 23:48:27 +02:00
2021-09-02 10:25:22 +02:00
if contents == kw {
return (Some(idx), span_idx);
2021-08-26 23:48:27 +02:00
}
}
}
}
2021-09-02 10:25:22 +02:00
(None, spans.len())
}
2021-08-26 23:48:27 +02:00
2021-09-02 10:25:22 +02:00
fn calculate_end_span(
working_set: &StateWorkingSet,
signature: &Signature,
spans: &[Span],
spans_idx: usize,
positional_idx: usize,
) -> usize {
if signature.rest_positional.is_some() {
spans.len()
} else {
let (kw_pos, kw_idx) =
first_kw_idx(working_set, signature, spans, spans_idx, positional_idx);
2021-08-26 23:48:27 +02:00
2021-09-02 10:25:22 +02:00
if let Some(kw_pos) = kw_pos {
// We found a keyword. Keywords, once found, create a guidepost to
// show us where the positionals will lay into the arguments. Because they're
// keywords, they get to set this by being present
2021-08-26 23:48:27 +02:00
2021-09-02 10:25:22 +02:00
let positionals_between = kw_pos - positional_idx - 1;
if positionals_between > (kw_idx - spans_idx) {
kw_idx
2021-08-26 23:48:27 +02:00
} else {
2021-09-02 10:25:22 +02:00
kw_idx - positionals_between
}
} else {
// Make space for the remaining require positionals, if we can
if signature.num_positionals_after(positional_idx) == 0 {
spans.len()
} else if positional_idx < signature.required_positional.len()
2021-09-02 10:25:22 +02:00
&& spans.len() > (signature.required_positional.len() - positional_idx)
{
spans.len() - (signature.required_positional.len() - positional_idx - 1)
} else {
2021-09-04 09:59:38 +02:00
spans_idx + 1
2021-07-24 07:57:17 +02:00
}
}
}
2021-09-02 10:25:22 +02:00
}
2021-07-24 07:57:17 +02:00
pub fn parse_multispan_value(
2021-09-02 10:25:22 +02:00
working_set: &mut StateWorkingSet,
spans: &[Span],
spans_idx: &mut usize,
shape: &SyntaxShape,
) -> Expression {
add raw-string literal support (#9956) # Description This PR adds raw string support by using `r#` at the beginning of single quoted strings and `#` at the end. Notice that escapes do not process, even within single quotes, parentheses don't mean anything, $variables don't mean anything. It's just a string. ```nushell ❯ echo r#'one\ntwo (blah) ($var)'# one\ntwo (blah) ($var) ``` Notice how they work without `echo` or `print` and how they work without carriage returns. ```nushell ❯ r#'adsfa'# adsfa ❯ r##"asdfa'@qpejq'## asdfa'@qpejq ❯ r#'asdfasdfasf ∙ foqwejfqo@'23rfjqf'# ``` They also have a special configurable color in the repl. (use single quotes though) ![image](https://github.com/nushell/nushell/assets/343840/8780e21d-de4c-45b3-9880-2425f5fe10ef) They should work like rust raw literals and allow `r##`, `r###`, `r####`, etc, to help with having one or many `#`'s in the middle of your raw-string. They should work with `let` as well. ```nushell r#'some\nraw\nstring'# | str upcase ``` closes https://github.com/nushell/nushell/issues/5091 # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: WindSoilder <WindSoilder@outlook.com> Co-authored-by: Ian Manske <ian.manske@pm.me>
2024-05-02 15:36:37 +02:00
trace!("parse multispan value");
2021-09-02 10:25:22 +02:00
match shape {
SyntaxShape::VarWithOptType => {
2022-01-01 22:42:50 +01:00
trace!("parsing: var with opt type");
Improve type hovers (#9515) # Description This PR does a few things to help improve type hovers and, in the process, fixes a few outstanding issues in the type system. Here's a list of the changes: * `for` now will try to infer the type of the iteration variable based on the expression it's given. This fixes things like `for x in [1, 2, 3] { }` where `x` now properly gets the int type. * Removed old input/output type fields from the signature, focuses on the vec of signatures. Updated a bunch of dataframe commands that hadn't moved over. This helps tie things together a bit better * Fixed inference of types from subexpressions to use the last expression in the block * Fixed handling of explicit types in `let` and `mut` calls, so we now respect that as the authoritative type I also tried to add `def` input/output type inference, but unfortunately we only know the predecl types universally, which means we won't have enough information to properly know what the types of the custom commands are. # User-Facing Changes Script typechecking will get tighter in some cases Hovers should be more accurate in some cases that previously resorted to any. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: Darren Schroeder <343840+fdncred@users.noreply.github.com>
2023-06-28 19:19:48 +02:00
parse_var_with_opt_type(working_set, spans, spans_idx, false).0
2021-09-02 10:25:22 +02:00
}
SyntaxShape::RowCondition => {
2022-01-01 22:42:50 +01:00
trace!("parsing: row condition");
let arg = parse_row_condition(working_set, &spans[*spans_idx..]);
2021-09-02 10:25:22 +02:00
*spans_idx = spans.len() - 1;
2021-07-16 08:24:46 +02:00
arg
2021-09-02 10:25:22 +02:00
}
SyntaxShape::MathExpression => {
trace!("parsing: math expression");
let arg = parse_math_expression(working_set, &spans[*spans_idx..], None);
*spans_idx = spans.len() - 1;
arg
}
SyntaxShape::OneOf(shapes) => {
// handle for `if` command.
//let block_then_exp = shapes.as_slice() == [SyntaxShape::Block, SyntaxShape::Expression];
for shape in shapes.iter() {
let starting_error_count = working_set.parse_errors.len();
let s = parse_multispan_value(working_set, spans, spans_idx, shape);
if starting_error_count == working_set.parse_errors.len() {
return s;
} else if let Some(
ParseError::Expected(..) | ParseError::ExpectedWithStringMsg(..),
) = working_set.parse_errors.last()
{
working_set.parse_errors.truncate(starting_error_count);
continue;
}
// `if` is parsing block first and then expression.
// when we're writing something like `else if $a`, parsing as a
// block will result to error(because it's not a block)
//
// If parse as a expression also failed, user is more likely concerned
// about expression failure rather than "expect block failure"".
// FIXME FIXME FIXME
// if block_then_exp {
// match &err {
// Some(ParseError::Expected(expected, _)) => {
// if expected.starts_with("block") {
// err = e
// }
// }
// _ => err = err.or(e),
// }
// } else {
// err = err.or(e)
// }
}
let span = spans[*spans_idx];
if working_set.parse_errors.is_empty() {
working_set.error(ParseError::ExpectedWithStringMsg(
format!("one of a list of accepted shapes: {shapes:?}"),
span,
));
}
Expression::garbage(working_set, span)
}
2021-09-02 10:25:22 +02:00
SyntaxShape::Expression => {
2022-01-01 22:42:50 +01:00
trace!("parsing: expression");
Make external command substitution works friendly(like fish shell, trailing ending newlines) (#7156) # Description As title, when execute external sub command, auto-trimming end new-lines, like how fish shell does. And if the command is executed directly like: `cat tmp`, the result won't change. Fixes: #6816 Fixes: #3980 Note that although nushell works correctly by directly replace output of external command to variable(or other places like string interpolation), it's not friendly to user, and users almost want to use `str trim` to trim trailing newline, I think that's why fish shell do this automatically. If the pr is ok, as a result, no more `str trim -r` is required when user is writing scripts which using external commands. # User-Facing Changes Before: <img width="523" alt="img" src="https://user-images.githubusercontent.com/22256154/202468810-86b04dbb-c147-459a-96a5-e0095eeaab3d.png"> After: <img width="505" alt="img" src="https://user-images.githubusercontent.com/22256154/202468599-7b537488-3d6b-458e-9d75-d85780826db0.png"> # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace --features=extra -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace --features=extra` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2022-11-23 04:51:57 +01:00
// is it subexpression?
// Not sure, but let's make it not, so the behavior is the same as previous version of nushell.
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
let arg = parse_expression(working_set, &spans[*spans_idx..]);
2021-09-02 10:25:22 +02:00
*spans_idx = spans.len() - 1;
2021-07-08 23:16:25 +02:00
arg
2021-09-02 10:25:22 +02:00
}
Custom command input/output types (#9690) # Description This adds input/output types to custom commands. These are input/output pairs that related an input type to an output type. For example (a single int-to-int input/output pair): ``` def foo []: int -> int { ... } ``` You can also have multiple input/output pairs: ``` def bar []: [int -> string, string -> list<string>] { ... } ``` These types are checked during definition time in the parser. If the block does not match the type, the user will get a parser error. This `:` to begin the input/output signatures should immediately follow the argument signature as shown above. The PR also improves type parsing by re-using the shape parser. The shape parser is now the canonical way to parse types/shapes in user code. This PR also splits `extern` into `extern`/`extern-wrapped` because of the parser limitation that a multi-span argument (which Signature now is) can't precede an optional argument. `extern-wrapped` now takes the required block that was previously optional. # User-Facing Changes The change to `extern` to split into `extern` and `extern-wrapped` is a breaking change. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-14 23:51:28 +02:00
SyntaxShape::Signature => {
trace!("parsing: signature");
let sig = parse_full_signature(working_set, &spans[*spans_idx..]);
*spans_idx = spans.len() - 1;
sig
}
2021-09-02 10:25:22 +02:00
SyntaxShape::Keyword(keyword, arg) => {
2022-01-01 22:42:50 +01:00
trace!(
"parsing: keyword({}) {:?}",
String::from_utf8_lossy(keyword),
arg
);
2021-09-02 10:25:22 +02:00
let arg_span = spans[*spans_idx];
2021-07-08 23:16:25 +02:00
2021-09-02 10:25:22 +02:00
let arg_contents = working_set.get_span_contents(arg_span);
2021-07-17 07:28:25 +02:00
2021-09-02 10:25:22 +02:00
if arg_contents != keyword {
// When keywords mismatch, this is a strong indicator of something going wrong.
// We won't often override the current error, but as this is a strong indicator
// go ahead and override the current error and tell the user about the missing
// keyword/literal.
working_set.error(ParseError::ExpectedKeyword(
2021-09-02 10:25:22 +02:00
String::from_utf8_lossy(keyword).into(),
arg_span,
))
}
2021-07-17 07:28:25 +02:00
2021-09-02 10:25:22 +02:00
*spans_idx += 1;
if *spans_idx >= spans.len() {
working_set.error(ParseError::KeywordMissingArgument(
arg.to_string(),
String::from_utf8_lossy(keyword).into(),
Span::new(spans[*spans_idx - 1].end, spans[*spans_idx - 1].end),
));
let keyword = Keyword {
keyword: keyword.as_slice().into(),
span: spans[*spans_idx - 1],
expr: Expression::garbage(working_set, arg_span),
};
return Expression::new(
working_set,
Expr::Keyword(Box::new(keyword)),
arg_span,
Type::Any,
);
2021-07-08 23:16:25 +02:00
}
let keyword = Keyword {
keyword: keyword.as_slice().into(),
span: spans[*spans_idx - 1],
expr: parse_multispan_value(working_set, spans, spans_idx, arg),
};
2021-07-17 07:28:25 +02:00
Expression::new(
working_set,
Expr::Keyword(Box::new(keyword.clone())),
arg_span,
keyword.expr.ty,
)
2021-09-02 10:25:22 +02:00
}
_ => {
// All other cases are single-span values
let arg_span = spans[*spans_idx];
2021-07-08 23:16:25 +02:00
parse_value(working_set, arg_span, shape)
2021-07-08 23:16:25 +02:00
}
}
2021-09-02 10:25:22 +02:00
}
2021-07-08 23:16:25 +02:00
pub struct ParsedInternalCall {
pub call: Box<Call>,
pub output: Type,
}
2021-09-02 10:25:22 +02:00
pub fn parse_internal_call(
working_set: &mut StateWorkingSet,
command_span: Span,
spans: &[Span],
decl_id: usize,
) -> ParsedInternalCall {
2022-01-01 22:42:50 +01:00
trace!("parsing: internal call (decl id: {})", decl_id);
let mut call = Call::new(command_span);
2021-09-02 10:25:22 +02:00
call.decl_id = decl_id;
call.head = command_span;
let _ = working_set.add_span(call.head);
2021-07-08 08:19:38 +02:00
let decl = working_set.get_decl(decl_id);
let signature = decl.signature();
Improve type hovers (#9515) # Description This PR does a few things to help improve type hovers and, in the process, fixes a few outstanding issues in the type system. Here's a list of the changes: * `for` now will try to infer the type of the iteration variable based on the expression it's given. This fixes things like `for x in [1, 2, 3] { }` where `x` now properly gets the int type. * Removed old input/output type fields from the signature, focuses on the vec of signatures. Updated a bunch of dataframe commands that hadn't moved over. This helps tie things together a bit better * Fixed inference of types from subexpressions to use the last expression in the block * Fixed handling of explicit types in `let` and `mut` calls, so we now respect that as the authoritative type I also tried to add `def` input/output type inference, but unfortunately we only know the predecl types universally, which means we won't have enough information to properly know what the types of the custom commands are. # User-Facing Changes Script typechecking will get tighter in some cases Hovers should be more accurate in some cases that previously resorted to any. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: Darren Schroeder <343840+fdncred@users.noreply.github.com>
2023-06-28 19:19:48 +02:00
let output = signature.get_output_type();
// storing the var ID for later due to borrowing issues
let lib_dirs_var_id = if decl.is_builtin() {
match decl.name() {
"use" | "overlay use" | "source-env" | "nu-check" => {
find_dirs_var(working_set, LIB_DIRS_VAR)
}
_ => None,
}
} else {
None
};
2021-09-02 10:25:22 +02:00
// The index into the positional parameter in the definition
let mut positional_idx = 0;
2021-07-08 08:19:38 +02:00
2021-09-02 10:25:22 +02:00
// The index into the spans of argument data given to parse
// Starting at the first argument
let mut spans_idx = 0;
2021-07-02 00:40:08 +02:00
Re-implement aliases (#8123) # Description This PR adds an alternative alias implementation. Old aliases still work but you need to use `old-alias` instead of `alias`. Instead of replacing spans in the original code and re-parsing, which proved to be extremely error-prone and a constant source of panics, the new implementation creates a new command that references the old command. Consider the new alias defined as `alias ll = ls -l`. The parser creates a new command called `ll` and remembers that it is actually a `ls` command called with the `-l` flag. Then, when the parser sees the `ll` command, it will translate it to `ls -l` and passes to it any parameters that were passed to the call to `ll`. It works quite similar to how known externals defined with `extern` are implemented. The new alias implementation should work the same way as the old aliases, including exporting from modules, referencing both known and unknown externals. It seems to preserve custom completions and pipeline metadata. It is quite robust in most cases but there are some rough edges (see later). Fixes https://github.com/nushell/nushell/issues/7648, https://github.com/nushell/nushell/issues/8026, https://github.com/nushell/nushell/issues/7512, https://github.com/nushell/nushell/issues/5780, https://github.com/nushell/nushell/issues/7754 No effect: https://github.com/nushell/nushell/issues/8122 (we might revisit the completions code after this PR) Should use custom command instead: https://github.com/nushell/nushell/issues/6048 # User-Facing Changes Since aliases are now basically commands, it has some new implications: 1. `alias spam = "spam"` (requires command call) * **workaround**: use `alias spam = echo "spam"` 2. `def foo [] { 'foo' }; alias foo = ls -l` (foo defined more than once) * **workaround**: use different name (commands also have this limitation) 4. `alias ls = (ls | sort-by type name -i)` * **workaround**: Use custom command. _The common issue with this is that it is currently not easy to pass flags through custom commands and command referencing itself will lead to stack overflow. Both of these issues are meant to be addressed._ 5. TODO: Help messages, `which` command, `$nu.scope.aliases`, etc. * Should we treat the aliases as commands or should they be separated from regular commands? 6. Needs better error message and syntax highlight for recursed alias (`alias f = f`) 7. Can't create alias with the same name as existing command (`alias ls = ls -a`) * Might be possible to add support for it (not 100% sure) 8. Standalone `alias` doesn't list aliases anymore 9. Can't alias parser keywords (e.g., stuff like `alias ou = overlay use` won't work) * TODO: Needs a better error message when attempting to do so # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-02-27 08:44:05 +01:00
if let Some(alias) = decl.as_alias() {
if let Expression {
expr: Expr::Call(wrapped_call),
..
} = &alias.wrapped_call
{
// Replace this command's call with the aliased call, but keep the alias name
call = *wrapped_call.clone();
call.head = command_span;
// Skip positionals passed to aliased call
positional_idx = call.positional_len();
} else {
working_set.error(ParseError::UnknownState(
"Alias does not point to internal call.".to_string(),
command_span,
));
Re-implement aliases (#8123) # Description This PR adds an alternative alias implementation. Old aliases still work but you need to use `old-alias` instead of `alias`. Instead of replacing spans in the original code and re-parsing, which proved to be extremely error-prone and a constant source of panics, the new implementation creates a new command that references the old command. Consider the new alias defined as `alias ll = ls -l`. The parser creates a new command called `ll` and remembers that it is actually a `ls` command called with the `-l` flag. Then, when the parser sees the `ll` command, it will translate it to `ls -l` and passes to it any parameters that were passed to the call to `ll`. It works quite similar to how known externals defined with `extern` are implemented. The new alias implementation should work the same way as the old aliases, including exporting from modules, referencing both known and unknown externals. It seems to preserve custom completions and pipeline metadata. It is quite robust in most cases but there are some rough edges (see later). Fixes https://github.com/nushell/nushell/issues/7648, https://github.com/nushell/nushell/issues/8026, https://github.com/nushell/nushell/issues/7512, https://github.com/nushell/nushell/issues/5780, https://github.com/nushell/nushell/issues/7754 No effect: https://github.com/nushell/nushell/issues/8122 (we might revisit the completions code after this PR) Should use custom command instead: https://github.com/nushell/nushell/issues/6048 # User-Facing Changes Since aliases are now basically commands, it has some new implications: 1. `alias spam = "spam"` (requires command call) * **workaround**: use `alias spam = echo "spam"` 2. `def foo [] { 'foo' }; alias foo = ls -l` (foo defined more than once) * **workaround**: use different name (commands also have this limitation) 4. `alias ls = (ls | sort-by type name -i)` * **workaround**: Use custom command. _The common issue with this is that it is currently not easy to pass flags through custom commands and command referencing itself will lead to stack overflow. Both of these issues are meant to be addressed._ 5. TODO: Help messages, `which` command, `$nu.scope.aliases`, etc. * Should we treat the aliases as commands or should they be separated from regular commands? 6. Needs better error message and syntax highlight for recursed alias (`alias f = f`) 7. Can't create alias with the same name as existing command (`alias ls = ls -a`) * Might be possible to add support for it (not 100% sure) 8. Standalone `alias` doesn't list aliases anymore 9. Can't alias parser keywords (e.g., stuff like `alias ou = overlay use` won't work) * TODO: Needs a better error message when attempting to do so # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-02-27 08:44:05 +01:00
return ParsedInternalCall {
call: Box::new(call),
output: Type::Any,
};
}
}
if let Some(var_id) = lib_dirs_var_id {
call.set_parser_info(
DIR_VAR_PARSER_INFO.to_owned(),
Expression::new(working_set, Expr::Var(var_id), call.head, Type::Any),
);
}
Re-implement aliases (#8123) # Description This PR adds an alternative alias implementation. Old aliases still work but you need to use `old-alias` instead of `alias`. Instead of replacing spans in the original code and re-parsing, which proved to be extremely error-prone and a constant source of panics, the new implementation creates a new command that references the old command. Consider the new alias defined as `alias ll = ls -l`. The parser creates a new command called `ll` and remembers that it is actually a `ls` command called with the `-l` flag. Then, when the parser sees the `ll` command, it will translate it to `ls -l` and passes to it any parameters that were passed to the call to `ll`. It works quite similar to how known externals defined with `extern` are implemented. The new alias implementation should work the same way as the old aliases, including exporting from modules, referencing both known and unknown externals. It seems to preserve custom completions and pipeline metadata. It is quite robust in most cases but there are some rough edges (see later). Fixes https://github.com/nushell/nushell/issues/7648, https://github.com/nushell/nushell/issues/8026, https://github.com/nushell/nushell/issues/7512, https://github.com/nushell/nushell/issues/5780, https://github.com/nushell/nushell/issues/7754 No effect: https://github.com/nushell/nushell/issues/8122 (we might revisit the completions code after this PR) Should use custom command instead: https://github.com/nushell/nushell/issues/6048 # User-Facing Changes Since aliases are now basically commands, it has some new implications: 1. `alias spam = "spam"` (requires command call) * **workaround**: use `alias spam = echo "spam"` 2. `def foo [] { 'foo' }; alias foo = ls -l` (foo defined more than once) * **workaround**: use different name (commands also have this limitation) 4. `alias ls = (ls | sort-by type name -i)` * **workaround**: Use custom command. _The common issue with this is that it is currently not easy to pass flags through custom commands and command referencing itself will lead to stack overflow. Both of these issues are meant to be addressed._ 5. TODO: Help messages, `which` command, `$nu.scope.aliases`, etc. * Should we treat the aliases as commands or should they be separated from regular commands? 6. Needs better error message and syntax highlight for recursed alias (`alias f = f`) 7. Can't create alias with the same name as existing command (`alias ls = ls -a`) * Might be possible to add support for it (not 100% sure) 8. Standalone `alias` doesn't list aliases anymore 9. Can't alias parser keywords (e.g., stuff like `alias ou = overlay use` won't work) * TODO: Needs a better error message when attempting to do so # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-02-27 08:44:05 +01:00
if signature.creates_scope {
working_set.enter_scope();
}
2021-09-02 10:25:22 +02:00
while spans_idx < spans.len() {
let arg_span = spans[spans_idx];
2021-07-08 23:16:25 +02:00
let starting_error_count = working_set.parse_errors.len();
2021-09-02 10:25:22 +02:00
// Check if we're on a long flag, if so, parse
let (long_name, arg) = parse_long_flag(working_set, spans, &mut spans_idx, &signature);
2021-09-02 10:25:22 +02:00
if let Some(long_name) = long_name {
// We found a long flag, like --bar
if working_set.parse_errors[starting_error_count..]
.iter()
.any(|x| matches!(x, ParseError::UnknownFlag(_, _, _, _)))
&& signature.allows_unknown_args
{
working_set.parse_errors.truncate(starting_error_count);
let arg = parse_value(working_set, arg_span, &SyntaxShape::Any);
call.add_unknown(arg);
} else {
call.add_named((long_name, None, arg));
}
2021-09-02 10:25:22 +02:00
spans_idx += 1;
continue;
}
2021-07-02 00:40:08 +02:00
let starting_error_count = working_set.parse_errors.len();
2021-09-02 10:25:22 +02:00
// Check if we're on a short flag or group of short flags, if so, parse
let short_flags = parse_short_flags(
2021-09-02 10:25:22 +02:00
working_set,
spans,
&mut spans_idx,
positional_idx,
&signature,
);
2021-07-08 22:29:00 +02:00
2022-07-17 14:46:40 +02:00
if let Some(mut short_flags) = short_flags {
if short_flags.is_empty() {
// workaround for completions (PR #6067)
2022-07-17 14:46:40 +02:00
short_flags.push(Flag {
long: "".to_string(),
short: Some('a'),
arg: None,
required: false,
desc: "".to_string(),
var_id: None,
default_value: None,
})
}
2021-09-02 10:25:22 +02:00
if working_set.parse_errors[starting_error_count..]
.iter()
.any(|x| matches!(x, ParseError::UnknownFlag(_, _, _, _)))
&& signature.allows_unknown_args
{
working_set.parse_errors.truncate(starting_error_count);
let arg = parse_value(working_set, arg_span, &SyntaxShape::Any);
call.add_unknown(arg);
} else {
for flag in short_flags {
let _ = working_set.add_span(spans[spans_idx]);
if let Some(arg_shape) = flag.arg {
if let Some(arg) = spans.get(spans_idx + 1) {
let arg = parse_value(working_set, *arg, &arg_shape);
if flag.long.is_empty() {
if let Some(short) = flag.short {
call.add_named((
Spanned {
item: String::new(),
span: spans[spans_idx],
},
Some(Spanned {
item: short.to_string(),
span: spans[spans_idx],
}),
Some(arg),
));
}
} else {
call.add_named((
Spanned {
item: flag.long.clone(),
span: spans[spans_idx],
},
None,
Some(arg),
));
}
spans_idx += 1;
} else {
working_set.error(ParseError::MissingFlagParam(
arg_shape.to_string(),
arg_span,
))
}
} else if flag.long.is_empty() {
if let Some(short) = flag.short {
call.add_named((
Spanned {
item: String::new(),
span: spans[spans_idx],
},
Some(Spanned {
item: short.to_string(),
span: spans[spans_idx],
}),
None,
));
}
2021-07-08 08:19:38 +02:00
} else {
call.add_named((
Spanned {
item: flag.long.clone(),
span: spans[spans_idx],
},
None,
None,
));
}
2021-07-08 08:19:38 +02:00
}
2021-07-08 22:29:00 +02:00
}
2021-09-02 10:25:22 +02:00
spans_idx += 1;
continue;
}
2021-07-08 22:29:00 +02:00
Allow spreading arguments to commands (#11289) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> Finishes implementing https://github.com/nushell/nushell/issues/10598, which asks for a spread operator in lists, in records, and when calling commands. # Description <!-- Thank you for improving Nushell. Please, check our [contributing guide](../CONTRIBUTING.md) and talk to the core team before making major changes. Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience. --> This PR will allow spreading arguments to commands (both internal and external). It will also deprecate spreading arguments automatically when passing to external commands. # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> - Users will be able to use `...` to spread arguments to custom/builtin commands that have rest parameters or allow unknown arguments, or to any external command - If a custom command doesn't have a rest parameter and it doesn't allow unknown arguments either, the spread operator will not be allowed - Passing lists to external commands without `...` will work for now but will cause a deprecation warning saying that it'll stop working in 0.91 (is 2 versions enough time?) Here's a function to help with demonstrating some behavior: ```nushell > def foo [ a, b, c?, d?, ...rest ] { [$a $b $c $d $rest] | to nuon } ``` You can pass a list of arguments to fill in the `rest` parameter using `...`: ```nushell > foo 1 2 3 4 ...[5 6] [1, 2, 3, 4, [5, 6]] ``` If you don't use `...`, the list `[5 6]` will be treated as a single argument: ```nushell > foo 1 2 3 4 [5 6] # Note the double [[]] [1, 2, 3, 4, [[5, 6]]] ``` You can omit optional parameters before the spread arguments: ```nushell > foo 1 2 3 ...[4 5] # d is omitted here [1, 2, 3, null, [4, 5]] ``` If you have multiple lists, you can spread them all: ```nushell > foo 1 2 3 ...[4 5] 6 7 ...[8] ...[] [1, 2, 3, null, [4, 5, 6, 7, 8]] ``` Here's the kind of error you get when you try to spread arguments to a command with no rest parameter: ![image](https://github.com/nushell/nushell/assets/45539777/93faceae-00eb-4e59-ac3f-17f98436e6e4) And this is the warning you get when you pass a list to an external now (without `...`): ![image](https://github.com/nushell/nushell/assets/45539777/d368f590-201e-49fb-8b20-68476ced415e) # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass (on Windows make sure to [enable developer mode](https://learn.microsoft.com/en-us/windows/apps/get-started/developer-mode-features-and-debugging)) - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> Added tests to cover the following cases: - Spreading arguments to a command that doesn't have a rest parameter (unexpected spread argument error) - Spreading arguments to a command that doesn't have a rest parameter *but* there's also a missing positional argument (missing positional error) - Spreading arguments to a command that doesn't have a rest parameter but does allow unknown arguments, such as `exec` (allowed) - Spreading a list literal containing arguments of the wrong type (parse error) - Spreading a non-list value, both to internal and external commands - Having named arguments in the middle of rest arguments - `explain`ing a command call that spreads its arguments # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> # Examples Suppose you have multiple tables: ```nushell let people = [[id name age]; [0 alice 100] [1 bob 200] [2 eve 300]] let evil_twins = [[id name age]; [0 ecila 100] [-1 bob 200] [-2 eve 300]] ``` Maybe you often find yourself needing to merge multiple tables and want a utility to do that. You could write a function like this: ```nushell def merge_all [ ...tables ] { $tables | reduce { |it, acc| $acc | merge $it } } ``` Then you can use it like this: ```nushell > merge_all ...([$people $evil_twins] | each { |$it| $it | select name age }) ╭───┬───────┬─────╮ │ # │ name │ age │ ├───┼───────┼─────┤ │ 0 │ ecila │ 100 │ │ 1 │ bob │ 200 │ │ 2 │ eve │ 300 │ ╰───┴───────┴─────╯ ``` Except they had duplicate columns, so now you first want to suffix every column with a number to tell you which table the column came from. You can make a command for that: ```nushell def select_and_merge [ --cols: list<string>, ...tables ] { let renamed_tables = $tables | enumerate | each { |it| $it.item | select $cols | rename ...($cols | each { |col| $col + ($it.index | into string) }) }; merge_all ...$renamed_tables } ``` And call it like this: ```nushell > select_and_merge --cols [name age] $people $evil_twins ╭───┬───────┬──────┬───────┬──────╮ │ # │ name0 │ age0 │ name1 │ age1 │ ├───┼───────┼──────┼───────┼──────┤ │ 0 │ alice │ 100 │ ecila │ 100 │ │ 1 │ bob │ 200 │ bob │ 200 │ │ 2 │ eve │ 300 │ eve │ 300 │ ╰───┴───────┴──────┴───────┴──────╯ ``` --- Suppose someone's made a command to search for APT packages: ```nushell # The main command def search-pkgs [ --install # Whether to install any packages it finds log_level: int # Pretend it's a good idea to make this a required positional parameter exclude?: list<string> # Packages to exclude repositories?: list<string> # Which repositories to look in (searches in all if not given) ...pkgs # Package names to search for ] { { install: $install, log_level: $log_level, exclude: ($exclude | to nuon), repositories: ($repositories | to nuon), pkgs: ($pkgs | to nuon) } } ``` It has a lot of parameters to configure it, so you might make your own helper commands to wrap around it for specific cases. Here's one example: ```nushell # Only look for packages locally def search-pkgs-local [ --install # Whether to install any packages it finds log_level: int exclude?: list<string> # Packages to exclude ...pkgs # Package names to search for ] { # All required and optional positional parameters are given search-pkgs --install=$install $log_level [] ["<local URI or something>"] ...$pkgs } ``` And you can run it like this: ```nushell > search-pkgs-local --install=false 5 ...["python2.7" "vim"] ╭──────────────┬──────────────────────────────╮ │ install │ false │ │ log_level │ 5 │ │ exclude │ [] │ │ repositories │ ["<local URI or something>"] │ │ pkgs │ ["python2.7", vim] │ ╰──────────────┴──────────────────────────────╯ ``` One thing I realized when writing this was that if we decide to not allow passing optional arguments using the spread operator, then you can (mis?)use the spread operator to skip optional parameters. Here, I didn't want to give `exclude` explicitly, so I used a spread operator to pass the packages to install. Without it, I would've needed to do `search-pkgs-local --install=false 5 [] "python2.7" "vim"` (explicitly pass `[]` (or `null`, in the general case) to `exclude`). There are probably more idiomatic ways to do this, but I just thought it was something interesting. If you're a virologist of the [xkcd](https://xkcd.com/350/) kind, another helper command you might make is this: ```nushell # Install any packages it finds def live-dangerously [ ...pkgs ] { # One optional argument was given (exclude), while another was not (repositories) search-pkgs 0 [] ...$pkgs --install # Flags can go after spread arguments } ``` Running it: ```nushell > live-dangerously "git" "*vi*" # *vi* because I don't feel like typing out vim and neovim ╭──────────────┬─────────────╮ │ install │ true │ │ log_level │ 0 │ │ exclude │ [] │ │ repositories │ null │ │ pkgs │ [git, *vi*] │ ╰──────────────┴─────────────╯ ``` Here's an example that uses the spread operator more than once within the same command call: ```nushell let extras = [ chrome firefox python java git ] def search-pkgs-curated [ ...pkgs ] { (search-pkgs 1 [emacs] ["example.com", "foo.com"] vim # A must for everyone! ...($pkgs | filter { |p| not ($p | str contains "*") }) # Remove packages with globs python # Good tool to have ...$extras --install=false python3) # I forget, did I already put Python in extras? } ``` Running it: ```nushell > search-pkgs-curated "git" "*vi*" ╭──────────────┬───────────────────────────────────────────────────────────────────╮ │ install │ false │ │ log_level │ 1 │ │ exclude │ [emacs] │ │ repositories │ [example.com, foo.com] │ │ pkgs │ [vim, git, python, chrome, firefox, python, java, git, "python3"] │ ╰──────────────┴───────────────────────────────────────────────────────────────────╯ ```
2023-12-28 08:43:20 +01:00
{
let contents = working_set.get_span_contents(spans[spans_idx]);
if contents.len() > 3
&& contents.starts_with(b"...")
&& (contents[3] == b'$' || contents[3] == b'[' || contents[3] == b'(')
{
if signature.rest_positional.is_none() && !signature.allows_unknown_args {
working_set.error(ParseError::UnexpectedSpreadArg(
signature.call_signature(),
arg_span,
));
call.add_positional(Expression::garbage(working_set, arg_span));
Allow spreading arguments to commands (#11289) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> Finishes implementing https://github.com/nushell/nushell/issues/10598, which asks for a spread operator in lists, in records, and when calling commands. # Description <!-- Thank you for improving Nushell. Please, check our [contributing guide](../CONTRIBUTING.md) and talk to the core team before making major changes. Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience. --> This PR will allow spreading arguments to commands (both internal and external). It will also deprecate spreading arguments automatically when passing to external commands. # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> - Users will be able to use `...` to spread arguments to custom/builtin commands that have rest parameters or allow unknown arguments, or to any external command - If a custom command doesn't have a rest parameter and it doesn't allow unknown arguments either, the spread operator will not be allowed - Passing lists to external commands without `...` will work for now but will cause a deprecation warning saying that it'll stop working in 0.91 (is 2 versions enough time?) Here's a function to help with demonstrating some behavior: ```nushell > def foo [ a, b, c?, d?, ...rest ] { [$a $b $c $d $rest] | to nuon } ``` You can pass a list of arguments to fill in the `rest` parameter using `...`: ```nushell > foo 1 2 3 4 ...[5 6] [1, 2, 3, 4, [5, 6]] ``` If you don't use `...`, the list `[5 6]` will be treated as a single argument: ```nushell > foo 1 2 3 4 [5 6] # Note the double [[]] [1, 2, 3, 4, [[5, 6]]] ``` You can omit optional parameters before the spread arguments: ```nushell > foo 1 2 3 ...[4 5] # d is omitted here [1, 2, 3, null, [4, 5]] ``` If you have multiple lists, you can spread them all: ```nushell > foo 1 2 3 ...[4 5] 6 7 ...[8] ...[] [1, 2, 3, null, [4, 5, 6, 7, 8]] ``` Here's the kind of error you get when you try to spread arguments to a command with no rest parameter: ![image](https://github.com/nushell/nushell/assets/45539777/93faceae-00eb-4e59-ac3f-17f98436e6e4) And this is the warning you get when you pass a list to an external now (without `...`): ![image](https://github.com/nushell/nushell/assets/45539777/d368f590-201e-49fb-8b20-68476ced415e) # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass (on Windows make sure to [enable developer mode](https://learn.microsoft.com/en-us/windows/apps/get-started/developer-mode-features-and-debugging)) - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> Added tests to cover the following cases: - Spreading arguments to a command that doesn't have a rest parameter (unexpected spread argument error) - Spreading arguments to a command that doesn't have a rest parameter *but* there's also a missing positional argument (missing positional error) - Spreading arguments to a command that doesn't have a rest parameter but does allow unknown arguments, such as `exec` (allowed) - Spreading a list literal containing arguments of the wrong type (parse error) - Spreading a non-list value, both to internal and external commands - Having named arguments in the middle of rest arguments - `explain`ing a command call that spreads its arguments # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> # Examples Suppose you have multiple tables: ```nushell let people = [[id name age]; [0 alice 100] [1 bob 200] [2 eve 300]] let evil_twins = [[id name age]; [0 ecila 100] [-1 bob 200] [-2 eve 300]] ``` Maybe you often find yourself needing to merge multiple tables and want a utility to do that. You could write a function like this: ```nushell def merge_all [ ...tables ] { $tables | reduce { |it, acc| $acc | merge $it } } ``` Then you can use it like this: ```nushell > merge_all ...([$people $evil_twins] | each { |$it| $it | select name age }) ╭───┬───────┬─────╮ │ # │ name │ age │ ├───┼───────┼─────┤ │ 0 │ ecila │ 100 │ │ 1 │ bob │ 200 │ │ 2 │ eve │ 300 │ ╰───┴───────┴─────╯ ``` Except they had duplicate columns, so now you first want to suffix every column with a number to tell you which table the column came from. You can make a command for that: ```nushell def select_and_merge [ --cols: list<string>, ...tables ] { let renamed_tables = $tables | enumerate | each { |it| $it.item | select $cols | rename ...($cols | each { |col| $col + ($it.index | into string) }) }; merge_all ...$renamed_tables } ``` And call it like this: ```nushell > select_and_merge --cols [name age] $people $evil_twins ╭───┬───────┬──────┬───────┬──────╮ │ # │ name0 │ age0 │ name1 │ age1 │ ├───┼───────┼──────┼───────┼──────┤ │ 0 │ alice │ 100 │ ecila │ 100 │ │ 1 │ bob │ 200 │ bob │ 200 │ │ 2 │ eve │ 300 │ eve │ 300 │ ╰───┴───────┴──────┴───────┴──────╯ ``` --- Suppose someone's made a command to search for APT packages: ```nushell # The main command def search-pkgs [ --install # Whether to install any packages it finds log_level: int # Pretend it's a good idea to make this a required positional parameter exclude?: list<string> # Packages to exclude repositories?: list<string> # Which repositories to look in (searches in all if not given) ...pkgs # Package names to search for ] { { install: $install, log_level: $log_level, exclude: ($exclude | to nuon), repositories: ($repositories | to nuon), pkgs: ($pkgs | to nuon) } } ``` It has a lot of parameters to configure it, so you might make your own helper commands to wrap around it for specific cases. Here's one example: ```nushell # Only look for packages locally def search-pkgs-local [ --install # Whether to install any packages it finds log_level: int exclude?: list<string> # Packages to exclude ...pkgs # Package names to search for ] { # All required and optional positional parameters are given search-pkgs --install=$install $log_level [] ["<local URI or something>"] ...$pkgs } ``` And you can run it like this: ```nushell > search-pkgs-local --install=false 5 ...["python2.7" "vim"] ╭──────────────┬──────────────────────────────╮ │ install │ false │ │ log_level │ 5 │ │ exclude │ [] │ │ repositories │ ["<local URI or something>"] │ │ pkgs │ ["python2.7", vim] │ ╰──────────────┴──────────────────────────────╯ ``` One thing I realized when writing this was that if we decide to not allow passing optional arguments using the spread operator, then you can (mis?)use the spread operator to skip optional parameters. Here, I didn't want to give `exclude` explicitly, so I used a spread operator to pass the packages to install. Without it, I would've needed to do `search-pkgs-local --install=false 5 [] "python2.7" "vim"` (explicitly pass `[]` (or `null`, in the general case) to `exclude`). There are probably more idiomatic ways to do this, but I just thought it was something interesting. If you're a virologist of the [xkcd](https://xkcd.com/350/) kind, another helper command you might make is this: ```nushell # Install any packages it finds def live-dangerously [ ...pkgs ] { # One optional argument was given (exclude), while another was not (repositories) search-pkgs 0 [] ...$pkgs --install # Flags can go after spread arguments } ``` Running it: ```nushell > live-dangerously "git" "*vi*" # *vi* because I don't feel like typing out vim and neovim ╭──────────────┬─────────────╮ │ install │ true │ │ log_level │ 0 │ │ exclude │ [] │ │ repositories │ null │ │ pkgs │ [git, *vi*] │ ╰──────────────┴─────────────╯ ``` Here's an example that uses the spread operator more than once within the same command call: ```nushell let extras = [ chrome firefox python java git ] def search-pkgs-curated [ ...pkgs ] { (search-pkgs 1 [emacs] ["example.com", "foo.com"] vim # A must for everyone! ...($pkgs | filter { |p| not ($p | str contains "*") }) # Remove packages with globs python # Good tool to have ...$extras --install=false python3) # I forget, did I already put Python in extras? } ``` Running it: ```nushell > search-pkgs-curated "git" "*vi*" ╭──────────────┬───────────────────────────────────────────────────────────────────╮ │ install │ false │ │ log_level │ 1 │ │ exclude │ [emacs] │ │ repositories │ [example.com, foo.com] │ │ pkgs │ [vim, git, python, chrome, firefox, python, java, git, "python3"] │ ╰──────────────┴───────────────────────────────────────────────────────────────────╯ ```
2023-12-28 08:43:20 +01:00
} else if positional_idx < signature.required_positional.len() {
working_set.error(ParseError::MissingPositional(
signature.required_positional[positional_idx].name.clone(),
Span::new(spans[spans_idx].start, spans[spans_idx].start),
signature.call_signature(),
));
call.add_positional(Expression::garbage(working_set, arg_span));
Allow spreading arguments to commands (#11289) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> Finishes implementing https://github.com/nushell/nushell/issues/10598, which asks for a spread operator in lists, in records, and when calling commands. # Description <!-- Thank you for improving Nushell. Please, check our [contributing guide](../CONTRIBUTING.md) and talk to the core team before making major changes. Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience. --> This PR will allow spreading arguments to commands (both internal and external). It will also deprecate spreading arguments automatically when passing to external commands. # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> - Users will be able to use `...` to spread arguments to custom/builtin commands that have rest parameters or allow unknown arguments, or to any external command - If a custom command doesn't have a rest parameter and it doesn't allow unknown arguments either, the spread operator will not be allowed - Passing lists to external commands without `...` will work for now but will cause a deprecation warning saying that it'll stop working in 0.91 (is 2 versions enough time?) Here's a function to help with demonstrating some behavior: ```nushell > def foo [ a, b, c?, d?, ...rest ] { [$a $b $c $d $rest] | to nuon } ``` You can pass a list of arguments to fill in the `rest` parameter using `...`: ```nushell > foo 1 2 3 4 ...[5 6] [1, 2, 3, 4, [5, 6]] ``` If you don't use `...`, the list `[5 6]` will be treated as a single argument: ```nushell > foo 1 2 3 4 [5 6] # Note the double [[]] [1, 2, 3, 4, [[5, 6]]] ``` You can omit optional parameters before the spread arguments: ```nushell > foo 1 2 3 ...[4 5] # d is omitted here [1, 2, 3, null, [4, 5]] ``` If you have multiple lists, you can spread them all: ```nushell > foo 1 2 3 ...[4 5] 6 7 ...[8] ...[] [1, 2, 3, null, [4, 5, 6, 7, 8]] ``` Here's the kind of error you get when you try to spread arguments to a command with no rest parameter: ![image](https://github.com/nushell/nushell/assets/45539777/93faceae-00eb-4e59-ac3f-17f98436e6e4) And this is the warning you get when you pass a list to an external now (without `...`): ![image](https://github.com/nushell/nushell/assets/45539777/d368f590-201e-49fb-8b20-68476ced415e) # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass (on Windows make sure to [enable developer mode](https://learn.microsoft.com/en-us/windows/apps/get-started/developer-mode-features-and-debugging)) - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> Added tests to cover the following cases: - Spreading arguments to a command that doesn't have a rest parameter (unexpected spread argument error) - Spreading arguments to a command that doesn't have a rest parameter *but* there's also a missing positional argument (missing positional error) - Spreading arguments to a command that doesn't have a rest parameter but does allow unknown arguments, such as `exec` (allowed) - Spreading a list literal containing arguments of the wrong type (parse error) - Spreading a non-list value, both to internal and external commands - Having named arguments in the middle of rest arguments - `explain`ing a command call that spreads its arguments # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> # Examples Suppose you have multiple tables: ```nushell let people = [[id name age]; [0 alice 100] [1 bob 200] [2 eve 300]] let evil_twins = [[id name age]; [0 ecila 100] [-1 bob 200] [-2 eve 300]] ``` Maybe you often find yourself needing to merge multiple tables and want a utility to do that. You could write a function like this: ```nushell def merge_all [ ...tables ] { $tables | reduce { |it, acc| $acc | merge $it } } ``` Then you can use it like this: ```nushell > merge_all ...([$people $evil_twins] | each { |$it| $it | select name age }) ╭───┬───────┬─────╮ │ # │ name │ age │ ├───┼───────┼─────┤ │ 0 │ ecila │ 100 │ │ 1 │ bob │ 200 │ │ 2 │ eve │ 300 │ ╰───┴───────┴─────╯ ``` Except they had duplicate columns, so now you first want to suffix every column with a number to tell you which table the column came from. You can make a command for that: ```nushell def select_and_merge [ --cols: list<string>, ...tables ] { let renamed_tables = $tables | enumerate | each { |it| $it.item | select $cols | rename ...($cols | each { |col| $col + ($it.index | into string) }) }; merge_all ...$renamed_tables } ``` And call it like this: ```nushell > select_and_merge --cols [name age] $people $evil_twins ╭───┬───────┬──────┬───────┬──────╮ │ # │ name0 │ age0 │ name1 │ age1 │ ├───┼───────┼──────┼───────┼──────┤ │ 0 │ alice │ 100 │ ecila │ 100 │ │ 1 │ bob │ 200 │ bob │ 200 │ │ 2 │ eve │ 300 │ eve │ 300 │ ╰───┴───────┴──────┴───────┴──────╯ ``` --- Suppose someone's made a command to search for APT packages: ```nushell # The main command def search-pkgs [ --install # Whether to install any packages it finds log_level: int # Pretend it's a good idea to make this a required positional parameter exclude?: list<string> # Packages to exclude repositories?: list<string> # Which repositories to look in (searches in all if not given) ...pkgs # Package names to search for ] { { install: $install, log_level: $log_level, exclude: ($exclude | to nuon), repositories: ($repositories | to nuon), pkgs: ($pkgs | to nuon) } } ``` It has a lot of parameters to configure it, so you might make your own helper commands to wrap around it for specific cases. Here's one example: ```nushell # Only look for packages locally def search-pkgs-local [ --install # Whether to install any packages it finds log_level: int exclude?: list<string> # Packages to exclude ...pkgs # Package names to search for ] { # All required and optional positional parameters are given search-pkgs --install=$install $log_level [] ["<local URI or something>"] ...$pkgs } ``` And you can run it like this: ```nushell > search-pkgs-local --install=false 5 ...["python2.7" "vim"] ╭──────────────┬──────────────────────────────╮ │ install │ false │ │ log_level │ 5 │ │ exclude │ [] │ │ repositories │ ["<local URI or something>"] │ │ pkgs │ ["python2.7", vim] │ ╰──────────────┴──────────────────────────────╯ ``` One thing I realized when writing this was that if we decide to not allow passing optional arguments using the spread operator, then you can (mis?)use the spread operator to skip optional parameters. Here, I didn't want to give `exclude` explicitly, so I used a spread operator to pass the packages to install. Without it, I would've needed to do `search-pkgs-local --install=false 5 [] "python2.7" "vim"` (explicitly pass `[]` (or `null`, in the general case) to `exclude`). There are probably more idiomatic ways to do this, but I just thought it was something interesting. If you're a virologist of the [xkcd](https://xkcd.com/350/) kind, another helper command you might make is this: ```nushell # Install any packages it finds def live-dangerously [ ...pkgs ] { # One optional argument was given (exclude), while another was not (repositories) search-pkgs 0 [] ...$pkgs --install # Flags can go after spread arguments } ``` Running it: ```nushell > live-dangerously "git" "*vi*" # *vi* because I don't feel like typing out vim and neovim ╭──────────────┬─────────────╮ │ install │ true │ │ log_level │ 0 │ │ exclude │ [] │ │ repositories │ null │ │ pkgs │ [git, *vi*] │ ╰──────────────┴─────────────╯ ``` Here's an example that uses the spread operator more than once within the same command call: ```nushell let extras = [ chrome firefox python java git ] def search-pkgs-curated [ ...pkgs ] { (search-pkgs 1 [emacs] ["example.com", "foo.com"] vim # A must for everyone! ...($pkgs | filter { |p| not ($p | str contains "*") }) # Remove packages with globs python # Good tool to have ...$extras --install=false python3) # I forget, did I already put Python in extras? } ``` Running it: ```nushell > search-pkgs-curated "git" "*vi*" ╭──────────────┬───────────────────────────────────────────────────────────────────╮ │ install │ false │ │ log_level │ 1 │ │ exclude │ [emacs] │ │ repositories │ [example.com, foo.com] │ │ pkgs │ [vim, git, python, chrome, firefox, python, java, git, "python3"] │ ╰──────────────┴───────────────────────────────────────────────────────────────────╯ ```
2023-12-28 08:43:20 +01:00
} else {
let rest_shape = match &signature.rest_positional {
Some(arg) => arg.shape.clone(),
None => SyntaxShape::Any,
};
// Parse list of arguments to be spread
let args = parse_value(
working_set,
Span::new(arg_span.start + 3, arg_span.end),
&SyntaxShape::List(Box::new(rest_shape)),
);
call.add_spread(args);
// Let the parser know that it's parsing rest arguments now
positional_idx =
signature.required_positional.len() + signature.optional_positional.len();
}
spans_idx += 1;
continue;
}
}
2021-09-02 10:25:22 +02:00
// Parse a positional arg if there is one
if let Some(positional) = signature.get_positional(positional_idx) {
let end = calculate_end_span(working_set, &signature, spans, spans_idx, positional_idx);
2021-07-17 01:22:01 +02:00
let end = if spans.len() > spans_idx && end == spans_idx {
end + 1
} else {
end
};
if spans[..end].is_empty() || spans_idx == end {
working_set.error(ParseError::MissingPositional(
positional.name.clone(),
Span::new(spans[spans_idx].end, spans[spans_idx].end),
signature.call_signature(),
));
positional_idx += 1;
continue;
}
let arg = parse_multispan_value(
2021-09-02 10:25:22 +02:00
working_set,
&spans[..end],
&mut spans_idx,
&positional.shape,
);
2021-07-23 23:46:55 +02:00
2021-09-02 10:25:22 +02:00
let arg = if !type_compatible(&positional.shape.to_type(), &arg.ty) {
working_set.error(ParseError::TypeMismatch(
positional.shape.to_type(),
arg.ty,
arg.span,
));
Expression::garbage(working_set, arg.span)
2021-07-08 08:19:38 +02:00
} else {
2021-09-02 10:25:22 +02:00
arg
};
call.add_positional(arg);
2021-09-02 10:25:22 +02:00
positional_idx += 1;
} else if signature.allows_unknown_args {
let arg = parse_value(working_set, arg_span, &SyntaxShape::Any);
call.add_unknown(arg);
2021-09-02 10:25:22 +02:00
} else {
call.add_positional(Expression::garbage(working_set, arg_span));
working_set.error(ParseError::ExtraPositional(
signature.call_signature(),
arg_span,
))
2021-07-08 08:19:38 +02:00
}
2021-07-02 00:40:08 +02:00
2021-09-02 10:25:22 +02:00
spans_idx += 1;
2021-07-08 08:19:38 +02:00
}
check_call(working_set, command_span, &signature, &call);
2021-07-31 07:20:40 +02:00
2021-10-09 18:10:46 +02:00
if signature.creates_scope {
working_set.exit_scope();
}
ParsedInternalCall {
call: Box::new(call),
output,
}
2021-09-02 10:25:22 +02:00
}
2021-07-31 07:20:40 +02:00
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
pub fn parse_call(working_set: &mut StateWorkingSet, spans: &[Span], head: Span) -> Expression {
2022-01-01 22:42:50 +01:00
trace!("parsing: call");
2021-10-29 22:50:28 +02:00
if spans.is_empty() {
working_set.error(ParseError::UnknownState(
"Encountered command with zero spans".into(),
Span::concat(spans),
));
return garbage(working_set, head);
}
2021-09-02 10:25:22 +02:00
let mut pos = 0;
let cmd_start = pos;
2021-10-29 22:50:28 +02:00
let mut name_spans = vec![];
2022-01-10 03:52:01 +01:00
let mut name = vec![];
2021-08-17 01:00:00 +02:00
2021-10-29 22:50:28 +02:00
for word_span in spans[cmd_start..].iter() {
// Find the longest group of words that could form a command
2021-10-29 22:50:28 +02:00
name_spans.push(*word_span);
2021-09-02 10:25:22 +02:00
2022-01-10 03:52:01 +01:00
let name_part = working_set.get_span_contents(*word_span);
if name.is_empty() {
name.extend(name_part);
} else {
name.push(b' ');
name.extend(name_part);
}
2021-09-02 10:25:22 +02:00
2021-10-29 22:50:28 +02:00
pos += 1;
}
Remove broken compile-time overload system (#9677) # Description This PR removes the compile-time overload system. Unfortunately, this system never worked correctly because in a gradual type system where types can be `Any`, you don't have enough information to correctly resolve function calls with overloads. These resolutions must be done at runtime, if they're supported. That said, there's a bit of work that needs to go into resolving input/output types (here overloads do not execute separate commands, but the same command and each overload explains how each output type corresponds to input types). This PR also removes the type scope, which would give incorrect answers in cases where multiple subexpressions were used in a pipeline. # User-Facing Changes Finishes removing compile-time overloads. These were only used in a few places in the code base, but it's possible it may impact user code. I'll mark this as breaking change so we can review. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-13 21:05:03 +02:00
let mut maybe_decl_id = working_set.find_decl(&name);
2021-10-29 22:50:28 +02:00
while maybe_decl_id.is_none() {
// Find the longest command match
if name_spans.len() <= 1 {
// Keep the first word even if it does not match -- could be external command
break;
2021-06-30 03:42:56 +02:00
}
2021-09-11 14:07:19 +02:00
2021-10-29 22:50:28 +02:00
name_spans.pop();
pos -= 1;
2022-01-10 03:52:01 +01:00
let mut name = vec![];
for name_span in &name_spans {
let name_part = working_set.get_span_contents(*name_span);
if name.is_empty() {
name.extend(name_part);
} else {
name.push(b' ');
name.extend(name_part);
}
}
Remove broken compile-time overload system (#9677) # Description This PR removes the compile-time overload system. Unfortunately, this system never worked correctly because in a gradual type system where types can be `Any`, you don't have enough information to correctly resolve function calls with overloads. These resolutions must be done at runtime, if they're supported. That said, there's a bit of work that needs to go into resolving input/output types (here overloads do not execute separate commands, but the same command and each overload explains how each output type corresponds to input types). This PR also removes the type scope, which would give incorrect answers in cases where multiple subexpressions were used in a pipeline. # User-Facing Changes Finishes removing compile-time overloads. These were only used in a few places in the code base, but it's possible it may impact user code. I'll mark this as breaking change so we can review. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-13 21:05:03 +02:00
maybe_decl_id = working_set.find_decl(&name);
2021-10-29 22:50:28 +02:00
}
if let Some(decl_id) = maybe_decl_id {
2021-09-11 14:07:19 +02:00
// Before the internal parsing we check if there is no let or alias declarations
// that are missing their name, e.g.: let = 1 or alias = 2
if spans.len() > 1 {
let test_equal = working_set.get_span_contents(spans[1]);
2021-09-11 14:16:40 +02:00
if test_equal == [b'='] {
2022-01-01 22:42:50 +01:00
trace!("incomplete statement");
working_set.error(ParseError::UnknownState(
"Incomplete statement".into(),
Span::concat(spans),
));
return garbage(working_set, Span::concat(spans));
2021-09-11 14:07:19 +02:00
}
}
Re-implement aliases (#8123) # Description This PR adds an alternative alias implementation. Old aliases still work but you need to use `old-alias` instead of `alias`. Instead of replacing spans in the original code and re-parsing, which proved to be extremely error-prone and a constant source of panics, the new implementation creates a new command that references the old command. Consider the new alias defined as `alias ll = ls -l`. The parser creates a new command called `ll` and remembers that it is actually a `ls` command called with the `-l` flag. Then, when the parser sees the `ll` command, it will translate it to `ls -l` and passes to it any parameters that were passed to the call to `ll`. It works quite similar to how known externals defined with `extern` are implemented. The new alias implementation should work the same way as the old aliases, including exporting from modules, referencing both known and unknown externals. It seems to preserve custom completions and pipeline metadata. It is quite robust in most cases but there are some rough edges (see later). Fixes https://github.com/nushell/nushell/issues/7648, https://github.com/nushell/nushell/issues/8026, https://github.com/nushell/nushell/issues/7512, https://github.com/nushell/nushell/issues/5780, https://github.com/nushell/nushell/issues/7754 No effect: https://github.com/nushell/nushell/issues/8122 (we might revisit the completions code after this PR) Should use custom command instead: https://github.com/nushell/nushell/issues/6048 # User-Facing Changes Since aliases are now basically commands, it has some new implications: 1. `alias spam = "spam"` (requires command call) * **workaround**: use `alias spam = echo "spam"` 2. `def foo [] { 'foo' }; alias foo = ls -l` (foo defined more than once) * **workaround**: use different name (commands also have this limitation) 4. `alias ls = (ls | sort-by type name -i)` * **workaround**: Use custom command. _The common issue with this is that it is currently not easy to pass flags through custom commands and command referencing itself will lead to stack overflow. Both of these issues are meant to be addressed._ 5. TODO: Help messages, `which` command, `$nu.scope.aliases`, etc. * Should we treat the aliases as commands or should they be separated from regular commands? 6. Needs better error message and syntax highlight for recursed alias (`alias f = f`) 7. Can't create alias with the same name as existing command (`alias ls = ls -a`) * Might be possible to add support for it (not 100% sure) 8. Standalone `alias` doesn't list aliases anymore 9. Can't alias parser keywords (e.g., stuff like `alias ou = overlay use` won't work) * TODO: Needs a better error message when attempting to do so # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-02-27 08:44:05 +01:00
// TODO: Try to remove the clone
Remove broken compile-time overload system (#9677) # Description This PR removes the compile-time overload system. Unfortunately, this system never worked correctly because in a gradual type system where types can be `Any`, you don't have enough information to correctly resolve function calls with overloads. These resolutions must be done at runtime, if they're supported. That said, there's a bit of work that needs to go into resolving input/output types (here overloads do not execute separate commands, but the same command and each overload explains how each output type corresponds to input types). This PR also removes the type scope, which would give incorrect answers in cases where multiple subexpressions were used in a pipeline. # User-Facing Changes Finishes removing compile-time overloads. These were only used in a few places in the code base, but it's possible it may impact user code. I'll mark this as breaking change so we can review. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-13 21:05:03 +02:00
let decl = working_set.get_decl(decl_id);
2022-01-01 22:42:50 +01:00
Re-implement aliases (#8123) # Description This PR adds an alternative alias implementation. Old aliases still work but you need to use `old-alias` instead of `alias`. Instead of replacing spans in the original code and re-parsing, which proved to be extremely error-prone and a constant source of panics, the new implementation creates a new command that references the old command. Consider the new alias defined as `alias ll = ls -l`. The parser creates a new command called `ll` and remembers that it is actually a `ls` command called with the `-l` flag. Then, when the parser sees the `ll` command, it will translate it to `ls -l` and passes to it any parameters that were passed to the call to `ll`. It works quite similar to how known externals defined with `extern` are implemented. The new alias implementation should work the same way as the old aliases, including exporting from modules, referencing both known and unknown externals. It seems to preserve custom completions and pipeline metadata. It is quite robust in most cases but there are some rough edges (see later). Fixes https://github.com/nushell/nushell/issues/7648, https://github.com/nushell/nushell/issues/8026, https://github.com/nushell/nushell/issues/7512, https://github.com/nushell/nushell/issues/5780, https://github.com/nushell/nushell/issues/7754 No effect: https://github.com/nushell/nushell/issues/8122 (we might revisit the completions code after this PR) Should use custom command instead: https://github.com/nushell/nushell/issues/6048 # User-Facing Changes Since aliases are now basically commands, it has some new implications: 1. `alias spam = "spam"` (requires command call) * **workaround**: use `alias spam = echo "spam"` 2. `def foo [] { 'foo' }; alias foo = ls -l` (foo defined more than once) * **workaround**: use different name (commands also have this limitation) 4. `alias ls = (ls | sort-by type name -i)` * **workaround**: Use custom command. _The common issue with this is that it is currently not easy to pass flags through custom commands and command referencing itself will lead to stack overflow. Both of these issues are meant to be addressed._ 5. TODO: Help messages, `which` command, `$nu.scope.aliases`, etc. * Should we treat the aliases as commands or should they be separated from regular commands? 6. Needs better error message and syntax highlight for recursed alias (`alias f = f`) 7. Can't create alias with the same name as existing command (`alias ls = ls -a`) * Might be possible to add support for it (not 100% sure) 8. Standalone `alias` doesn't list aliases anymore 9. Can't alias parser keywords (e.g., stuff like `alias ou = overlay use` won't work) * TODO: Needs a better error message when attempting to do so # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-02-27 08:44:05 +01:00
let parsed_call = if let Some(alias) = decl.as_alias() {
if let Expression {
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
expr: Expr::ExternalCall(head, args),
Re-implement aliases (#8123) # Description This PR adds an alternative alias implementation. Old aliases still work but you need to use `old-alias` instead of `alias`. Instead of replacing spans in the original code and re-parsing, which proved to be extremely error-prone and a constant source of panics, the new implementation creates a new command that references the old command. Consider the new alias defined as `alias ll = ls -l`. The parser creates a new command called `ll` and remembers that it is actually a `ls` command called with the `-l` flag. Then, when the parser sees the `ll` command, it will translate it to `ls -l` and passes to it any parameters that were passed to the call to `ll`. It works quite similar to how known externals defined with `extern` are implemented. The new alias implementation should work the same way as the old aliases, including exporting from modules, referencing both known and unknown externals. It seems to preserve custom completions and pipeline metadata. It is quite robust in most cases but there are some rough edges (see later). Fixes https://github.com/nushell/nushell/issues/7648, https://github.com/nushell/nushell/issues/8026, https://github.com/nushell/nushell/issues/7512, https://github.com/nushell/nushell/issues/5780, https://github.com/nushell/nushell/issues/7754 No effect: https://github.com/nushell/nushell/issues/8122 (we might revisit the completions code after this PR) Should use custom command instead: https://github.com/nushell/nushell/issues/6048 # User-Facing Changes Since aliases are now basically commands, it has some new implications: 1. `alias spam = "spam"` (requires command call) * **workaround**: use `alias spam = echo "spam"` 2. `def foo [] { 'foo' }; alias foo = ls -l` (foo defined more than once) * **workaround**: use different name (commands also have this limitation) 4. `alias ls = (ls | sort-by type name -i)` * **workaround**: Use custom command. _The common issue with this is that it is currently not easy to pass flags through custom commands and command referencing itself will lead to stack overflow. Both of these issues are meant to be addressed._ 5. TODO: Help messages, `which` command, `$nu.scope.aliases`, etc. * Should we treat the aliases as commands or should they be separated from regular commands? 6. Needs better error message and syntax highlight for recursed alias (`alias f = f`) 7. Can't create alias with the same name as existing command (`alias ls = ls -a`) * Might be possible to add support for it (not 100% sure) 8. Standalone `alias` doesn't list aliases anymore 9. Can't alias parser keywords (e.g., stuff like `alias ou = overlay use` won't work) * TODO: Needs a better error message when attempting to do so # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-02-27 08:44:05 +01:00
span: _,
span_id: _,
Re-implement aliases (#8123) # Description This PR adds an alternative alias implementation. Old aliases still work but you need to use `old-alias` instead of `alias`. Instead of replacing spans in the original code and re-parsing, which proved to be extremely error-prone and a constant source of panics, the new implementation creates a new command that references the old command. Consider the new alias defined as `alias ll = ls -l`. The parser creates a new command called `ll` and remembers that it is actually a `ls` command called with the `-l` flag. Then, when the parser sees the `ll` command, it will translate it to `ls -l` and passes to it any parameters that were passed to the call to `ll`. It works quite similar to how known externals defined with `extern` are implemented. The new alias implementation should work the same way as the old aliases, including exporting from modules, referencing both known and unknown externals. It seems to preserve custom completions and pipeline metadata. It is quite robust in most cases but there are some rough edges (see later). Fixes https://github.com/nushell/nushell/issues/7648, https://github.com/nushell/nushell/issues/8026, https://github.com/nushell/nushell/issues/7512, https://github.com/nushell/nushell/issues/5780, https://github.com/nushell/nushell/issues/7754 No effect: https://github.com/nushell/nushell/issues/8122 (we might revisit the completions code after this PR) Should use custom command instead: https://github.com/nushell/nushell/issues/6048 # User-Facing Changes Since aliases are now basically commands, it has some new implications: 1. `alias spam = "spam"` (requires command call) * **workaround**: use `alias spam = echo "spam"` 2. `def foo [] { 'foo' }; alias foo = ls -l` (foo defined more than once) * **workaround**: use different name (commands also have this limitation) 4. `alias ls = (ls | sort-by type name -i)` * **workaround**: Use custom command. _The common issue with this is that it is currently not easy to pass flags through custom commands and command referencing itself will lead to stack overflow. Both of these issues are meant to be addressed._ 5. TODO: Help messages, `which` command, `$nu.scope.aliases`, etc. * Should we treat the aliases as commands or should they be separated from regular commands? 6. Needs better error message and syntax highlight for recursed alias (`alias f = f`) 7. Can't create alias with the same name as existing command (`alias ls = ls -a`) * Might be possible to add support for it (not 100% sure) 8. Standalone `alias` doesn't list aliases anymore 9. Can't alias parser keywords (e.g., stuff like `alias ou = overlay use` won't work) * TODO: Needs a better error message when attempting to do so # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-02-27 08:44:05 +01:00
ty,
custom_completion,
Remove broken compile-time overload system (#9677) # Description This PR removes the compile-time overload system. Unfortunately, this system never worked correctly because in a gradual type system where types can be `Any`, you don't have enough information to correctly resolve function calls with overloads. These resolutions must be done at runtime, if they're supported. That said, there's a bit of work that needs to go into resolving input/output types (here overloads do not execute separate commands, but the same command and each overload explains how each output type corresponds to input types). This PR also removes the type scope, which would give incorrect answers in cases where multiple subexpressions were used in a pipeline. # User-Facing Changes Finishes removing compile-time overloads. These were only used in a few places in the code base, but it's possible it may impact user code. I'll mark this as breaking change so we can review. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-13 21:05:03 +02:00
} = &alias.clone().wrapped_call
Re-implement aliases (#8123) # Description This PR adds an alternative alias implementation. Old aliases still work but you need to use `old-alias` instead of `alias`. Instead of replacing spans in the original code and re-parsing, which proved to be extremely error-prone and a constant source of panics, the new implementation creates a new command that references the old command. Consider the new alias defined as `alias ll = ls -l`. The parser creates a new command called `ll` and remembers that it is actually a `ls` command called with the `-l` flag. Then, when the parser sees the `ll` command, it will translate it to `ls -l` and passes to it any parameters that were passed to the call to `ll`. It works quite similar to how known externals defined with `extern` are implemented. The new alias implementation should work the same way as the old aliases, including exporting from modules, referencing both known and unknown externals. It seems to preserve custom completions and pipeline metadata. It is quite robust in most cases but there are some rough edges (see later). Fixes https://github.com/nushell/nushell/issues/7648, https://github.com/nushell/nushell/issues/8026, https://github.com/nushell/nushell/issues/7512, https://github.com/nushell/nushell/issues/5780, https://github.com/nushell/nushell/issues/7754 No effect: https://github.com/nushell/nushell/issues/8122 (we might revisit the completions code after this PR) Should use custom command instead: https://github.com/nushell/nushell/issues/6048 # User-Facing Changes Since aliases are now basically commands, it has some new implications: 1. `alias spam = "spam"` (requires command call) * **workaround**: use `alias spam = echo "spam"` 2. `def foo [] { 'foo' }; alias foo = ls -l` (foo defined more than once) * **workaround**: use different name (commands also have this limitation) 4. `alias ls = (ls | sort-by type name -i)` * **workaround**: Use custom command. _The common issue with this is that it is currently not easy to pass flags through custom commands and command referencing itself will lead to stack overflow. Both of these issues are meant to be addressed._ 5. TODO: Help messages, `which` command, `$nu.scope.aliases`, etc. * Should we treat the aliases as commands or should they be separated from regular commands? 6. Needs better error message and syntax highlight for recursed alias (`alias f = f`) 7. Can't create alias with the same name as existing command (`alias ls = ls -a`) * Might be possible to add support for it (not 100% sure) 8. Standalone `alias` doesn't list aliases anymore 9. Can't alias parser keywords (e.g., stuff like `alias ou = overlay use` won't work) * TODO: Needs a better error message when attempting to do so # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-02-27 08:44:05 +01:00
{
trace!("parsing: alias of external call");
let mut head = head.clone();
head.span = spans[0]; // replacing the spans preserves syntax highlighting
Re-implement aliases (#8123) # Description This PR adds an alternative alias implementation. Old aliases still work but you need to use `old-alias` instead of `alias`. Instead of replacing spans in the original code and re-parsing, which proved to be extremely error-prone and a constant source of panics, the new implementation creates a new command that references the old command. Consider the new alias defined as `alias ll = ls -l`. The parser creates a new command called `ll` and remembers that it is actually a `ls` command called with the `-l` flag. Then, when the parser sees the `ll` command, it will translate it to `ls -l` and passes to it any parameters that were passed to the call to `ll`. It works quite similar to how known externals defined with `extern` are implemented. The new alias implementation should work the same way as the old aliases, including exporting from modules, referencing both known and unknown externals. It seems to preserve custom completions and pipeline metadata. It is quite robust in most cases but there are some rough edges (see later). Fixes https://github.com/nushell/nushell/issues/7648, https://github.com/nushell/nushell/issues/8026, https://github.com/nushell/nushell/issues/7512, https://github.com/nushell/nushell/issues/5780, https://github.com/nushell/nushell/issues/7754 No effect: https://github.com/nushell/nushell/issues/8122 (we might revisit the completions code after this PR) Should use custom command instead: https://github.com/nushell/nushell/issues/6048 # User-Facing Changes Since aliases are now basically commands, it has some new implications: 1. `alias spam = "spam"` (requires command call) * **workaround**: use `alias spam = echo "spam"` 2. `def foo [] { 'foo' }; alias foo = ls -l` (foo defined more than once) * **workaround**: use different name (commands also have this limitation) 4. `alias ls = (ls | sort-by type name -i)` * **workaround**: Use custom command. _The common issue with this is that it is currently not easy to pass flags through custom commands and command referencing itself will lead to stack overflow. Both of these issues are meant to be addressed._ 5. TODO: Help messages, `which` command, `$nu.scope.aliases`, etc. * Should we treat the aliases as commands or should they be separated from regular commands? 6. Needs better error message and syntax highlight for recursed alias (`alias f = f`) 7. Can't create alias with the same name as existing command (`alias ls = ls -a`) * Might be possible to add support for it (not 100% sure) 8. Standalone `alias` doesn't list aliases anymore 9. Can't alias parser keywords (e.g., stuff like `alias ou = overlay use` won't work) * TODO: Needs a better error message when attempting to do so # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-02-27 08:44:05 +01:00
let mut final_args = args.clone().into_vec();
for arg_span in &spans[1..] {
let arg = parse_external_arg(working_set, *arg_span);
Re-implement aliases (#8123) # Description This PR adds an alternative alias implementation. Old aliases still work but you need to use `old-alias` instead of `alias`. Instead of replacing spans in the original code and re-parsing, which proved to be extremely error-prone and a constant source of panics, the new implementation creates a new command that references the old command. Consider the new alias defined as `alias ll = ls -l`. The parser creates a new command called `ll` and remembers that it is actually a `ls` command called with the `-l` flag. Then, when the parser sees the `ll` command, it will translate it to `ls -l` and passes to it any parameters that were passed to the call to `ll`. It works quite similar to how known externals defined with `extern` are implemented. The new alias implementation should work the same way as the old aliases, including exporting from modules, referencing both known and unknown externals. It seems to preserve custom completions and pipeline metadata. It is quite robust in most cases but there are some rough edges (see later). Fixes https://github.com/nushell/nushell/issues/7648, https://github.com/nushell/nushell/issues/8026, https://github.com/nushell/nushell/issues/7512, https://github.com/nushell/nushell/issues/5780, https://github.com/nushell/nushell/issues/7754 No effect: https://github.com/nushell/nushell/issues/8122 (we might revisit the completions code after this PR) Should use custom command instead: https://github.com/nushell/nushell/issues/6048 # User-Facing Changes Since aliases are now basically commands, it has some new implications: 1. `alias spam = "spam"` (requires command call) * **workaround**: use `alias spam = echo "spam"` 2. `def foo [] { 'foo' }; alias foo = ls -l` (foo defined more than once) * **workaround**: use different name (commands also have this limitation) 4. `alias ls = (ls | sort-by type name -i)` * **workaround**: Use custom command. _The common issue with this is that it is currently not easy to pass flags through custom commands and command referencing itself will lead to stack overflow. Both of these issues are meant to be addressed._ 5. TODO: Help messages, `which` command, `$nu.scope.aliases`, etc. * Should we treat the aliases as commands or should they be separated from regular commands? 6. Needs better error message and syntax highlight for recursed alias (`alias f = f`) 7. Can't create alias with the same name as existing command (`alias ls = ls -a`) * Might be possible to add support for it (not 100% sure) 8. Standalone `alias` doesn't list aliases anymore 9. Can't alias parser keywords (e.g., stuff like `alias ou = overlay use` won't work) * TODO: Needs a better error message when attempting to do so # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-02-27 08:44:05 +01:00
final_args.push(arg);
}
let mut expression = Expression::new(
working_set,
Expr::ExternalCall(head, final_args.into()),
Span::concat(spans),
ty.clone(),
);
expression.custom_completion = *custom_completion;
return expression;
Re-implement aliases (#8123) # Description This PR adds an alternative alias implementation. Old aliases still work but you need to use `old-alias` instead of `alias`. Instead of replacing spans in the original code and re-parsing, which proved to be extremely error-prone and a constant source of panics, the new implementation creates a new command that references the old command. Consider the new alias defined as `alias ll = ls -l`. The parser creates a new command called `ll` and remembers that it is actually a `ls` command called with the `-l` flag. Then, when the parser sees the `ll` command, it will translate it to `ls -l` and passes to it any parameters that were passed to the call to `ll`. It works quite similar to how known externals defined with `extern` are implemented. The new alias implementation should work the same way as the old aliases, including exporting from modules, referencing both known and unknown externals. It seems to preserve custom completions and pipeline metadata. It is quite robust in most cases but there are some rough edges (see later). Fixes https://github.com/nushell/nushell/issues/7648, https://github.com/nushell/nushell/issues/8026, https://github.com/nushell/nushell/issues/7512, https://github.com/nushell/nushell/issues/5780, https://github.com/nushell/nushell/issues/7754 No effect: https://github.com/nushell/nushell/issues/8122 (we might revisit the completions code after this PR) Should use custom command instead: https://github.com/nushell/nushell/issues/6048 # User-Facing Changes Since aliases are now basically commands, it has some new implications: 1. `alias spam = "spam"` (requires command call) * **workaround**: use `alias spam = echo "spam"` 2. `def foo [] { 'foo' }; alias foo = ls -l` (foo defined more than once) * **workaround**: use different name (commands also have this limitation) 4. `alias ls = (ls | sort-by type name -i)` * **workaround**: Use custom command. _The common issue with this is that it is currently not easy to pass flags through custom commands and command referencing itself will lead to stack overflow. Both of these issues are meant to be addressed._ 5. TODO: Help messages, `which` command, `$nu.scope.aliases`, etc. * Should we treat the aliases as commands or should they be separated from regular commands? 6. Needs better error message and syntax highlight for recursed alias (`alias f = f`) 7. Can't create alias with the same name as existing command (`alias ls = ls -a`) * Might be possible to add support for it (not 100% sure) 8. Standalone `alias` doesn't list aliases anymore 9. Can't alias parser keywords (e.g., stuff like `alias ou = overlay use` won't work) * TODO: Needs a better error message when attempting to do so # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-02-27 08:44:05 +01:00
} else {
trace!("parsing: alias of internal call");
parse_internal_call(
working_set,
Span::concat(&spans[cmd_start..pos]),
Re-implement aliases (#8123) # Description This PR adds an alternative alias implementation. Old aliases still work but you need to use `old-alias` instead of `alias`. Instead of replacing spans in the original code and re-parsing, which proved to be extremely error-prone and a constant source of panics, the new implementation creates a new command that references the old command. Consider the new alias defined as `alias ll = ls -l`. The parser creates a new command called `ll` and remembers that it is actually a `ls` command called with the `-l` flag. Then, when the parser sees the `ll` command, it will translate it to `ls -l` and passes to it any parameters that were passed to the call to `ll`. It works quite similar to how known externals defined with `extern` are implemented. The new alias implementation should work the same way as the old aliases, including exporting from modules, referencing both known and unknown externals. It seems to preserve custom completions and pipeline metadata. It is quite robust in most cases but there are some rough edges (see later). Fixes https://github.com/nushell/nushell/issues/7648, https://github.com/nushell/nushell/issues/8026, https://github.com/nushell/nushell/issues/7512, https://github.com/nushell/nushell/issues/5780, https://github.com/nushell/nushell/issues/7754 No effect: https://github.com/nushell/nushell/issues/8122 (we might revisit the completions code after this PR) Should use custom command instead: https://github.com/nushell/nushell/issues/6048 # User-Facing Changes Since aliases are now basically commands, it has some new implications: 1. `alias spam = "spam"` (requires command call) * **workaround**: use `alias spam = echo "spam"` 2. `def foo [] { 'foo' }; alias foo = ls -l` (foo defined more than once) * **workaround**: use different name (commands also have this limitation) 4. `alias ls = (ls | sort-by type name -i)` * **workaround**: Use custom command. _The common issue with this is that it is currently not easy to pass flags through custom commands and command referencing itself will lead to stack overflow. Both of these issues are meant to be addressed._ 5. TODO: Help messages, `which` command, `$nu.scope.aliases`, etc. * Should we treat the aliases as commands or should they be separated from regular commands? 6. Needs better error message and syntax highlight for recursed alias (`alias f = f`) 7. Can't create alias with the same name as existing command (`alias ls = ls -a`) * Might be possible to add support for it (not 100% sure) 8. Standalone `alias` doesn't list aliases anymore 9. Can't alias parser keywords (e.g., stuff like `alias ou = overlay use` won't work) * TODO: Needs a better error message when attempting to do so # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-02-27 08:44:05 +01:00
&spans[pos..],
decl_id,
)
}
} else {
trace!("parsing: internal call");
parse_internal_call(
working_set,
Span::concat(&spans[cmd_start..pos]),
Re-implement aliases (#8123) # Description This PR adds an alternative alias implementation. Old aliases still work but you need to use `old-alias` instead of `alias`. Instead of replacing spans in the original code and re-parsing, which proved to be extremely error-prone and a constant source of panics, the new implementation creates a new command that references the old command. Consider the new alias defined as `alias ll = ls -l`. The parser creates a new command called `ll` and remembers that it is actually a `ls` command called with the `-l` flag. Then, when the parser sees the `ll` command, it will translate it to `ls -l` and passes to it any parameters that were passed to the call to `ll`. It works quite similar to how known externals defined with `extern` are implemented. The new alias implementation should work the same way as the old aliases, including exporting from modules, referencing both known and unknown externals. It seems to preserve custom completions and pipeline metadata. It is quite robust in most cases but there are some rough edges (see later). Fixes https://github.com/nushell/nushell/issues/7648, https://github.com/nushell/nushell/issues/8026, https://github.com/nushell/nushell/issues/7512, https://github.com/nushell/nushell/issues/5780, https://github.com/nushell/nushell/issues/7754 No effect: https://github.com/nushell/nushell/issues/8122 (we might revisit the completions code after this PR) Should use custom command instead: https://github.com/nushell/nushell/issues/6048 # User-Facing Changes Since aliases are now basically commands, it has some new implications: 1. `alias spam = "spam"` (requires command call) * **workaround**: use `alias spam = echo "spam"` 2. `def foo [] { 'foo' }; alias foo = ls -l` (foo defined more than once) * **workaround**: use different name (commands also have this limitation) 4. `alias ls = (ls | sort-by type name -i)` * **workaround**: Use custom command. _The common issue with this is that it is currently not easy to pass flags through custom commands and command referencing itself will lead to stack overflow. Both of these issues are meant to be addressed._ 5. TODO: Help messages, `which` command, `$nu.scope.aliases`, etc. * Should we treat the aliases as commands or should they be separated from regular commands? 6. Needs better error message and syntax highlight for recursed alias (`alias f = f`) 7. Can't create alias with the same name as existing command (`alias ls = ls -a`) * Might be possible to add support for it (not 100% sure) 8. Standalone `alias` doesn't list aliases anymore 9. Can't alias parser keywords (e.g., stuff like `alias ou = overlay use` won't work) * TODO: Needs a better error message when attempting to do so # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-02-27 08:44:05 +01:00
&spans[pos..],
decl_id,
)
};
Expression::new(
working_set,
Expr::Call(parsed_call.call),
Span::concat(spans),
parsed_call.output,
)
2021-09-02 10:25:22 +02:00
} else {
// We might be parsing left-unbounded range ("..10")
let bytes = working_set.get_span_contents(spans[0]);
2022-01-01 22:42:50 +01:00
trace!("parsing: range {:?} ", bytes);
if let (Some(b'.'), Some(b'.')) = (bytes.first(), bytes.get(1)) {
2022-01-01 22:42:50 +01:00
trace!("-- found leading range indicator");
let starting_error_count = working_set.parse_errors.len();
let range_expr = parse_range(working_set, spans[0]);
if working_set.parse_errors.len() == starting_error_count {
2022-01-01 22:42:50 +01:00
trace!("-- successfully parsed range");
return range_expr;
}
working_set.parse_errors.truncate(starting_error_count);
}
2022-01-01 22:42:50 +01:00
trace!("parsing: external call");
2021-10-29 22:50:28 +02:00
// Otherwise, try external command
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
parse_external_call(working_set, spans)
2021-09-02 10:25:22 +02:00
}
}
pub fn parse_binary(working_set: &mut StateWorkingSet, span: Span) -> Expression {
Improve type hovers (#9515) # Description This PR does a few things to help improve type hovers and, in the process, fixes a few outstanding issues in the type system. Here's a list of the changes: * `for` now will try to infer the type of the iteration variable based on the expression it's given. This fixes things like `for x in [1, 2, 3] { }` where `x` now properly gets the int type. * Removed old input/output type fields from the signature, focuses on the vec of signatures. Updated a bunch of dataframe commands that hadn't moved over. This helps tie things together a bit better * Fixed inference of types from subexpressions to use the last expression in the block * Fixed handling of explicit types in `let` and `mut` calls, so we now respect that as the authoritative type I also tried to add `def` input/output type inference, but unfortunately we only know the predecl types universally, which means we won't have enough information to properly know what the types of the custom commands are. # User-Facing Changes Script typechecking will get tighter in some cases Hovers should be more accurate in some cases that previously resorted to any. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: Darren Schroeder <343840+fdncred@users.noreply.github.com>
2023-06-28 19:19:48 +02:00
trace!("parsing: binary");
let contents = working_set.get_span_contents(span);
if contents.starts_with(b"0x[") {
parse_binary_with_base(working_set, span, 16, 2, b"0x[", b"]")
} else if contents.starts_with(b"0o[") {
parse_binary_with_base(working_set, span, 8, 3, b"0o[", b"]")
} else if contents.starts_with(b"0b[") {
parse_binary_with_base(working_set, span, 2, 8, b"0b[", b"]")
} else {
working_set.error(ParseError::Expected("binary", span));
garbage(working_set, span)
2022-03-01 00:31:53 +01:00
}
}
2022-03-01 00:31:53 +01:00
fn parse_binary_with_base(
working_set: &mut StateWorkingSet,
span: Span,
base: u32,
min_digits_per_byte: usize,
prefix: &[u8],
suffix: &[u8],
) -> Expression {
2022-03-01 00:31:53 +01:00
let token = working_set.get_span_contents(span);
if let Some(token) = token.strip_prefix(prefix) {
if let Some(token) = token.strip_suffix(suffix) {
let (lexed, err) = lex(
token,
span.start + prefix.len(),
&[b',', b'\r', b'\n'],
&[],
true,
);
if let Some(err) = err {
working_set.error(err);
}
2022-03-01 00:31:53 +01:00
let mut binary_value = vec![];
2022-03-01 00:31:53 +01:00
for token in lexed {
match token.contents {
TokenContents::Item => {
let contents = working_set.get_span_contents(token.span);
binary_value.extend_from_slice(contents);
2022-03-01 00:31:53 +01:00
}
Better errors when bash-like operators are used (#7241) # Description Adds improved errors for when a user uses a bashism that nu doesn't support. fixes #7237 Examples: ``` Error: nu::parser::shell_andand (link) × The '&&' operator is not supported in Nushell ╭─[entry #1:1:1] 1 │ ls && ls · ─┬ · ╰── instead of '&&', use ';' or 'and' ╰──── help: use ';' instead of the shell '&&', or 'and' instead of the boolean '&&' ``` ``` Error: nu::parser::shell_oror (link) × The '||' operator is not supported in Nushell ╭─[entry #8:1:1] 1 │ ls || ls · ─┬ · ╰── instead of '||', use 'try' or 'or' ╰──── help: use 'try' instead of the shell '||', or 'or' instead of the boolean '||' ``` ``` Error: nu::parser::shell_err (link) × The '2>' shell operation is 'err>' in Nushell. ╭─[entry #9:1:1] 1 │ foo 2> bar.txt · ─┬ · ╰── use 'err>' instead of '2>' in Nushell ╰──── ``` ``` Error: nu::parser::shell_outerr (link) × The '2>&1' shell operation is 'out+err>' in Nushell. ╭─[entry #10:1:1] 1 │ foo 2>&1 bar.txt · ──┬─ · ╰── use 'out+err>' instead of '2>&1' in Nushell ╰──── help: Nushell redirection will write all of stdout before stderr. ``` # User-Facing Changes **BREAKING CHANGES** This removes the `&&` and `||` operators. We previously supported by `&&`/`and` and `||`/`or`. With this change, only `and` and `or` are valid boolean operators. # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2022-12-08 00:02:11 +01:00
TokenContents::Pipe
| TokenContents::PipePipe
| TokenContents::ErrGreaterPipe
Better errors when bash-like operators are used (#7241) # Description Adds improved errors for when a user uses a bashism that nu doesn't support. fixes #7237 Examples: ``` Error: nu::parser::shell_andand (link) × The '&&' operator is not supported in Nushell ╭─[entry #1:1:1] 1 │ ls && ls · ─┬ · ╰── instead of '&&', use ';' or 'and' ╰──── help: use ';' instead of the shell '&&', or 'and' instead of the boolean '&&' ``` ``` Error: nu::parser::shell_oror (link) × The '||' operator is not supported in Nushell ╭─[entry #8:1:1] 1 │ ls || ls · ─┬ · ╰── instead of '||', use 'try' or 'or' ╰──── help: use 'try' instead of the shell '||', or 'or' instead of the boolean '||' ``` ``` Error: nu::parser::shell_err (link) × The '2>' shell operation is 'err>' in Nushell. ╭─[entry #9:1:1] 1 │ foo 2> bar.txt · ─┬ · ╰── use 'err>' instead of '2>' in Nushell ╰──── ``` ``` Error: nu::parser::shell_outerr (link) × The '2>&1' shell operation is 'out+err>' in Nushell. ╭─[entry #10:1:1] 1 │ foo 2>&1 bar.txt · ──┬─ · ╰── use 'out+err>' instead of '2>&1' in Nushell ╰──── help: Nushell redirection will write all of stdout before stderr. ``` # User-Facing Changes **BREAKING CHANGES** This removes the `&&` and `||` operators. We previously supported by `&&`/`and` and `||`/`or`. With this change, only `and` and `or` are valid boolean operators. # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2022-12-08 00:02:11 +01:00
| TokenContents::OutGreaterThan
| TokenContents::OutErrGreaterPipe
| TokenContents::OutGreaterGreaterThan
Better errors when bash-like operators are used (#7241) # Description Adds improved errors for when a user uses a bashism that nu doesn't support. fixes #7237 Examples: ``` Error: nu::parser::shell_andand (link) × The '&&' operator is not supported in Nushell ╭─[entry #1:1:1] 1 │ ls && ls · ─┬ · ╰── instead of '&&', use ';' or 'and' ╰──── help: use ';' instead of the shell '&&', or 'and' instead of the boolean '&&' ``` ``` Error: nu::parser::shell_oror (link) × The '||' operator is not supported in Nushell ╭─[entry #8:1:1] 1 │ ls || ls · ─┬ · ╰── instead of '||', use 'try' or 'or' ╰──── help: use 'try' instead of the shell '||', or 'or' instead of the boolean '||' ``` ``` Error: nu::parser::shell_err (link) × The '2>' shell operation is 'err>' in Nushell. ╭─[entry #9:1:1] 1 │ foo 2> bar.txt · ─┬ · ╰── use 'err>' instead of '2>' in Nushell ╰──── ``` ``` Error: nu::parser::shell_outerr (link) × The '2>&1' shell operation is 'out+err>' in Nushell. ╭─[entry #10:1:1] 1 │ foo 2>&1 bar.txt · ──┬─ · ╰── use 'out+err>' instead of '2>&1' in Nushell ╰──── help: Nushell redirection will write all of stdout before stderr. ``` # User-Facing Changes **BREAKING CHANGES** This removes the `&&` and `||` operators. We previously supported by `&&`/`and` and `||`/`or`. With this change, only `and` and `or` are valid boolean operators. # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2022-12-08 00:02:11 +01:00
| TokenContents::ErrGreaterThan
| TokenContents::ErrGreaterGreaterThan
| TokenContents::OutErrGreaterThan
| TokenContents::OutErrGreaterGreaterThan => {
working_set.error(ParseError::Expected("binary", span));
return garbage(working_set, span);
2022-03-01 00:31:53 +01:00
}
Better errors when bash-like operators are used (#7241) # Description Adds improved errors for when a user uses a bashism that nu doesn't support. fixes #7237 Examples: ``` Error: nu::parser::shell_andand (link) × The '&&' operator is not supported in Nushell ╭─[entry #1:1:1] 1 │ ls && ls · ─┬ · ╰── instead of '&&', use ';' or 'and' ╰──── help: use ';' instead of the shell '&&', or 'and' instead of the boolean '&&' ``` ``` Error: nu::parser::shell_oror (link) × The '||' operator is not supported in Nushell ╭─[entry #8:1:1] 1 │ ls || ls · ─┬ · ╰── instead of '||', use 'try' or 'or' ╰──── help: use 'try' instead of the shell '||', or 'or' instead of the boolean '||' ``` ``` Error: nu::parser::shell_err (link) × The '2>' shell operation is 'err>' in Nushell. ╭─[entry #9:1:1] 1 │ foo 2> bar.txt · ─┬ · ╰── use 'err>' instead of '2>' in Nushell ╰──── ``` ``` Error: nu::parser::shell_outerr (link) × The '2>&1' shell operation is 'out+err>' in Nushell. ╭─[entry #10:1:1] 1 │ foo 2>&1 bar.txt · ──┬─ · ╰── use 'out+err>' instead of '2>&1' in Nushell ╰──── help: Nushell redirection will write all of stdout before stderr. ``` # User-Facing Changes **BREAKING CHANGES** This removes the `&&` and `||` operators. We previously supported by `&&`/`and` and `||`/`or`. With this change, only `and` and `or` are valid boolean operators. # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2022-12-08 00:02:11 +01:00
TokenContents::Comment | TokenContents::Semicolon | TokenContents::Eol => {}
2022-03-01 00:31:53 +01:00
}
}
let required_padding = (min_digits_per_byte - binary_value.len() % min_digits_per_byte)
% min_digits_per_byte;
if required_padding != 0 {
binary_value = {
let mut tail = binary_value;
let mut binary_value: Vec<u8> = vec![b'0'; required_padding];
binary_value.append(&mut tail);
binary_value
};
2022-03-01 00:31:53 +01:00
}
let str = String::from_utf8_lossy(&binary_value).to_string();
2022-03-01 00:31:53 +01:00
match decode_with_base(&str, base, min_digits_per_byte) {
Ok(v) => return Expression::new(working_set, Expr::Binary(v), span, Type::Binary),
2022-03-01 00:31:53 +01:00
Err(x) => {
working_set.error(ParseError::IncorrectValue(
"not a binary value".into(),
span,
x.to_string(),
));
return garbage(working_set, span);
2022-03-01 00:31:53 +01:00
}
}
}
}
working_set.error(ParseError::Expected("binary", span));
garbage(working_set, span)
2022-03-01 00:31:53 +01:00
}
fn decode_with_base(s: &str, base: u32, digits_per_byte: usize) -> Result<Vec<u8>, ParseIntError> {
2022-06-17 20:11:48 +02:00
s.chars()
.chunks(digits_per_byte)
.into_iter()
.map(|chunk| {
let str: String = chunk.collect();
u8::from_str_radix(&str, base)
})
.collect()
}
fn strip_underscores(token: &[u8]) -> String {
String::from_utf8_lossy(token)
.chars()
.filter(|c| *c != '_')
.collect()
}
pub fn parse_int(working_set: &mut StateWorkingSet, span: Span) -> Expression {
let token = working_set.get_span_contents(span);
fn extract_int(
working_set: &mut StateWorkingSet,
token: &str,
span: Span,
radix: u32,
) -> Expression {
if let Ok(num) = i64::from_str_radix(token, radix) {
Expression::new(working_set, Expr::Int(num), span, Type::Int)
2021-08-08 22:21:21 +02:00
} else {
working_set.error(ParseError::InvalidLiteral(
format!("invalid digits for radix {}", radix),
"int".into(),
span,
));
garbage(working_set, span)
2021-08-08 22:21:21 +02:00
}
}
let token = strip_underscores(token);
Syntax errors for string and int (#7952) # Description Added a few syntax errors in ints and strings, changed parser to stop and show that error rather than continue trying to parse those tokens as some other shape. However, I don't see how to push this direction much further, and most of the classic confusing errors can't be changed. Flagged as WIP for the moment, but passes all checks and works better than current release: 1. I have yet to figure out how to make these errors refer back to the book, as I see some other errors do. 2. How to give syntax error when malformed int is first token in line? Currently parsed as external command, user gets confusing error message. 3. Would like to be more strict with *decimal* int literals (lacking, e.g, `0x' prefix). Need to tinker more with the order of parse shape calls, currently, float is tried after int, so '1.4' has to be passed. _(Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience.)_ ```bash 〉"\z" Error: ╭─[entry #3:1:1] 1 │ "\z" · ─┬─ · ╰── Syntax error in string, unrecognized character after escape '\'. ╰──── ``` Canonic presentation of a syntax error. ```bash 〉" \u{01ffbogus}" Error: × Invalid syntax ╭─[entry #2:1:1] 1 │ " \u{01ffbogus}" · ───────┬────── · ╰── Syntax error in string, expecting 1 to 6 hex digits in unicode escape '\u{X...}', max value 10FFFF. ╰──── ``` Malformed unicode escape in string, flagged as error. String parse can be opinionated, it's the last shape tried. ```bash 〉0x22bogus Error: nu::shell::external_command (link) × External command failed ╭─[entry #4:1:1] 1 │ 0x22bogus · ────┬──── · ╰── executable was not found ╰──── help: No such file or directory (os error 2) ``` A *correct* number in first token would be evaluated, but an *incorrect* one is treated as external command? Confusing to users. ```bash 〉0 + 0x22bogus Error: × Invalid syntax ╭─[entry #5:1:1] 1 │ 0 + 0x22bogus · ────┬──── · ╰── Syntax error in int, invalid digits in radix 16 int. ╰──── ``` Can give syntax error if token is unambiguously int literal. e.g has 0b or 0x prefix, could not be a float. ```bash 〉0 + 098bogus Error: nu::parser::unsupported_operation (link) × Types mismatched for operation. ╭─[entry #6:1:1] 1 │ 0 + 098bogus · ┬ ┬ ────┬─── · │ │ ╰── string · │ ╰── doesn't support these values. · ╰── int ╰──── help: Change int or string to be the right types and try again. ``` But *decimal* literal (no prefix) can't be too strict. Parser is going to try float later. So '1.4' must be passed. # User-Facing Changes First and foremost, more specific error messages for typos in string and int literals. Probably improves interactive user experience. But a script that was causing and then checking for specific error might notice a different error message. _(List of all changes that impact the user experience here. This helps us keep track of breaking changes.)_ # Tests + Formatting Added (positive and negative unit tests in `cargo test -p nu-parser`. Didn't add integration tests. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [x] `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [x] `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --------- Co-authored-by: Stefan Holderbach <sholderbach@users.noreply.github.com>
2023-02-13 17:09:50 +01:00
if token.is_empty() {
working_set.error(ParseError::Expected("int", span));
return garbage(working_set, span);
Syntax errors for string and int (#7952) # Description Added a few syntax errors in ints and strings, changed parser to stop and show that error rather than continue trying to parse those tokens as some other shape. However, I don't see how to push this direction much further, and most of the classic confusing errors can't be changed. Flagged as WIP for the moment, but passes all checks and works better than current release: 1. I have yet to figure out how to make these errors refer back to the book, as I see some other errors do. 2. How to give syntax error when malformed int is first token in line? Currently parsed as external command, user gets confusing error message. 3. Would like to be more strict with *decimal* int literals (lacking, e.g, `0x' prefix). Need to tinker more with the order of parse shape calls, currently, float is tried after int, so '1.4' has to be passed. _(Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience.)_ ```bash 〉"\z" Error: ╭─[entry #3:1:1] 1 │ "\z" · ─┬─ · ╰── Syntax error in string, unrecognized character after escape '\'. ╰──── ``` Canonic presentation of a syntax error. ```bash 〉" \u{01ffbogus}" Error: × Invalid syntax ╭─[entry #2:1:1] 1 │ " \u{01ffbogus}" · ───────┬────── · ╰── Syntax error in string, expecting 1 to 6 hex digits in unicode escape '\u{X...}', max value 10FFFF. ╰──── ``` Malformed unicode escape in string, flagged as error. String parse can be opinionated, it's the last shape tried. ```bash 〉0x22bogus Error: nu::shell::external_command (link) × External command failed ╭─[entry #4:1:1] 1 │ 0x22bogus · ────┬──── · ╰── executable was not found ╰──── help: No such file or directory (os error 2) ``` A *correct* number in first token would be evaluated, but an *incorrect* one is treated as external command? Confusing to users. ```bash 〉0 + 0x22bogus Error: × Invalid syntax ╭─[entry #5:1:1] 1 │ 0 + 0x22bogus · ────┬──── · ╰── Syntax error in int, invalid digits in radix 16 int. ╰──── ``` Can give syntax error if token is unambiguously int literal. e.g has 0b or 0x prefix, could not be a float. ```bash 〉0 + 098bogus Error: nu::parser::unsupported_operation (link) × Types mismatched for operation. ╭─[entry #6:1:1] 1 │ 0 + 098bogus · ┬ ┬ ────┬─── · │ │ ╰── string · │ ╰── doesn't support these values. · ╰── int ╰──── help: Change int or string to be the right types and try again. ``` But *decimal* literal (no prefix) can't be too strict. Parser is going to try float later. So '1.4' must be passed. # User-Facing Changes First and foremost, more specific error messages for typos in string and int literals. Probably improves interactive user experience. But a script that was causing and then checking for specific error might notice a different error message. _(List of all changes that impact the user experience here. This helps us keep track of breaking changes.)_ # Tests + Formatting Added (positive and negative unit tests in `cargo test -p nu-parser`. Didn't add integration tests. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [x] `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [x] `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --------- Co-authored-by: Stefan Holderbach <sholderbach@users.noreply.github.com>
2023-02-13 17:09:50 +01:00
}
if let Some(num) = token.strip_prefix("0b") {
extract_int(working_set, num, span, 2)
} else if let Some(num) = token.strip_prefix("0o") {
extract_int(working_set, num, span, 8)
} else if let Some(num) = token.strip_prefix("0x") {
extract_int(working_set, num, span, 16)
} else if let Ok(num) = token.parse::<i64>() {
Expression::new(working_set, Expr::Int(num), span, Type::Int)
2021-09-02 10:25:22 +02:00
} else {
working_set.error(ParseError::Expected("int", span));
garbage(working_set, span)
2021-07-01 02:01:04 +02:00
}
2021-09-02 10:25:22 +02:00
}
2021-07-01 02:01:04 +02:00
pub fn parse_float(working_set: &mut StateWorkingSet, span: Span) -> Expression {
let token = working_set.get_span_contents(span);
let token = strip_underscores(token);
if let Ok(x) = token.parse::<f64>() {
Expression::new(working_set, Expr::Float(x), span, Type::Float)
2021-09-02 10:25:22 +02:00
} else {
working_set.error(ParseError::Expected("float", span));
garbage(working_set, span)
2021-09-02 10:25:22 +02:00
}
}
2021-07-30 05:26:06 +02:00
pub fn parse_number(working_set: &mut StateWorkingSet, span: Span) -> Expression {
let starting_error_count = working_set.parse_errors.len();
let result = parse_int(working_set, span);
if starting_error_count == working_set.parse_errors.len() {
return result;
} else if !matches!(
working_set.parse_errors.last(),
Some(ParseError::Expected(_, _))
) {
} else {
working_set.parse_errors.truncate(starting_error_count);
2021-07-30 05:26:06 +02:00
}
let result = parse_float(working_set, span);
if starting_error_count == working_set.parse_errors.len() {
return result;
Syntax errors for string and int (#7952) # Description Added a few syntax errors in ints and strings, changed parser to stop and show that error rather than continue trying to parse those tokens as some other shape. However, I don't see how to push this direction much further, and most of the classic confusing errors can't be changed. Flagged as WIP for the moment, but passes all checks and works better than current release: 1. I have yet to figure out how to make these errors refer back to the book, as I see some other errors do. 2. How to give syntax error when malformed int is first token in line? Currently parsed as external command, user gets confusing error message. 3. Would like to be more strict with *decimal* int literals (lacking, e.g, `0x' prefix). Need to tinker more with the order of parse shape calls, currently, float is tried after int, so '1.4' has to be passed. _(Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience.)_ ```bash 〉"\z" Error: ╭─[entry #3:1:1] 1 │ "\z" · ─┬─ · ╰── Syntax error in string, unrecognized character after escape '\'. ╰──── ``` Canonic presentation of a syntax error. ```bash 〉" \u{01ffbogus}" Error: × Invalid syntax ╭─[entry #2:1:1] 1 │ " \u{01ffbogus}" · ───────┬────── · ╰── Syntax error in string, expecting 1 to 6 hex digits in unicode escape '\u{X...}', max value 10FFFF. ╰──── ``` Malformed unicode escape in string, flagged as error. String parse can be opinionated, it's the last shape tried. ```bash 〉0x22bogus Error: nu::shell::external_command (link) × External command failed ╭─[entry #4:1:1] 1 │ 0x22bogus · ────┬──── · ╰── executable was not found ╰──── help: No such file or directory (os error 2) ``` A *correct* number in first token would be evaluated, but an *incorrect* one is treated as external command? Confusing to users. ```bash 〉0 + 0x22bogus Error: × Invalid syntax ╭─[entry #5:1:1] 1 │ 0 + 0x22bogus · ────┬──── · ╰── Syntax error in int, invalid digits in radix 16 int. ╰──── ``` Can give syntax error if token is unambiguously int literal. e.g has 0b or 0x prefix, could not be a float. ```bash 〉0 + 098bogus Error: nu::parser::unsupported_operation (link) × Types mismatched for operation. ╭─[entry #6:1:1] 1 │ 0 + 098bogus · ┬ ┬ ────┬─── · │ │ ╰── string · │ ╰── doesn't support these values. · ╰── int ╰──── help: Change int or string to be the right types and try again. ``` But *decimal* literal (no prefix) can't be too strict. Parser is going to try float later. So '1.4' must be passed. # User-Facing Changes First and foremost, more specific error messages for typos in string and int literals. Probably improves interactive user experience. But a script that was causing and then checking for specific error might notice a different error message. _(List of all changes that impact the user experience here. This helps us keep track of breaking changes.)_ # Tests + Formatting Added (positive and negative unit tests in `cargo test -p nu-parser`. Didn't add integration tests. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [x] `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [x] `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --------- Co-authored-by: Stefan Holderbach <sholderbach@users.noreply.github.com>
2023-02-13 17:09:50 +01:00
}
working_set.parse_errors.truncate(starting_error_count);
Syntax errors for string and int (#7952) # Description Added a few syntax errors in ints and strings, changed parser to stop and show that error rather than continue trying to parse those tokens as some other shape. However, I don't see how to push this direction much further, and most of the classic confusing errors can't be changed. Flagged as WIP for the moment, but passes all checks and works better than current release: 1. I have yet to figure out how to make these errors refer back to the book, as I see some other errors do. 2. How to give syntax error when malformed int is first token in line? Currently parsed as external command, user gets confusing error message. 3. Would like to be more strict with *decimal* int literals (lacking, e.g, `0x' prefix). Need to tinker more with the order of parse shape calls, currently, float is tried after int, so '1.4' has to be passed. _(Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience.)_ ```bash 〉"\z" Error: ╭─[entry #3:1:1] 1 │ "\z" · ─┬─ · ╰── Syntax error in string, unrecognized character after escape '\'. ╰──── ``` Canonic presentation of a syntax error. ```bash 〉" \u{01ffbogus}" Error: × Invalid syntax ╭─[entry #2:1:1] 1 │ " \u{01ffbogus}" · ───────┬────── · ╰── Syntax error in string, expecting 1 to 6 hex digits in unicode escape '\u{X...}', max value 10FFFF. ╰──── ``` Malformed unicode escape in string, flagged as error. String parse can be opinionated, it's the last shape tried. ```bash 〉0x22bogus Error: nu::shell::external_command (link) × External command failed ╭─[entry #4:1:1] 1 │ 0x22bogus · ────┬──── · ╰── executable was not found ╰──── help: No such file or directory (os error 2) ``` A *correct* number in first token would be evaluated, but an *incorrect* one is treated as external command? Confusing to users. ```bash 〉0 + 0x22bogus Error: × Invalid syntax ╭─[entry #5:1:1] 1 │ 0 + 0x22bogus · ────┬──── · ╰── Syntax error in int, invalid digits in radix 16 int. ╰──── ``` Can give syntax error if token is unambiguously int literal. e.g has 0b or 0x prefix, could not be a float. ```bash 〉0 + 098bogus Error: nu::parser::unsupported_operation (link) × Types mismatched for operation. ╭─[entry #6:1:1] 1 │ 0 + 098bogus · ┬ ┬ ────┬─── · │ │ ╰── string · │ ╰── doesn't support these values. · ╰── int ╰──── help: Change int or string to be the right types and try again. ``` But *decimal* literal (no prefix) can't be too strict. Parser is going to try float later. So '1.4' must be passed. # User-Facing Changes First and foremost, more specific error messages for typos in string and int literals. Probably improves interactive user experience. But a script that was causing and then checking for specific error might notice a different error message. _(List of all changes that impact the user experience here. This helps us keep track of breaking changes.)_ # Tests + Formatting Added (positive and negative unit tests in `cargo test -p nu-parser`. Didn't add integration tests. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [x] `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [x] `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --------- Co-authored-by: Stefan Holderbach <sholderbach@users.noreply.github.com>
2023-02-13 17:09:50 +01:00
working_set.error(ParseError::Expected("number", span));
garbage(working_set, span)
2021-09-02 10:25:22 +02:00
}
2021-07-30 05:26:06 +02:00
pub fn parse_range(working_set: &mut StateWorkingSet, span: Span) -> Expression {
2022-01-01 22:42:50 +01:00
trace!("parsing: range");
// Range follows the following syntax: [<from>][<next_operator><next>]<range_operator>[<to>]
// where <next_operator> is ".."
range operator accepts bot..=top as well as bot..top (#8382) # Description A compromise fix for #8162. Nushell range operator now accepts `..=` to mean the range includes the top value, so you can use your Rust habits. But the unadorned `..` range operator also includes the value, so you can also use your Nushell habits. _(Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience.)_ ```nushell 〉1..5 ╭───┬───╮ │ 0 │ 1 │ │ 1 │ 2 │ │ 2 │ 3 │ │ 3 │ 4 │ │ 4 │ 5 │ ╰───┴───╯ -------------------------------------------- /home/bobhy/src/rust/nushell -------------------------------------------- 〉1..=5 ╭───┬───╮ │ 0 │ 1 │ │ 1 │ 2 │ │ 2 │ 3 │ │ 3 │ 4 │ │ 4 │ 5 │ ╰───┴───╯ -------------------------------------------- /home/bobhy/src/rust/nushell -------------------------------------------- 〉1..<5 ╭───┬───╮ │ 0 │ 1 │ │ 1 │ 2 │ │ 2 │ 3 │ │ 3 │ 4 │ ╰───┴───╯ ``` # User-Facing Changes Existing scripts with range operator will continue to operate as heretofore. _(List of all changes that impact the user experience here. This helps us keep track of breaking changes.)_ # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [x] `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [x] `cargo test --workspace` to check that all tests pass # After Submitting Will update the book to include new syntax.
2023-04-07 13:40:05 +02:00
// and <range_operator> is "..", "..=" or "..<"
// and one of the <from> or <to> bounds must be present (just '..' is not allowed since it
// looks like parent directory)
Syntax errors for string and int (#7952) # Description Added a few syntax errors in ints and strings, changed parser to stop and show that error rather than continue trying to parse those tokens as some other shape. However, I don't see how to push this direction much further, and most of the classic confusing errors can't be changed. Flagged as WIP for the moment, but passes all checks and works better than current release: 1. I have yet to figure out how to make these errors refer back to the book, as I see some other errors do. 2. How to give syntax error when malformed int is first token in line? Currently parsed as external command, user gets confusing error message. 3. Would like to be more strict with *decimal* int literals (lacking, e.g, `0x' prefix). Need to tinker more with the order of parse shape calls, currently, float is tried after int, so '1.4' has to be passed. _(Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience.)_ ```bash 〉"\z" Error: ╭─[entry #3:1:1] 1 │ "\z" · ─┬─ · ╰── Syntax error in string, unrecognized character after escape '\'. ╰──── ``` Canonic presentation of a syntax error. ```bash 〉" \u{01ffbogus}" Error: × Invalid syntax ╭─[entry #2:1:1] 1 │ " \u{01ffbogus}" · ───────┬────── · ╰── Syntax error in string, expecting 1 to 6 hex digits in unicode escape '\u{X...}', max value 10FFFF. ╰──── ``` Malformed unicode escape in string, flagged as error. String parse can be opinionated, it's the last shape tried. ```bash 〉0x22bogus Error: nu::shell::external_command (link) × External command failed ╭─[entry #4:1:1] 1 │ 0x22bogus · ────┬──── · ╰── executable was not found ╰──── help: No such file or directory (os error 2) ``` A *correct* number in first token would be evaluated, but an *incorrect* one is treated as external command? Confusing to users. ```bash 〉0 + 0x22bogus Error: × Invalid syntax ╭─[entry #5:1:1] 1 │ 0 + 0x22bogus · ────┬──── · ╰── Syntax error in int, invalid digits in radix 16 int. ╰──── ``` Can give syntax error if token is unambiguously int literal. e.g has 0b or 0x prefix, could not be a float. ```bash 〉0 + 098bogus Error: nu::parser::unsupported_operation (link) × Types mismatched for operation. ╭─[entry #6:1:1] 1 │ 0 + 098bogus · ┬ ┬ ────┬─── · │ │ ╰── string · │ ╰── doesn't support these values. · ╰── int ╰──── help: Change int or string to be the right types and try again. ``` But *decimal* literal (no prefix) can't be too strict. Parser is going to try float later. So '1.4' must be passed. # User-Facing Changes First and foremost, more specific error messages for typos in string and int literals. Probably improves interactive user experience. But a script that was causing and then checking for specific error might notice a different error message. _(List of all changes that impact the user experience here. This helps us keep track of breaking changes.)_ # Tests + Formatting Added (positive and negative unit tests in `cargo test -p nu-parser`. Didn't add integration tests. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [x] `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [x] `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --------- Co-authored-by: Stefan Holderbach <sholderbach@users.noreply.github.com>
2023-02-13 17:09:50 +01:00
//bugbug range cannot be [..] because that looks like parent directory
let contents = working_set.get_span_contents(span);
let token = if let Ok(s) = String::from_utf8(contents.into()) {
s
} else {
working_set.error(ParseError::NonUtf8(span));
return garbage(working_set, span);
};
if !token.contains("..") {
working_set.error(ParseError::Expected("at least one range bound set", span));
return garbage(working_set, span);
}
// First, figure out what exact operators are used and determine their positions
let dotdot_pos: Vec<_> = token.match_indices("..").map(|(pos, _)| pos).collect();
let (next_op_pos, range_op_pos) = match dotdot_pos.len() {
1 => (None, dotdot_pos[0]),
2 => (Some(dotdot_pos[0]), dotdot_pos[1]),
_ => {
working_set.error(ParseError::Expected(
"one range operator ('..' or '..<') and optionally one next operator ('..')",
span,
));
return garbage(working_set, span);
}
};
Prevent cubic time on nested parentheses (#10467) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> # Description <!-- Thank you for improving Nushell. Please, check our [contributing guide](../CONTRIBUTING.md) and talk to the core team before making major changes. Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience. --> When parse_range get an item like ((((1..2)))) it would try to parse "((((1" with a long chain of recursive parsers, namely: - parse_value - parse_paren_expr - parse_full_cell_path - parse_block - parse_pipeline - parse_builtin_commands - parse_expression - parse_math_expression - parse_value - ... where `parse_paren_expr` calls `parse_range` in turn. Because at any time in the chain `parse_paren_expr` can call `parse_range`, which will then continue the chain, we get quadratic number of function calls, each linear on the size of the input By checking with the lexer that the parens are matched, we prevent the long chain from being called on unmatched braces. Now, this is still more quadratic than it needs to be, to fix that, we should process parens only once, instead of on each recursive call # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> Speed improvements in some edge cases # Tests + Formatting Not sure how to test this, maybe I could add a benchmark <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass (on Windows make sure to [enable developer mode](https://learn.microsoft.com/en-us/windows/apps/get-started/developer-mode-features-and-debugging)) - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> # Other notes Found using the fuzzer, by setting a timeout on max run-time. It also found a stack-overflow on too many parentheses, which this doesn't fix.
2023-09-22 18:24:35 +02:00
// Avoid calling sub-parsers on unmatched parens, to prevent quadratic time on things like ((((1..2))))
// No need to call the expensive parse_value on "((((1"
if dotdot_pos[0] > 0 {
let (_tokens, err) = lex(
&contents[..dotdot_pos[0]],
span.start,
&[],
&[b'.', b'?'],
true,
);
if let Some(_err) = err {
working_set.error(ParseError::Expected("Valid expression before ..", span));
return garbage(working_set, span);
Prevent cubic time on nested parentheses (#10467) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> # Description <!-- Thank you for improving Nushell. Please, check our [contributing guide](../CONTRIBUTING.md) and talk to the core team before making major changes. Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience. --> When parse_range get an item like ((((1..2)))) it would try to parse "((((1" with a long chain of recursive parsers, namely: - parse_value - parse_paren_expr - parse_full_cell_path - parse_block - parse_pipeline - parse_builtin_commands - parse_expression - parse_math_expression - parse_value - ... where `parse_paren_expr` calls `parse_range` in turn. Because at any time in the chain `parse_paren_expr` can call `parse_range`, which will then continue the chain, we get quadratic number of function calls, each linear on the size of the input By checking with the lexer that the parens are matched, we prevent the long chain from being called on unmatched braces. Now, this is still more quadratic than it needs to be, to fix that, we should process parens only once, instead of on each recursive call # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> Speed improvements in some edge cases # Tests + Formatting Not sure how to test this, maybe I could add a benchmark <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass (on Windows make sure to [enable developer mode](https://learn.microsoft.com/en-us/windows/apps/get-started/developer-mode-features-and-debugging)) - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> # Other notes Found using the fuzzer, by setting a timeout on max run-time. It also found a stack-overflow on too many parentheses, which this doesn't fix.
2023-09-22 18:24:35 +02:00
}
}
let (inclusion, range_op_str, range_op_span) = if let Some(pos) = token.find("..<") {
if pos == range_op_pos {
let op_str = "..<";
let op_span = Span::new(
span.start + range_op_pos,
span.start + range_op_pos + op_str.len(),
);
(RangeInclusion::RightExclusive, "..<", op_span)
} else {
working_set.error(ParseError::Expected(
"inclusive operator preceding second range bound",
span,
));
return garbage(working_set, span);
}
} else {
range operator accepts bot..=top as well as bot..top (#8382) # Description A compromise fix for #8162. Nushell range operator now accepts `..=` to mean the range includes the top value, so you can use your Rust habits. But the unadorned `..` range operator also includes the value, so you can also use your Nushell habits. _(Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience.)_ ```nushell 〉1..5 ╭───┬───╮ │ 0 │ 1 │ │ 1 │ 2 │ │ 2 │ 3 │ │ 3 │ 4 │ │ 4 │ 5 │ ╰───┴───╯ -------------------------------------------- /home/bobhy/src/rust/nushell -------------------------------------------- 〉1..=5 ╭───┬───╮ │ 0 │ 1 │ │ 1 │ 2 │ │ 2 │ 3 │ │ 3 │ 4 │ │ 4 │ 5 │ ╰───┴───╯ -------------------------------------------- /home/bobhy/src/rust/nushell -------------------------------------------- 〉1..<5 ╭───┬───╮ │ 0 │ 1 │ │ 1 │ 2 │ │ 2 │ 3 │ │ 3 │ 4 │ ╰───┴───╯ ``` # User-Facing Changes Existing scripts with range operator will continue to operate as heretofore. _(List of all changes that impact the user experience here. This helps us keep track of breaking changes.)_ # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [x] `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [x] `cargo test --workspace` to check that all tests pass # After Submitting Will update the book to include new syntax.
2023-04-07 13:40:05 +02:00
let op_str = if token.contains("..=") { "..=" } else { ".." };
let op_span = Span::new(
span.start + range_op_pos,
span.start + range_op_pos + op_str.len(),
);
range operator accepts bot..=top as well as bot..top (#8382) # Description A compromise fix for #8162. Nushell range operator now accepts `..=` to mean the range includes the top value, so you can use your Rust habits. But the unadorned `..` range operator also includes the value, so you can also use your Nushell habits. _(Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience.)_ ```nushell 〉1..5 ╭───┬───╮ │ 0 │ 1 │ │ 1 │ 2 │ │ 2 │ 3 │ │ 3 │ 4 │ │ 4 │ 5 │ ╰───┴───╯ -------------------------------------------- /home/bobhy/src/rust/nushell -------------------------------------------- 〉1..=5 ╭───┬───╮ │ 0 │ 1 │ │ 1 │ 2 │ │ 2 │ 3 │ │ 3 │ 4 │ │ 4 │ 5 │ ╰───┴───╯ -------------------------------------------- /home/bobhy/src/rust/nushell -------------------------------------------- 〉1..<5 ╭───┬───╮ │ 0 │ 1 │ │ 1 │ 2 │ │ 2 │ 3 │ │ 3 │ 4 │ ╰───┴───╯ ``` # User-Facing Changes Existing scripts with range operator will continue to operate as heretofore. _(List of all changes that impact the user experience here. This helps us keep track of breaking changes.)_ # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [x] `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [x] `cargo test --workspace` to check that all tests pass # After Submitting Will update the book to include new syntax.
2023-04-07 13:40:05 +02:00
(RangeInclusion::Inclusive, op_str, op_span)
};
// Now, based on the operator positions, figure out where the bounds & next are located and
// parse them
2021-10-12 19:44:23 +02:00
// TODO: Actually parse the next number in the range
let from = if token.starts_with("..") {
// token starts with either next operator, or range operator -- we don't care which one
None
} else {
let from_span = Span::new(span.start, span.start + dotdot_pos[0]);
Some(parse_value(working_set, from_span, &SyntaxShape::Number))
};
let to = if token.ends_with(range_op_str) {
None
} else {
let to_span = Span::new(range_op_span.end, span.end);
Some(parse_value(working_set, to_span, &SyntaxShape::Number))
};
2022-01-01 22:42:50 +01:00
trace!("-- from: {:?} to: {:?}", from, to);
if let (None, None) = (&from, &to) {
working_set.error(ParseError::Expected("at least one range bound set", span));
return garbage(working_set, span);
}
let (next, next_op_span) = if let Some(pos) = next_op_pos {
let next_op_span = Span::new(span.start + pos, span.start + pos + "..".len());
let next_span = Span::new(next_op_span.end, range_op_span.start);
(
Some(parse_value(working_set, next_span, &SyntaxShape::Number)),
next_op_span,
)
} else {
2021-12-19 08:46:13 +01:00
(None, span)
};
let operator = RangeOperator {
inclusion,
span: range_op_span,
next_op_span,
};
let range = Range {
from,
next,
to,
operator,
};
Expression::new(working_set, Expr::Range(Box::new(range)), span, Type::Range)
}
pub(crate) fn parse_dollar_expr(working_set: &mut StateWorkingSet, span: Span) -> Expression {
trace!("parsing: dollar expression");
2021-09-02 10:25:22 +02:00
let contents = working_set.get_span_contents(span);
2021-07-30 05:26:06 +02:00
if contents.starts_with(b"$\"") || contents.starts_with(b"$'") {
parse_string_interpolation(working_set, span)
} else if contents.starts_with(b"$.") {
parse_simple_cell_path(working_set, Span::new(span.start + 2, span.end))
2021-09-02 10:25:22 +02:00
} else {
let starting_error_count = working_set.parse_errors.len();
let expr = parse_range(working_set, span);
if starting_error_count == working_set.parse_errors.len() {
expr
} else {
working_set.parse_errors.truncate(starting_error_count);
parse_full_cell_path(working_set, None, span)
}
2021-09-02 10:25:22 +02:00
}
}
2021-07-30 05:26:06 +02:00
add raw-string literal support (#9956) # Description This PR adds raw string support by using `r#` at the beginning of single quoted strings and `#` at the end. Notice that escapes do not process, even within single quotes, parentheses don't mean anything, $variables don't mean anything. It's just a string. ```nushell ❯ echo r#'one\ntwo (blah) ($var)'# one\ntwo (blah) ($var) ``` Notice how they work without `echo` or `print` and how they work without carriage returns. ```nushell ❯ r#'adsfa'# adsfa ❯ r##"asdfa'@qpejq'## asdfa'@qpejq ❯ r#'asdfasdfasf ∙ foqwejfqo@'23rfjqf'# ``` They also have a special configurable color in the repl. (use single quotes though) ![image](https://github.com/nushell/nushell/assets/343840/8780e21d-de4c-45b3-9880-2425f5fe10ef) They should work like rust raw literals and allow `r##`, `r###`, `r####`, etc, to help with having one or many `#`'s in the middle of your raw-string. They should work with `let` as well. ```nushell r#'some\nraw\nstring'# | str upcase ``` closes https://github.com/nushell/nushell/issues/5091 # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: WindSoilder <WindSoilder@outlook.com> Co-authored-by: Ian Manske <ian.manske@pm.me>
2024-05-02 15:36:37 +02:00
pub fn parse_raw_string(working_set: &mut StateWorkingSet, span: Span) -> Expression {
trace!("parsing: raw-string, with required delimiters");
let bytes = working_set.get_span_contents(span);
let prefix_sharp_cnt = if bytes.starts_with(b"r#") {
// actually `sharp_cnt` is always `index - 1`
// but create a variable here to make it clearer.
let mut sharp_cnt = 1;
let mut index = 2;
while index < bytes.len() && bytes[index] == b'#' {
index += 1;
sharp_cnt += 1;
}
sharp_cnt
} else {
working_set.error(ParseError::Expected("r#", span));
return garbage(working_set, span);
add raw-string literal support (#9956) # Description This PR adds raw string support by using `r#` at the beginning of single quoted strings and `#` at the end. Notice that escapes do not process, even within single quotes, parentheses don't mean anything, $variables don't mean anything. It's just a string. ```nushell ❯ echo r#'one\ntwo (blah) ($var)'# one\ntwo (blah) ($var) ``` Notice how they work without `echo` or `print` and how they work without carriage returns. ```nushell ❯ r#'adsfa'# adsfa ❯ r##"asdfa'@qpejq'## asdfa'@qpejq ❯ r#'asdfasdfasf ∙ foqwejfqo@'23rfjqf'# ``` They also have a special configurable color in the repl. (use single quotes though) ![image](https://github.com/nushell/nushell/assets/343840/8780e21d-de4c-45b3-9880-2425f5fe10ef) They should work like rust raw literals and allow `r##`, `r###`, `r####`, etc, to help with having one or many `#`'s in the middle of your raw-string. They should work with `let` as well. ```nushell r#'some\nraw\nstring'# | str upcase ``` closes https://github.com/nushell/nushell/issues/5091 # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: WindSoilder <WindSoilder@outlook.com> Co-authored-by: Ian Manske <ian.manske@pm.me>
2024-05-02 15:36:37 +02:00
};
let expect_postfix_sharp_cnt = prefix_sharp_cnt;
// check the length of whole raw string.
// the whole raw string should contains at least
// 1(r) + prefix_sharp_cnt + 1(') + 1(') + postfix_sharp characters
if bytes.len() < prefix_sharp_cnt + expect_postfix_sharp_cnt + 3 {
working_set.error(ParseError::Unclosed('\''.into(), span));
return garbage(working_set, span);
add raw-string literal support (#9956) # Description This PR adds raw string support by using `r#` at the beginning of single quoted strings and `#` at the end. Notice that escapes do not process, even within single quotes, parentheses don't mean anything, $variables don't mean anything. It's just a string. ```nushell ❯ echo r#'one\ntwo (blah) ($var)'# one\ntwo (blah) ($var) ``` Notice how they work without `echo` or `print` and how they work without carriage returns. ```nushell ❯ r#'adsfa'# adsfa ❯ r##"asdfa'@qpejq'## asdfa'@qpejq ❯ r#'asdfasdfasf ∙ foqwejfqo@'23rfjqf'# ``` They also have a special configurable color in the repl. (use single quotes though) ![image](https://github.com/nushell/nushell/assets/343840/8780e21d-de4c-45b3-9880-2425f5fe10ef) They should work like rust raw literals and allow `r##`, `r###`, `r####`, etc, to help with having one or many `#`'s in the middle of your raw-string. They should work with `let` as well. ```nushell r#'some\nraw\nstring'# | str upcase ``` closes https://github.com/nushell/nushell/issues/5091 # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: WindSoilder <WindSoilder@outlook.com> Co-authored-by: Ian Manske <ian.manske@pm.me>
2024-05-02 15:36:37 +02:00
}
// check for unbalanced # and single quotes.
let postfix_bytes = &bytes[bytes.len() - expect_postfix_sharp_cnt..bytes.len()];
if postfix_bytes.iter().any(|b| *b != b'#') {
working_set.error(ParseError::Unbalanced(
"prefix #".to_string(),
"postfix #".to_string(),
span,
));
return garbage(working_set, span);
add raw-string literal support (#9956) # Description This PR adds raw string support by using `r#` at the beginning of single quoted strings and `#` at the end. Notice that escapes do not process, even within single quotes, parentheses don't mean anything, $variables don't mean anything. It's just a string. ```nushell ❯ echo r#'one\ntwo (blah) ($var)'# one\ntwo (blah) ($var) ``` Notice how they work without `echo` or `print` and how they work without carriage returns. ```nushell ❯ r#'adsfa'# adsfa ❯ r##"asdfa'@qpejq'## asdfa'@qpejq ❯ r#'asdfasdfasf ∙ foqwejfqo@'23rfjqf'# ``` They also have a special configurable color in the repl. (use single quotes though) ![image](https://github.com/nushell/nushell/assets/343840/8780e21d-de4c-45b3-9880-2425f5fe10ef) They should work like rust raw literals and allow `r##`, `r###`, `r####`, etc, to help with having one or many `#`'s in the middle of your raw-string. They should work with `let` as well. ```nushell r#'some\nraw\nstring'# | str upcase ``` closes https://github.com/nushell/nushell/issues/5091 # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: WindSoilder <WindSoilder@outlook.com> Co-authored-by: Ian Manske <ian.manske@pm.me>
2024-05-02 15:36:37 +02:00
}
// check for unblanaced single quotes.
if bytes[1 + prefix_sharp_cnt] != b'\''
|| bytes[bytes.len() - expect_postfix_sharp_cnt - 1] != b'\''
{
working_set.error(ParseError::Unclosed('\''.into(), span));
return garbage(working_set, span);
add raw-string literal support (#9956) # Description This PR adds raw string support by using `r#` at the beginning of single quoted strings and `#` at the end. Notice that escapes do not process, even within single quotes, parentheses don't mean anything, $variables don't mean anything. It's just a string. ```nushell ❯ echo r#'one\ntwo (blah) ($var)'# one\ntwo (blah) ($var) ``` Notice how they work without `echo` or `print` and how they work without carriage returns. ```nushell ❯ r#'adsfa'# adsfa ❯ r##"asdfa'@qpejq'## asdfa'@qpejq ❯ r#'asdfasdfasf ∙ foqwejfqo@'23rfjqf'# ``` They also have a special configurable color in the repl. (use single quotes though) ![image](https://github.com/nushell/nushell/assets/343840/8780e21d-de4c-45b3-9880-2425f5fe10ef) They should work like rust raw literals and allow `r##`, `r###`, `r####`, etc, to help with having one or many `#`'s in the middle of your raw-string. They should work with `let` as well. ```nushell r#'some\nraw\nstring'# | str upcase ``` closes https://github.com/nushell/nushell/issues/5091 # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: WindSoilder <WindSoilder@outlook.com> Co-authored-by: Ian Manske <ian.manske@pm.me>
2024-05-02 15:36:37 +02:00
}
let bytes = &bytes[prefix_sharp_cnt + 1 + 1..bytes.len() - 1 - prefix_sharp_cnt];
if let Ok(token) = String::from_utf8(bytes.into()) {
Expression::new(working_set, Expr::RawString(token), span, Type::String)
add raw-string literal support (#9956) # Description This PR adds raw string support by using `r#` at the beginning of single quoted strings and `#` at the end. Notice that escapes do not process, even within single quotes, parentheses don't mean anything, $variables don't mean anything. It's just a string. ```nushell ❯ echo r#'one\ntwo (blah) ($var)'# one\ntwo (blah) ($var) ``` Notice how they work without `echo` or `print` and how they work without carriage returns. ```nushell ❯ r#'adsfa'# adsfa ❯ r##"asdfa'@qpejq'## asdfa'@qpejq ❯ r#'asdfasdfasf ∙ foqwejfqo@'23rfjqf'# ``` They also have a special configurable color in the repl. (use single quotes though) ![image](https://github.com/nushell/nushell/assets/343840/8780e21d-de4c-45b3-9880-2425f5fe10ef) They should work like rust raw literals and allow `r##`, `r###`, `r####`, etc, to help with having one or many `#`'s in the middle of your raw-string. They should work with `let` as well. ```nushell r#'some\nraw\nstring'# | str upcase ``` closes https://github.com/nushell/nushell/issues/5091 # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: WindSoilder <WindSoilder@outlook.com> Co-authored-by: Ian Manske <ian.manske@pm.me>
2024-05-02 15:36:37 +02:00
} else {
working_set.error(ParseError::Expected("utf8 raw-string", span));
garbage(working_set, span)
add raw-string literal support (#9956) # Description This PR adds raw string support by using `r#` at the beginning of single quoted strings and `#` at the end. Notice that escapes do not process, even within single quotes, parentheses don't mean anything, $variables don't mean anything. It's just a string. ```nushell ❯ echo r#'one\ntwo (blah) ($var)'# one\ntwo (blah) ($var) ``` Notice how they work without `echo` or `print` and how they work without carriage returns. ```nushell ❯ r#'adsfa'# adsfa ❯ r##"asdfa'@qpejq'## asdfa'@qpejq ❯ r#'asdfasdfasf ∙ foqwejfqo@'23rfjqf'# ``` They also have a special configurable color in the repl. (use single quotes though) ![image](https://github.com/nushell/nushell/assets/343840/8780e21d-de4c-45b3-9880-2425f5fe10ef) They should work like rust raw literals and allow `r##`, `r###`, `r####`, etc, to help with having one or many `#`'s in the middle of your raw-string. They should work with `let` as well. ```nushell r#'some\nraw\nstring'# | str upcase ``` closes https://github.com/nushell/nushell/issues/5091 # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: WindSoilder <WindSoilder@outlook.com> Co-authored-by: Ian Manske <ian.manske@pm.me>
2024-05-02 15:36:37 +02:00
}
}
pub fn parse_paren_expr(
working_set: &mut StateWorkingSet,
span: Span,
shape: &SyntaxShape,
) -> Expression {
let starting_error_count = working_set.parse_errors.len();
let expr = parse_range(working_set, span);
if starting_error_count == working_set.parse_errors.len() {
expr
} else {
working_set.parse_errors.truncate(starting_error_count);
if matches!(shape, SyntaxShape::Signature) {
parse_signature(working_set, span)
} else {
parse_full_cell_path(working_set, None, span)
}
}
}
pub fn parse_brace_expr(
working_set: &mut StateWorkingSet,
span: Span,
shape: &SyntaxShape,
) -> Expression {
// Try to detect what kind of value we're about to parse
// FIXME: In the future, we should work over the token stream so we only have to do this once
// before parsing begins
// FIXME: we're still using the shape because we rely on it to know how to handle syntax where
// the parse is ambiguous. We'll need to update the parts of the grammar where this is ambiguous
// and then revisit the parsing.
if span.end <= (span.start + 1) {
working_set.error(ParseError::ExpectedWithStringMsg(
format!("non-block value: {shape}"),
span,
));
return Expression::garbage(working_set, span);
}
let bytes = working_set.get_span_contents(Span::new(span.start + 1, span.end - 1));
let (tokens, _) = lex(bytes, span.start + 1, &[b'\r', b'\n', b'\t'], &[b':'], true);
let second_token = tokens
Apply nightly clippy fixes (#11083) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> # Description <!-- Thank you for improving Nushell. Please, check our [contributing guide](../CONTRIBUTING.md) and talk to the core team before making major changes. Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience. --> Clippy fixes for rust 1.76.0-nightly # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> N/A # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass (on Windows make sure to [enable developer mode](https://learn.microsoft.com/en-us/windows/apps/get-started/developer-mode-features-and-debugging)) - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-11-17 16:15:55 +01:00
.first()
.map(|token| working_set.get_span_contents(token.span));
Apply nightly clippy fixes (#11083) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> # Description <!-- Thank you for improving Nushell. Please, check our [contributing guide](../CONTRIBUTING.md) and talk to the core team before making major changes. Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience. --> Clippy fixes for rust 1.76.0-nightly # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> N/A # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass (on Windows make sure to [enable developer mode](https://learn.microsoft.com/en-us/windows/apps/get-started/developer-mode-features-and-debugging)) - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-11-17 16:15:55 +01:00
let second_token_contents = tokens.first().map(|token| token.contents);
let third_token = tokens
.get(1)
.map(|token| working_set.get_span_contents(token.span));
if second_token.is_none() {
// If we're empty, that means an empty record or closure
if matches!(shape, SyntaxShape::Closure(_)) {
parse_closure_expression(working_set, shape, span)
} else if matches!(shape, SyntaxShape::Block) {
parse_block_expression(working_set, span)
Add pattern matching (#8590) # Description This adds `match` and basic pattern matching. An example: ``` match $x { 1..10 => { print "Value is between 1 and 10" } { foo: $bar } => { print $"Value has a 'foo' field with value ($bar)" } [$a, $b] => { print $"Value is a list with two items: ($a) and ($b)" } _ => { print "Value is none of the above" } } ``` Like the recent changes to `if` to allow it to be used as an expression, `match` can also be used as an expression. This allows you to assign the result to a variable, eg) `let xyz = match ...` I've also included a short-hand pattern for matching records, as I think it might help when doing a lot of record patterns: `{$foo}` which is equivalent to `{foo: $foo}`. There are still missing components, so consider this the first step in full pattern matching support. Currently missing: * Patterns for strings * Or-patterns (like the `|` in Rust) * Patterns for tables (unclear how we want to match a table, so it'll need some design) * Patterns for binary values * And much more # User-Facing Changes [see above] # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-03-24 02:52:01 +01:00
} else if matches!(shape, SyntaxShape::MatchBlock) {
parse_match_block_expression(working_set, span)
} else {
parse_record(working_set, span)
}
} else if matches!(second_token_contents, Some(TokenContents::Pipe))
|| matches!(second_token_contents, Some(TokenContents::PipePipe))
{
parse_closure_expression(working_set, shape, span)
} else if matches!(third_token, Some(b":")) {
parse_full_cell_path(working_set, None, span)
Respect SyntaxShape when parsing spread operator (#11674) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> This fixes an issue brought up by nihilander in [Discord](https://discord.com/channels/601130461678272522/614593951969574961/1201594105986285649). # Description <!-- Thank you for improving Nushell. Please, check our [contributing guide](../CONTRIBUTING.md) and talk to the core team before making major changes. Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience. --> Nushell panics when the spread operator is used like this (the `...$rest` shouldn't actually be parsed as a spread operator at all): ```nu $ def foo [...rest: string] {...$rest} $ foo bar baz thread 'main' panicked at /root/.cargo/registry/src/index.crates.io-6f17d22bba15001f/nu-protocol-0.89.0/src/signature.rs:650:9: Internal error: can't run a predeclaration without a body stack backtrace: 0: rust_begin_unwind 1: core::panicking::panic_fmt 2: <nu_protocol::signature::Predeclaration as nu_protocol::engine::command::Command>::run 3: nu_engine::eval::eval_call 4: nu_engine::eval::eval_expression_with_input 5: nu_engine::eval::eval_element_with_input 6: nu_engine::eval::eval_block 7: nu_cli::util::eval_source 8: nu_cli::repl::evaluate_repl 9: nu::run::run_repl 10: nu::main note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace. ``` The problem was that whenever the parser saw something like `{...$`, `{...(`, or `{...[`, it would treat that as a record with a spread expression, ignoring the syntax shape of the block it was parsing. This should now be fixed, and the snippet above instead gives the following error: ```nu Error: nu::shell::external_command × External command failed ╭─[entry #1:1:1] 1 │ def foo [...rest] {...$rest} · ────┬─── · ╰── executable was not found ╰──── help: No such file or directory (os error 2) ``` # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> Stuff like `do { ...$rest }` will now try to run a command `...$rest` rather than complaining that variable `$rest` doesn't exist. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass (on Windows make sure to [enable developer mode](https://learn.microsoft.com/en-us/windows/apps/get-started/developer-mode-features-and-debugging)) - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> Sorry about the issue, I am not touching the parser again for a long time :)
2024-01-30 06:49:42 +01:00
} else if matches!(shape, SyntaxShape::Closure(_)) {
parse_closure_expression(working_set, shape, span)
} else if matches!(shape, SyntaxShape::Block) {
parse_block_expression(working_set, span)
Add pattern matching (#8590) # Description This adds `match` and basic pattern matching. An example: ``` match $x { 1..10 => { print "Value is between 1 and 10" } { foo: $bar } => { print $"Value has a 'foo' field with value ($bar)" } [$a, $b] => { print $"Value is a list with two items: ($a) and ($b)" } _ => { print "Value is none of the above" } } ``` Like the recent changes to `if` to allow it to be used as an expression, `match` can also be used as an expression. This allows you to assign the result to a variable, eg) `let xyz = match ...` I've also included a short-hand pattern for matching records, as I think it might help when doing a lot of record patterns: `{$foo}` which is equivalent to `{foo: $foo}`. There are still missing components, so consider this the first step in full pattern matching support. Currently missing: * Patterns for strings * Or-patterns (like the `|` in Rust) * Patterns for tables (unclear how we want to match a table, so it'll need some design) * Patterns for binary values * And much more # User-Facing Changes [see above] # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-03-24 02:52:01 +01:00
} else if matches!(shape, SyntaxShape::MatchBlock) {
parse_match_block_expression(working_set, span)
Respect SyntaxShape when parsing spread operator (#11674) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> This fixes an issue brought up by nihilander in [Discord](https://discord.com/channels/601130461678272522/614593951969574961/1201594105986285649). # Description <!-- Thank you for improving Nushell. Please, check our [contributing guide](../CONTRIBUTING.md) and talk to the core team before making major changes. Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience. --> Nushell panics when the spread operator is used like this (the `...$rest` shouldn't actually be parsed as a spread operator at all): ```nu $ def foo [...rest: string] {...$rest} $ foo bar baz thread 'main' panicked at /root/.cargo/registry/src/index.crates.io-6f17d22bba15001f/nu-protocol-0.89.0/src/signature.rs:650:9: Internal error: can't run a predeclaration without a body stack backtrace: 0: rust_begin_unwind 1: core::panicking::panic_fmt 2: <nu_protocol::signature::Predeclaration as nu_protocol::engine::command::Command>::run 3: nu_engine::eval::eval_call 4: nu_engine::eval::eval_expression_with_input 5: nu_engine::eval::eval_element_with_input 6: nu_engine::eval::eval_block 7: nu_cli::util::eval_source 8: nu_cli::repl::evaluate_repl 9: nu::run::run_repl 10: nu::main note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace. ``` The problem was that whenever the parser saw something like `{...$`, `{...(`, or `{...[`, it would treat that as a record with a spread expression, ignoring the syntax shape of the block it was parsing. This should now be fixed, and the snippet above instead gives the following error: ```nu Error: nu::shell::external_command × External command failed ╭─[entry #1:1:1] 1 │ def foo [...rest] {...$rest} · ────┬─── · ╰── executable was not found ╰──── help: No such file or directory (os error 2) ``` # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> Stuff like `do { ...$rest }` will now try to run a command `...$rest` rather than complaining that variable `$rest` doesn't exist. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass (on Windows make sure to [enable developer mode](https://learn.microsoft.com/en-us/windows/apps/get-started/developer-mode-features-and-debugging)) - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> Sorry about the issue, I am not touching the parser again for a long time :)
2024-01-30 06:49:42 +01:00
} else if second_token.is_some_and(|c| {
c.len() > 3 && c.starts_with(b"...") && (c[3] == b'$' || c[3] == b'{' || c[3] == b'(')
}) {
parse_record(working_set, span)
} else if matches!(shape, SyntaxShape::Any) {
parse_closure_expression(working_set, shape, span)
} else {
working_set.error(ParseError::ExpectedWithStringMsg(
format!("non-block value: {shape}"),
span,
));
Expression::garbage(working_set, span)
}
}
pub fn parse_string_interpolation(working_set: &mut StateWorkingSet, span: Span) -> Expression {
2021-09-02 10:25:22 +02:00
#[derive(PartialEq, Eq, Debug)]
enum InterpolationMode {
String,
Expression,
}
2021-07-30 05:26:06 +02:00
2021-09-02 10:25:22 +02:00
let contents = working_set.get_span_contents(span);
2021-07-30 05:26:06 +02:00
let mut double_quote = false;
let (start, end) = if contents.starts_with(b"$\"") {
double_quote = true;
let end = if contents.ends_with(b"\"") && contents.len() > 2 {
span.end - 1
} else {
span.end
};
(span.start + 2, end)
} else if contents.starts_with(b"$'") {
let end = if contents.ends_with(b"'") && contents.len() > 2 {
span.end - 1
} else {
span.end
};
(span.start + 2, end)
2021-09-02 10:25:22 +02:00
} else {
(span.start, span.end)
2021-09-02 10:25:22 +02:00
};
Protocol: debug_assert!() Span to reflect a valid slice (#6806) Also enforce this by #[non_exhaustive] span such that going forward we cannot, in debug builds (1), construct invalid spans. The motivation for this stems from #6431 where I've seen crashes due to invalid slice indexing. My hope is this will mitigate such senarios 1. https://github.com/nushell/nushell/pull/6431#issuecomment-1278147241 # Description (description of your pull request here) # Tests Make sure you've done the following: - [ ] Add tests that cover your changes, either in the command examples, the crate/tests folder, or in the /tests folder. - [ ] Try to think about corner cases and various ways how your changes could break. Cover them with tests. - [ ] If adding tests is not possible, please document in the PR body a minimal example with steps on how to reproduce so one can verify your change works. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [ ] `cargo clippy --workspace --features=extra -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [ ] `cargo test --workspace --features=extra` to check that all the tests pass # Documentation - [ ] If your PR touches a user-facing nushell feature then make sure that there is an entry in the documentation (https://github.com/nushell/nushell.github.io) for the feature, and update it if necessary.
2022-12-03 10:44:12 +01:00
let inner_span = Span::new(start, end);
2021-09-02 10:25:22 +02:00
let contents = working_set.get_span_contents(inner_span).to_vec();
let mut output = vec![];
let mut mode = InterpolationMode::String;
let mut token_start = start;
let mut delimiter_stack = vec![];
2021-09-02 10:25:22 +02:00
let mut consecutive_backslashes: usize = 0;
2021-09-02 10:25:22 +02:00
let mut b = start;
while b != end {
let current_byte = contents[b - start];
if mode == InterpolationMode::String {
let preceding_consecutive_backslashes = consecutive_backslashes;
let is_backslash = current_byte == b'\\';
consecutive_backslashes = if is_backslash {
preceding_consecutive_backslashes + 1
} else {
0
};
if current_byte == b'(' && (!double_quote || preceding_consecutive_backslashes % 2 == 0)
{
mode = InterpolationMode::Expression;
if token_start < b {
let span = Span::new(token_start, b);
let str_contents = working_set.get_span_contents(span);
let (str_contents, err) = if double_quote {
unescape_string(str_contents, span)
} else {
(str_contents.to_vec(), None)
};
if let Some(err) = err {
working_set.error(err);
}
output.push(Expression::new(
working_set,
Expr::String(String::from_utf8_lossy(&str_contents).to_string()),
span,
Type::String,
));
token_start = b;
}
}
}
if mode == InterpolationMode::Expression {
let byte = current_byte;
if let Some(b'\'') = delimiter_stack.last() {
if byte == b'\'' {
delimiter_stack.pop();
}
} else if let Some(b'"') = delimiter_stack.last() {
if byte == b'"' {
delimiter_stack.pop();
}
} else if let Some(b'`') = delimiter_stack.last() {
if byte == b'`' {
delimiter_stack.pop();
}
} else if byte == b'\'' {
delimiter_stack.push(b'\'')
} else if byte == b'"' {
delimiter_stack.push(b'"');
} else if byte == b'`' {
delimiter_stack.push(b'`')
} else if byte == b'(' {
delimiter_stack.push(b')');
} else if byte == b')' {
if let Some(b')') = delimiter_stack.last() {
delimiter_stack.pop();
}
if delimiter_stack.is_empty() {
mode = InterpolationMode::String;
2021-09-02 10:25:22 +02:00
if token_start < b {
Protocol: debug_assert!() Span to reflect a valid slice (#6806) Also enforce this by #[non_exhaustive] span such that going forward we cannot, in debug builds (1), construct invalid spans. The motivation for this stems from #6431 where I've seen crashes due to invalid slice indexing. My hope is this will mitigate such senarios 1. https://github.com/nushell/nushell/pull/6431#issuecomment-1278147241 # Description (description of your pull request here) # Tests Make sure you've done the following: - [ ] Add tests that cover your changes, either in the command examples, the crate/tests folder, or in the /tests folder. - [ ] Try to think about corner cases and various ways how your changes could break. Cover them with tests. - [ ] If adding tests is not possible, please document in the PR body a minimal example with steps on how to reproduce so one can verify your change works. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [ ] `cargo clippy --workspace --features=extra -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [ ] `cargo test --workspace --features=extra` to check that all the tests pass # Documentation - [ ] If your PR touches a user-facing nushell feature then make sure that there is an entry in the documentation (https://github.com/nushell/nushell.github.io) for the feature, and update it if necessary.
2022-12-03 10:44:12 +01:00
let span = Span::new(token_start, b + 1);
2021-07-30 05:26:06 +02:00
let expr = parse_full_cell_path(working_set, None, span);
output.push(expr);
}
2021-09-02 10:25:22 +02:00
token_start = b + 1;
continue;
2021-07-30 05:26:06 +02:00
}
}
}
2021-09-02 10:25:22 +02:00
b += 1;
}
2021-07-30 05:26:06 +02:00
2021-09-02 10:25:22 +02:00
match mode {
InterpolationMode::String => {
if token_start < end {
Protocol: debug_assert!() Span to reflect a valid slice (#6806) Also enforce this by #[non_exhaustive] span such that going forward we cannot, in debug builds (1), construct invalid spans. The motivation for this stems from #6431 where I've seen crashes due to invalid slice indexing. My hope is this will mitigate such senarios 1. https://github.com/nushell/nushell/pull/6431#issuecomment-1278147241 # Description (description of your pull request here) # Tests Make sure you've done the following: - [ ] Add tests that cover your changes, either in the command examples, the crate/tests folder, or in the /tests folder. - [ ] Try to think about corner cases and various ways how your changes could break. Cover them with tests. - [ ] If adding tests is not possible, please document in the PR body a minimal example with steps on how to reproduce so one can verify your change works. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [ ] `cargo clippy --workspace --features=extra -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [ ] `cargo test --workspace --features=extra` to check that all the tests pass # Documentation - [ ] If your PR touches a user-facing nushell feature then make sure that there is an entry in the documentation (https://github.com/nushell/nushell.github.io) for the feature, and update it if necessary.
2022-12-03 10:44:12 +01:00
let span = Span::new(token_start, end);
2021-09-02 10:25:22 +02:00
let str_contents = working_set.get_span_contents(span);
let (str_contents, err) = if double_quote {
unescape_string(str_contents, span)
} else {
(str_contents.to_vec(), None)
};
if let Some(err) = err {
working_set.error(err);
}
output.push(Expression::new(
working_set,
Expr::String(String::from_utf8_lossy(&str_contents).to_string()),
2021-07-30 05:26:06 +02:00
span,
Type::String,
));
2021-09-02 10:25:22 +02:00
}
}
InterpolationMode::Expression => {
if token_start < end {
Protocol: debug_assert!() Span to reflect a valid slice (#6806) Also enforce this by #[non_exhaustive] span such that going forward we cannot, in debug builds (1), construct invalid spans. The motivation for this stems from #6431 where I've seen crashes due to invalid slice indexing. My hope is this will mitigate such senarios 1. https://github.com/nushell/nushell/pull/6431#issuecomment-1278147241 # Description (description of your pull request here) # Tests Make sure you've done the following: - [ ] Add tests that cover your changes, either in the command examples, the crate/tests folder, or in the /tests folder. - [ ] Try to think about corner cases and various ways how your changes could break. Cover them with tests. - [ ] If adding tests is not possible, please document in the PR body a minimal example with steps on how to reproduce so one can verify your change works. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [ ] `cargo clippy --workspace --features=extra -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [ ] `cargo test --workspace --features=extra` to check that all the tests pass # Documentation - [ ] If your PR touches a user-facing nushell feature then make sure that there is an entry in the documentation (https://github.com/nushell/nushell.github.io) for the feature, and update it if necessary.
2022-12-03 10:44:12 +01:00
let span = Span::new(token_start, end);
2021-09-02 10:25:22 +02:00
let expr = parse_full_cell_path(working_set, None, span);
2021-09-02 10:25:22 +02:00
output.push(expr);
}
2021-07-30 05:26:06 +02:00
}
2021-07-24 07:57:17 +02:00
}
2021-07-02 09:15:30 +02:00
Expression::new(
working_set,
Expr::StringInterpolation(output),
span,
Type::String,
)
2021-09-02 10:25:22 +02:00
}
2021-07-24 07:57:17 +02:00
pub fn parse_variable_expr(working_set: &mut StateWorkingSet, span: Span) -> Expression {
2021-09-02 10:25:22 +02:00
let contents = working_set.get_span_contents(span);
if contents == b"$nu" {
return Expression::new(
working_set,
Expr::Var(nu_protocol::NU_VARIABLE_ID),
span,
Type::Any,
);
} else if contents == b"$in" {
return Expression::new(
working_set,
Expr::Var(nu_protocol::IN_VARIABLE_ID),
span,
Type::Any,
);
} else if contents == b"$env" {
return Expression::new(
working_set,
Expr::Var(nu_protocol::ENV_VARIABLE_ID),
span,
Type::Any,
);
2021-09-02 10:25:22 +02:00
}
Better error message if env var is used as var (#9522) # Description This PR improves the error message if an environment variable (that's visible before the parser begins) is used in the form of `$PATH` instead of `$env.PATH`. Before: ``` Error: nu::parser::variable_not_found × Variable not found. ╭─[entry #31:1:1] 1 │ echo $PATH · ──┬── · ╰── variable not found. ╰──── ``` After: ``` Error: nu::parser::env_var_not_var × Use $env.PATH instead of $PATH. ╭─[entry #1:1:1] 1 │ echo $PATH · ──┬── · ╰── use $env.PATH instead of $PATH ╰──── ``` # User-Facing Changes Just the improvement to the error message # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-06-25 19:59:56 +02:00
let name = if contents.starts_with(b"$") {
String::from_utf8_lossy(&contents[1..]).to_string()
} else {
String::from_utf8_lossy(contents).to_string()
};
if let Some(id) = parse_variable(working_set, span) {
Expression::new(
working_set,
Expr::Var(id),
span,
working_set.get_variable(id).ty.clone(),
)
Better error message if env var is used as var (#9522) # Description This PR improves the error message if an environment variable (that's visible before the parser begins) is used in the form of `$PATH` instead of `$env.PATH`. Before: ``` Error: nu::parser::variable_not_found × Variable not found. ╭─[entry #31:1:1] 1 │ echo $PATH · ──┬── · ╰── variable not found. ╰──── ``` After: ``` Error: nu::parser::env_var_not_var × Use $env.PATH instead of $PATH. ╭─[entry #1:1:1] 1 │ echo $PATH · ──┬── · ╰── use $env.PATH instead of $PATH ╰──── ``` # User-Facing Changes Just the improvement to the error message # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-06-25 19:59:56 +02:00
} else if working_set.get_env_var(&name).is_some() {
working_set.error(ParseError::EnvVarNotVar(name, span));
garbage(working_set, span)
2021-09-02 10:25:22 +02:00
} else {
let ws = &*working_set;
let suggestion = DidYouMean::new(&ws.list_variables(), ws.get_span_contents(span));
working_set.error(ParseError::VariableNotFound(suggestion, span));
garbage(working_set, span)
2021-09-02 10:25:22 +02:00
}
}
2021-07-02 09:15:30 +02:00
2021-10-02 04:59:11 +02:00
pub fn parse_cell_path(
working_set: &mut StateWorkingSet,
tokens: impl Iterator<Item = Token>,
Optional members in cell paths: Attempt 2 (#8379) This is a follow up from https://github.com/nushell/nushell/pull/7540. Please provide feedback if you have the time! ## Summary This PR lets you use `?` to indicate that a member in a cell path is optional and Nushell should return `null` if that member cannot be accessed. Unlike the previous PR, `?` is now a _postfix_ modifier for cell path members. A cell path of `.foo?.bar` means that `foo` is optional and `bar` is not. `?` does _not_ suppress all errors; it is intended to help in situations where data has "holes", i.e. the data types are correct but something is missing. Type mismatches (like trying to do a string path access on a date) will still fail. ### Record Examples ```bash { foo: 123 }.foo # returns 123 { foo: 123 }.bar # errors { foo: 123 }.bar? # returns null { foo: 123 } | get bar # errors { foo: 123 } | get bar? # returns null { foo: 123 }.bar.baz # errors { foo: 123 }.bar?.baz # errors because `baz` is not present on the result from `bar?` { foo: 123 }.bar.baz? # errors { foo: 123 }.bar?.baz? # returns null ``` ### List Examples ``` 〉[{foo: 1} {foo: 2} {}].foo Error: nu::shell::column_not_found × Cannot find column ╭─[entry #30:1:1] 1 │ [{foo: 1} {foo: 2} {}].foo · ─┬ ─┬─ · │ ╰── cannot find column 'foo' · ╰── value originates here ╰──── 〉[{foo: 1} {foo: 2} {}].foo? ╭───┬───╮ │ 0 │ 1 │ │ 1 │ 2 │ │ 2 │ │ ╰───┴───╯ 〉[{foo: 1} {foo: 2} {}].foo?.2 | describe nothing 〉[a b c].4? | describe nothing 〉[{foo: 1} {foo: 2} {}] | where foo? == 1 ╭───┬─────╮ │ # │ foo │ ├───┼─────┤ │ 0 │ 1 │ ╰───┴─────╯ ``` # Breaking changes 1. Column names with `?` in them now need to be quoted. 2. The `-i`/`--ignore-errors` flag has been removed from `get` and `select` 1. After this PR, most `get` error handling can be done with `?` and/or `try`/`catch`. 4. Cell path accesses like this no longer work without a `?`: ```bash 〉[{a:1 b:2} {a:3}].b.0 2 ``` We had some clever code that was able to recognize that since we only want row `0`, it's OK if other rows are missing column `b`. I removed that because it's tricky to maintain, and now that query needs to be written like: ```bash 〉[{a:1 b:2} {a:3}].b?.0 2 ``` I think the regression is acceptable for now. I plan to do more work in the future to enable streaming of cell path accesses, and when that happens I'll be able to make `.b.0` work again.
2023-03-16 04:50:58 +01:00
expect_dot: bool,
) -> Vec<PathMember> {
Optional members in cell paths: Attempt 2 (#8379) This is a follow up from https://github.com/nushell/nushell/pull/7540. Please provide feedback if you have the time! ## Summary This PR lets you use `?` to indicate that a member in a cell path is optional and Nushell should return `null` if that member cannot be accessed. Unlike the previous PR, `?` is now a _postfix_ modifier for cell path members. A cell path of `.foo?.bar` means that `foo` is optional and `bar` is not. `?` does _not_ suppress all errors; it is intended to help in situations where data has "holes", i.e. the data types are correct but something is missing. Type mismatches (like trying to do a string path access on a date) will still fail. ### Record Examples ```bash { foo: 123 }.foo # returns 123 { foo: 123 }.bar # errors { foo: 123 }.bar? # returns null { foo: 123 } | get bar # errors { foo: 123 } | get bar? # returns null { foo: 123 }.bar.baz # errors { foo: 123 }.bar?.baz # errors because `baz` is not present on the result from `bar?` { foo: 123 }.bar.baz? # errors { foo: 123 }.bar?.baz? # returns null ``` ### List Examples ``` 〉[{foo: 1} {foo: 2} {}].foo Error: nu::shell::column_not_found × Cannot find column ╭─[entry #30:1:1] 1 │ [{foo: 1} {foo: 2} {}].foo · ─┬ ─┬─ · │ ╰── cannot find column 'foo' · ╰── value originates here ╰──── 〉[{foo: 1} {foo: 2} {}].foo? ╭───┬───╮ │ 0 │ 1 │ │ 1 │ 2 │ │ 2 │ │ ╰───┴───╯ 〉[{foo: 1} {foo: 2} {}].foo?.2 | describe nothing 〉[a b c].4? | describe nothing 〉[{foo: 1} {foo: 2} {}] | where foo? == 1 ╭───┬─────╮ │ # │ foo │ ├───┼─────┤ │ 0 │ 1 │ ╰───┴─────╯ ``` # Breaking changes 1. Column names with `?` in them now need to be quoted. 2. The `-i`/`--ignore-errors` flag has been removed from `get` and `select` 1. After this PR, most `get` error handling can be done with `?` and/or `try`/`catch`. 4. Cell path accesses like this no longer work without a `?`: ```bash 〉[{a:1 b:2} {a:3}].b.0 2 ``` We had some clever code that was able to recognize that since we only want row `0`, it's OK if other rows are missing column `b`. I removed that because it's tricky to maintain, and now that query needs to be written like: ```bash 〉[{a:1 b:2} {a:3}].b?.0 2 ``` I think the regression is acceptable for now. I plan to do more work in the future to enable streaming of cell path accesses, and when that happens I'll be able to make `.b.0` work again.
2023-03-16 04:50:58 +01:00
enum TokenType {
Dot, // .
QuestionOrDot, // ? or .
PathMember, // an int or string, like `1` or `foo`
}
// Parsing a cell path is essentially a state machine, and this is the state
let mut expected_token = if expect_dot {
TokenType::Dot
} else {
TokenType::PathMember
};
2021-10-02 04:59:11 +02:00
let mut tail = vec![];
for path_element in tokens {
let bytes = working_set.get_span_contents(path_element.span);
Optional members in cell paths: Attempt 2 (#8379) This is a follow up from https://github.com/nushell/nushell/pull/7540. Please provide feedback if you have the time! ## Summary This PR lets you use `?` to indicate that a member in a cell path is optional and Nushell should return `null` if that member cannot be accessed. Unlike the previous PR, `?` is now a _postfix_ modifier for cell path members. A cell path of `.foo?.bar` means that `foo` is optional and `bar` is not. `?` does _not_ suppress all errors; it is intended to help in situations where data has "holes", i.e. the data types are correct but something is missing. Type mismatches (like trying to do a string path access on a date) will still fail. ### Record Examples ```bash { foo: 123 }.foo # returns 123 { foo: 123 }.bar # errors { foo: 123 }.bar? # returns null { foo: 123 } | get bar # errors { foo: 123 } | get bar? # returns null { foo: 123 }.bar.baz # errors { foo: 123 }.bar?.baz # errors because `baz` is not present on the result from `bar?` { foo: 123 }.bar.baz? # errors { foo: 123 }.bar?.baz? # returns null ``` ### List Examples ``` 〉[{foo: 1} {foo: 2} {}].foo Error: nu::shell::column_not_found × Cannot find column ╭─[entry #30:1:1] 1 │ [{foo: 1} {foo: 2} {}].foo · ─┬ ─┬─ · │ ╰── cannot find column 'foo' · ╰── value originates here ╰──── 〉[{foo: 1} {foo: 2} {}].foo? ╭───┬───╮ │ 0 │ 1 │ │ 1 │ 2 │ │ 2 │ │ ╰───┴───╯ 〉[{foo: 1} {foo: 2} {}].foo?.2 | describe nothing 〉[a b c].4? | describe nothing 〉[{foo: 1} {foo: 2} {}] | where foo? == 1 ╭───┬─────╮ │ # │ foo │ ├───┼─────┤ │ 0 │ 1 │ ╰───┴─────╯ ``` # Breaking changes 1. Column names with `?` in them now need to be quoted. 2. The `-i`/`--ignore-errors` flag has been removed from `get` and `select` 1. After this PR, most `get` error handling can be done with `?` and/or `try`/`catch`. 4. Cell path accesses like this no longer work without a `?`: ```bash 〉[{a:1 b:2} {a:3}].b.0 2 ``` We had some clever code that was able to recognize that since we only want row `0`, it's OK if other rows are missing column `b`. I removed that because it's tricky to maintain, and now that query needs to be written like: ```bash 〉[{a:1 b:2} {a:3}].b?.0 2 ``` I think the regression is acceptable for now. I plan to do more work in the future to enable streaming of cell path accesses, and when that happens I'll be able to make `.b.0` work again.
2023-03-16 04:50:58 +01:00
match expected_token {
TokenType::Dot => {
if bytes.len() != 1 || bytes[0] != b'.' {
working_set.error(ParseError::Expected(".", path_element.span));
return tail;
Optional members in cell paths: Attempt 2 (#8379) This is a follow up from https://github.com/nushell/nushell/pull/7540. Please provide feedback if you have the time! ## Summary This PR lets you use `?` to indicate that a member in a cell path is optional and Nushell should return `null` if that member cannot be accessed. Unlike the previous PR, `?` is now a _postfix_ modifier for cell path members. A cell path of `.foo?.bar` means that `foo` is optional and `bar` is not. `?` does _not_ suppress all errors; it is intended to help in situations where data has "holes", i.e. the data types are correct but something is missing. Type mismatches (like trying to do a string path access on a date) will still fail. ### Record Examples ```bash { foo: 123 }.foo # returns 123 { foo: 123 }.bar # errors { foo: 123 }.bar? # returns null { foo: 123 } | get bar # errors { foo: 123 } | get bar? # returns null { foo: 123 }.bar.baz # errors { foo: 123 }.bar?.baz # errors because `baz` is not present on the result from `bar?` { foo: 123 }.bar.baz? # errors { foo: 123 }.bar?.baz? # returns null ``` ### List Examples ``` 〉[{foo: 1} {foo: 2} {}].foo Error: nu::shell::column_not_found × Cannot find column ╭─[entry #30:1:1] 1 │ [{foo: 1} {foo: 2} {}].foo · ─┬ ─┬─ · │ ╰── cannot find column 'foo' · ╰── value originates here ╰──── 〉[{foo: 1} {foo: 2} {}].foo? ╭───┬───╮ │ 0 │ 1 │ │ 1 │ 2 │ │ 2 │ │ ╰───┴───╯ 〉[{foo: 1} {foo: 2} {}].foo?.2 | describe nothing 〉[a b c].4? | describe nothing 〉[{foo: 1} {foo: 2} {}] | where foo? == 1 ╭───┬─────╮ │ # │ foo │ ├───┼─────┤ │ 0 │ 1 │ ╰───┴─────╯ ``` # Breaking changes 1. Column names with `?` in them now need to be quoted. 2. The `-i`/`--ignore-errors` flag has been removed from `get` and `select` 1. After this PR, most `get` error handling can be done with `?` and/or `try`/`catch`. 4. Cell path accesses like this no longer work without a `?`: ```bash 〉[{a:1 b:2} {a:3}].b.0 2 ``` We had some clever code that was able to recognize that since we only want row `0`, it's OK if other rows are missing column `b`. I removed that because it's tricky to maintain, and now that query needs to be written like: ```bash 〉[{a:1 b:2} {a:3}].b?.0 2 ``` I think the regression is acceptable for now. I plan to do more work in the future to enable streaming of cell path accesses, and when that happens I'll be able to make `.b.0` work again.
2023-03-16 04:50:58 +01:00
}
expected_token = TokenType::PathMember;
2021-10-02 04:59:11 +02:00
}
Optional members in cell paths: Attempt 2 (#8379) This is a follow up from https://github.com/nushell/nushell/pull/7540. Please provide feedback if you have the time! ## Summary This PR lets you use `?` to indicate that a member in a cell path is optional and Nushell should return `null` if that member cannot be accessed. Unlike the previous PR, `?` is now a _postfix_ modifier for cell path members. A cell path of `.foo?.bar` means that `foo` is optional and `bar` is not. `?` does _not_ suppress all errors; it is intended to help in situations where data has "holes", i.e. the data types are correct but something is missing. Type mismatches (like trying to do a string path access on a date) will still fail. ### Record Examples ```bash { foo: 123 }.foo # returns 123 { foo: 123 }.bar # errors { foo: 123 }.bar? # returns null { foo: 123 } | get bar # errors { foo: 123 } | get bar? # returns null { foo: 123 }.bar.baz # errors { foo: 123 }.bar?.baz # errors because `baz` is not present on the result from `bar?` { foo: 123 }.bar.baz? # errors { foo: 123 }.bar?.baz? # returns null ``` ### List Examples ``` 〉[{foo: 1} {foo: 2} {}].foo Error: nu::shell::column_not_found × Cannot find column ╭─[entry #30:1:1] 1 │ [{foo: 1} {foo: 2} {}].foo · ─┬ ─┬─ · │ ╰── cannot find column 'foo' · ╰── value originates here ╰──── 〉[{foo: 1} {foo: 2} {}].foo? ╭───┬───╮ │ 0 │ 1 │ │ 1 │ 2 │ │ 2 │ │ ╰───┴───╯ 〉[{foo: 1} {foo: 2} {}].foo?.2 | describe nothing 〉[a b c].4? | describe nothing 〉[{foo: 1} {foo: 2} {}] | where foo? == 1 ╭───┬─────╮ │ # │ foo │ ├───┼─────┤ │ 0 │ 1 │ ╰───┴─────╯ ``` # Breaking changes 1. Column names with `?` in them now need to be quoted. 2. The `-i`/`--ignore-errors` flag has been removed from `get` and `select` 1. After this PR, most `get` error handling can be done with `?` and/or `try`/`catch`. 4. Cell path accesses like this no longer work without a `?`: ```bash 〉[{a:1 b:2} {a:3}].b.0 2 ``` We had some clever code that was able to recognize that since we only want row `0`, it's OK if other rows are missing column `b`. I removed that because it's tricky to maintain, and now that query needs to be written like: ```bash 〉[{a:1 b:2} {a:3}].b?.0 2 ``` I think the regression is acceptable for now. I plan to do more work in the future to enable streaming of cell path accesses, and when that happens I'll be able to make `.b.0` work again.
2023-03-16 04:50:58 +01:00
TokenType::QuestionOrDot => {
if bytes.len() == 1 && bytes[0] == b'.' {
expected_token = TokenType::PathMember;
} else if bytes.len() == 1 && bytes[0] == b'?' {
if let Some(last) = tail.last_mut() {
match last {
PathMember::String {
ref mut optional, ..
} => *optional = true,
PathMember::Int {
ref mut optional, ..
} => *optional = true,
}
}
expected_token = TokenType::Dot;
} else {
working_set.error(ParseError::Expected(". or ?", path_element.span));
return tail;
Optional members in cell paths: Attempt 2 (#8379) This is a follow up from https://github.com/nushell/nushell/pull/7540. Please provide feedback if you have the time! ## Summary This PR lets you use `?` to indicate that a member in a cell path is optional and Nushell should return `null` if that member cannot be accessed. Unlike the previous PR, `?` is now a _postfix_ modifier for cell path members. A cell path of `.foo?.bar` means that `foo` is optional and `bar` is not. `?` does _not_ suppress all errors; it is intended to help in situations where data has "holes", i.e. the data types are correct but something is missing. Type mismatches (like trying to do a string path access on a date) will still fail. ### Record Examples ```bash { foo: 123 }.foo # returns 123 { foo: 123 }.bar # errors { foo: 123 }.bar? # returns null { foo: 123 } | get bar # errors { foo: 123 } | get bar? # returns null { foo: 123 }.bar.baz # errors { foo: 123 }.bar?.baz # errors because `baz` is not present on the result from `bar?` { foo: 123 }.bar.baz? # errors { foo: 123 }.bar?.baz? # returns null ``` ### List Examples ``` 〉[{foo: 1} {foo: 2} {}].foo Error: nu::shell::column_not_found × Cannot find column ╭─[entry #30:1:1] 1 │ [{foo: 1} {foo: 2} {}].foo · ─┬ ─┬─ · │ ╰── cannot find column 'foo' · ╰── value originates here ╰──── 〉[{foo: 1} {foo: 2} {}].foo? ╭───┬───╮ │ 0 │ 1 │ │ 1 │ 2 │ │ 2 │ │ ╰───┴───╯ 〉[{foo: 1} {foo: 2} {}].foo?.2 | describe nothing 〉[a b c].4? | describe nothing 〉[{foo: 1} {foo: 2} {}] | where foo? == 1 ╭───┬─────╮ │ # │ foo │ ├───┼─────┤ │ 0 │ 1 │ ╰───┴─────╯ ``` # Breaking changes 1. Column names with `?` in them now need to be quoted. 2. The `-i`/`--ignore-errors` flag has been removed from `get` and `select` 1. After this PR, most `get` error handling can be done with `?` and/or `try`/`catch`. 4. Cell path accesses like this no longer work without a `?`: ```bash 〉[{a:1 b:2} {a:3}].b.0 2 ``` We had some clever code that was able to recognize that since we only want row `0`, it's OK if other rows are missing column `b`. I removed that because it's tricky to maintain, and now that query needs to be written like: ```bash 〉[{a:1 b:2} {a:3}].b?.0 2 ``` I think the regression is acceptable for now. I plan to do more work in the future to enable streaming of cell path accesses, and when that happens I'll be able to make `.b.0` work again.
2023-03-16 04:50:58 +01:00
}
}
TokenType::PathMember => {
let starting_error_count = working_set.parse_errors.len();
let expr = parse_int(working_set, path_element.span);
working_set.parse_errors.truncate(starting_error_count);
match expr {
Expression {
expr: Expr::Int(val),
span,
..
} => tail.push(PathMember::Int {
Optional members in cell paths: Attempt 2 (#8379) This is a follow up from https://github.com/nushell/nushell/pull/7540. Please provide feedback if you have the time! ## Summary This PR lets you use `?` to indicate that a member in a cell path is optional and Nushell should return `null` if that member cannot be accessed. Unlike the previous PR, `?` is now a _postfix_ modifier for cell path members. A cell path of `.foo?.bar` means that `foo` is optional and `bar` is not. `?` does _not_ suppress all errors; it is intended to help in situations where data has "holes", i.e. the data types are correct but something is missing. Type mismatches (like trying to do a string path access on a date) will still fail. ### Record Examples ```bash { foo: 123 }.foo # returns 123 { foo: 123 }.bar # errors { foo: 123 }.bar? # returns null { foo: 123 } | get bar # errors { foo: 123 } | get bar? # returns null { foo: 123 }.bar.baz # errors { foo: 123 }.bar?.baz # errors because `baz` is not present on the result from `bar?` { foo: 123 }.bar.baz? # errors { foo: 123 }.bar?.baz? # returns null ``` ### List Examples ``` 〉[{foo: 1} {foo: 2} {}].foo Error: nu::shell::column_not_found × Cannot find column ╭─[entry #30:1:1] 1 │ [{foo: 1} {foo: 2} {}].foo · ─┬ ─┬─ · │ ╰── cannot find column 'foo' · ╰── value originates here ╰──── 〉[{foo: 1} {foo: 2} {}].foo? ╭───┬───╮ │ 0 │ 1 │ │ 1 │ 2 │ │ 2 │ │ ╰───┴───╯ 〉[{foo: 1} {foo: 2} {}].foo?.2 | describe nothing 〉[a b c].4? | describe nothing 〉[{foo: 1} {foo: 2} {}] | where foo? == 1 ╭───┬─────╮ │ # │ foo │ ├───┼─────┤ │ 0 │ 1 │ ╰───┴─────╯ ``` # Breaking changes 1. Column names with `?` in them now need to be quoted. 2. The `-i`/`--ignore-errors` flag has been removed from `get` and `select` 1. After this PR, most `get` error handling can be done with `?` and/or `try`/`catch`. 4. Cell path accesses like this no longer work without a `?`: ```bash 〉[{a:1 b:2} {a:3}].b.0 2 ``` We had some clever code that was able to recognize that since we only want row `0`, it's OK if other rows are missing column `b`. I removed that because it's tricky to maintain, and now that query needs to be written like: ```bash 〉[{a:1 b:2} {a:3}].b?.0 2 ``` I think the regression is acceptable for now. I plan to do more work in the future to enable streaming of cell path accesses, and when that happens I'll be able to make `.b.0` work again.
2023-03-16 04:50:58 +01:00
val: val as usize,
span,
optional: false,
}),
_ => {
let result = parse_string(working_set, path_element.span);
Optional members in cell paths: Attempt 2 (#8379) This is a follow up from https://github.com/nushell/nushell/pull/7540. Please provide feedback if you have the time! ## Summary This PR lets you use `?` to indicate that a member in a cell path is optional and Nushell should return `null` if that member cannot be accessed. Unlike the previous PR, `?` is now a _postfix_ modifier for cell path members. A cell path of `.foo?.bar` means that `foo` is optional and `bar` is not. `?` does _not_ suppress all errors; it is intended to help in situations where data has "holes", i.e. the data types are correct but something is missing. Type mismatches (like trying to do a string path access on a date) will still fail. ### Record Examples ```bash { foo: 123 }.foo # returns 123 { foo: 123 }.bar # errors { foo: 123 }.bar? # returns null { foo: 123 } | get bar # errors { foo: 123 } | get bar? # returns null { foo: 123 }.bar.baz # errors { foo: 123 }.bar?.baz # errors because `baz` is not present on the result from `bar?` { foo: 123 }.bar.baz? # errors { foo: 123 }.bar?.baz? # returns null ``` ### List Examples ``` 〉[{foo: 1} {foo: 2} {}].foo Error: nu::shell::column_not_found × Cannot find column ╭─[entry #30:1:1] 1 │ [{foo: 1} {foo: 2} {}].foo · ─┬ ─┬─ · │ ╰── cannot find column 'foo' · ╰── value originates here ╰──── 〉[{foo: 1} {foo: 2} {}].foo? ╭───┬───╮ │ 0 │ 1 │ │ 1 │ 2 │ │ 2 │ │ ╰───┴───╯ 〉[{foo: 1} {foo: 2} {}].foo?.2 | describe nothing 〉[a b c].4? | describe nothing 〉[{foo: 1} {foo: 2} {}] | where foo? == 1 ╭───┬─────╮ │ # │ foo │ ├───┼─────┤ │ 0 │ 1 │ ╰───┴─────╯ ``` # Breaking changes 1. Column names with `?` in them now need to be quoted. 2. The `-i`/`--ignore-errors` flag has been removed from `get` and `select` 1. After this PR, most `get` error handling can be done with `?` and/or `try`/`catch`. 4. Cell path accesses like this no longer work without a `?`: ```bash 〉[{a:1 b:2} {a:3}].b.0 2 ``` We had some clever code that was able to recognize that since we only want row `0`, it's OK if other rows are missing column `b`. I removed that because it's tricky to maintain, and now that query needs to be written like: ```bash 〉[{a:1 b:2} {a:3}].b?.0 2 ``` I think the regression is acceptable for now. I plan to do more work in the future to enable streaming of cell path accesses, and when that happens I'll be able to make `.b.0` work again.
2023-03-16 04:50:58 +01:00
match result {
Expression {
expr: Expr::String(string),
span,
..
} => {
tail.push(PathMember::String {
val: string,
span,
optional: false,
});
}
_ => {
working_set
.error(ParseError::Expected("string", path_element.span));
return tail;
Optional members in cell paths: Attempt 2 (#8379) This is a follow up from https://github.com/nushell/nushell/pull/7540. Please provide feedback if you have the time! ## Summary This PR lets you use `?` to indicate that a member in a cell path is optional and Nushell should return `null` if that member cannot be accessed. Unlike the previous PR, `?` is now a _postfix_ modifier for cell path members. A cell path of `.foo?.bar` means that `foo` is optional and `bar` is not. `?` does _not_ suppress all errors; it is intended to help in situations where data has "holes", i.e. the data types are correct but something is missing. Type mismatches (like trying to do a string path access on a date) will still fail. ### Record Examples ```bash { foo: 123 }.foo # returns 123 { foo: 123 }.bar # errors { foo: 123 }.bar? # returns null { foo: 123 } | get bar # errors { foo: 123 } | get bar? # returns null { foo: 123 }.bar.baz # errors { foo: 123 }.bar?.baz # errors because `baz` is not present on the result from `bar?` { foo: 123 }.bar.baz? # errors { foo: 123 }.bar?.baz? # returns null ``` ### List Examples ``` 〉[{foo: 1} {foo: 2} {}].foo Error: nu::shell::column_not_found × Cannot find column ╭─[entry #30:1:1] 1 │ [{foo: 1} {foo: 2} {}].foo · ─┬ ─┬─ · │ ╰── cannot find column 'foo' · ╰── value originates here ╰──── 〉[{foo: 1} {foo: 2} {}].foo? ╭───┬───╮ │ 0 │ 1 │ │ 1 │ 2 │ │ 2 │ │ ╰───┴───╯ 〉[{foo: 1} {foo: 2} {}].foo?.2 | describe nothing 〉[a b c].4? | describe nothing 〉[{foo: 1} {foo: 2} {}] | where foo? == 1 ╭───┬─────╮ │ # │ foo │ ├───┼─────┤ │ 0 │ 1 │ ╰───┴─────╯ ``` # Breaking changes 1. Column names with `?` in them now need to be quoted. 2. The `-i`/`--ignore-errors` flag has been removed from `get` and `select` 1. After this PR, most `get` error handling can be done with `?` and/or `try`/`catch`. 4. Cell path accesses like this no longer work without a `?`: ```bash 〉[{a:1 b:2} {a:3}].b.0 2 ``` We had some clever code that was able to recognize that since we only want row `0`, it's OK if other rows are missing column `b`. I removed that because it's tricky to maintain, and now that query needs to be written like: ```bash 〉[{a:1 b:2} {a:3}].b?.0 2 ``` I think the regression is acceptable for now. I plan to do more work in the future to enable streaming of cell path accesses, and when that happens I'll be able to make `.b.0` work again.
2023-03-16 04:50:58 +01:00
}
2021-10-02 04:59:11 +02:00
}
}
}
Optional members in cell paths: Attempt 2 (#8379) This is a follow up from https://github.com/nushell/nushell/pull/7540. Please provide feedback if you have the time! ## Summary This PR lets you use `?` to indicate that a member in a cell path is optional and Nushell should return `null` if that member cannot be accessed. Unlike the previous PR, `?` is now a _postfix_ modifier for cell path members. A cell path of `.foo?.bar` means that `foo` is optional and `bar` is not. `?` does _not_ suppress all errors; it is intended to help in situations where data has "holes", i.e. the data types are correct but something is missing. Type mismatches (like trying to do a string path access on a date) will still fail. ### Record Examples ```bash { foo: 123 }.foo # returns 123 { foo: 123 }.bar # errors { foo: 123 }.bar? # returns null { foo: 123 } | get bar # errors { foo: 123 } | get bar? # returns null { foo: 123 }.bar.baz # errors { foo: 123 }.bar?.baz # errors because `baz` is not present on the result from `bar?` { foo: 123 }.bar.baz? # errors { foo: 123 }.bar?.baz? # returns null ``` ### List Examples ``` 〉[{foo: 1} {foo: 2} {}].foo Error: nu::shell::column_not_found × Cannot find column ╭─[entry #30:1:1] 1 │ [{foo: 1} {foo: 2} {}].foo · ─┬ ─┬─ · │ ╰── cannot find column 'foo' · ╰── value originates here ╰──── 〉[{foo: 1} {foo: 2} {}].foo? ╭───┬───╮ │ 0 │ 1 │ │ 1 │ 2 │ │ 2 │ │ ╰───┴───╯ 〉[{foo: 1} {foo: 2} {}].foo?.2 | describe nothing 〉[a b c].4? | describe nothing 〉[{foo: 1} {foo: 2} {}] | where foo? == 1 ╭───┬─────╮ │ # │ foo │ ├───┼─────┤ │ 0 │ 1 │ ╰───┴─────╯ ``` # Breaking changes 1. Column names with `?` in them now need to be quoted. 2. The `-i`/`--ignore-errors` flag has been removed from `get` and `select` 1. After this PR, most `get` error handling can be done with `?` and/or `try`/`catch`. 4. Cell path accesses like this no longer work without a `?`: ```bash 〉[{a:1 b:2} {a:3}].b.0 2 ``` We had some clever code that was able to recognize that since we only want row `0`, it's OK if other rows are missing column `b`. I removed that because it's tricky to maintain, and now that query needs to be written like: ```bash 〉[{a:1 b:2} {a:3}].b?.0 2 ``` I think the regression is acceptable for now. I plan to do more work in the future to enable streaming of cell path accesses, and when that happens I'll be able to make `.b.0` work again.
2023-03-16 04:50:58 +01:00
expected_token = TokenType::QuestionOrDot;
2021-10-02 04:59:11 +02:00
}
}
}
tail
2021-10-02 04:59:11 +02:00
}
pub fn parse_simple_cell_path(working_set: &mut StateWorkingSet, span: Span) -> Expression {
let source = working_set.get_span_contents(span);
let (tokens, err) = lex(source, span.start, &[b'\n', b'\r'], &[b'.', b'?'], true);
if let Some(err) = err {
working_set.error(err)
}
let tokens = tokens.into_iter().peekable();
let cell_path = parse_cell_path(working_set, tokens, false);
Expression::new(
working_set,
Expr::CellPath(CellPath { members: cell_path }),
span,
Type::CellPath,
)
}
2021-09-26 20:39:19 +02:00
pub fn parse_full_cell_path(
2021-09-02 10:25:22 +02:00
working_set: &mut StateWorkingSet,
2021-09-09 23:47:20 +02:00
implicit_head: Option<VarId>,
2021-09-02 10:25:22 +02:00
span: Span,
) -> Expression {
trace!("parsing: full cell path");
2021-09-26 20:39:19 +02:00
let full_cell_span = span;
2021-09-07 00:02:24 +02:00
let source = working_set.get_span_contents(span);
2021-07-08 08:19:38 +02:00
Optional members in cell paths: Attempt 2 (#8379) This is a follow up from https://github.com/nushell/nushell/pull/7540. Please provide feedback if you have the time! ## Summary This PR lets you use `?` to indicate that a member in a cell path is optional and Nushell should return `null` if that member cannot be accessed. Unlike the previous PR, `?` is now a _postfix_ modifier for cell path members. A cell path of `.foo?.bar` means that `foo` is optional and `bar` is not. `?` does _not_ suppress all errors; it is intended to help in situations where data has "holes", i.e. the data types are correct but something is missing. Type mismatches (like trying to do a string path access on a date) will still fail. ### Record Examples ```bash { foo: 123 }.foo # returns 123 { foo: 123 }.bar # errors { foo: 123 }.bar? # returns null { foo: 123 } | get bar # errors { foo: 123 } | get bar? # returns null { foo: 123 }.bar.baz # errors { foo: 123 }.bar?.baz # errors because `baz` is not present on the result from `bar?` { foo: 123 }.bar.baz? # errors { foo: 123 }.bar?.baz? # returns null ``` ### List Examples ``` 〉[{foo: 1} {foo: 2} {}].foo Error: nu::shell::column_not_found × Cannot find column ╭─[entry #30:1:1] 1 │ [{foo: 1} {foo: 2} {}].foo · ─┬ ─┬─ · │ ╰── cannot find column 'foo' · ╰── value originates here ╰──── 〉[{foo: 1} {foo: 2} {}].foo? ╭───┬───╮ │ 0 │ 1 │ │ 1 │ 2 │ │ 2 │ │ ╰───┴───╯ 〉[{foo: 1} {foo: 2} {}].foo?.2 | describe nothing 〉[a b c].4? | describe nothing 〉[{foo: 1} {foo: 2} {}] | where foo? == 1 ╭───┬─────╮ │ # │ foo │ ├───┼─────┤ │ 0 │ 1 │ ╰───┴─────╯ ``` # Breaking changes 1. Column names with `?` in them now need to be quoted. 2. The `-i`/`--ignore-errors` flag has been removed from `get` and `select` 1. After this PR, most `get` error handling can be done with `?` and/or `try`/`catch`. 4. Cell path accesses like this no longer work without a `?`: ```bash 〉[{a:1 b:2} {a:3}].b.0 2 ``` We had some clever code that was able to recognize that since we only want row `0`, it's OK if other rows are missing column `b`. I removed that because it's tricky to maintain, and now that query needs to be written like: ```bash 〉[{a:1 b:2} {a:3}].b?.0 2 ``` I think the regression is acceptable for now. I plan to do more work in the future to enable streaming of cell path accesses, and when that happens I'll be able to make `.b.0` work again.
2023-03-16 04:50:58 +01:00
let (tokens, err) = lex(source, span.start, &[b'\n', b'\r'], &[b'.', b'?'], true);
if let Some(err) = err {
working_set.error(err)
}
2021-09-02 10:25:22 +02:00
2021-09-09 23:47:20 +02:00
let mut tokens = tokens.into_iter().peekable();
if let Some(head) = tokens.peek() {
2021-09-07 00:02:24 +02:00
let bytes = working_set.get_span_contents(head.span);
2021-10-02 04:59:11 +02:00
let (head, expect_dot) = if bytes.starts_with(b"(") {
2022-01-01 22:42:50 +01:00
trace!("parsing: paren-head of full cell path");
let head_span = head.span;
2021-09-07 00:02:24 +02:00
let mut start = head.span.start;
let mut end = head.span.end;
2021-07-08 08:19:38 +02:00
2021-09-07 00:02:24 +02:00
if bytes.starts_with(b"(") {
start += 1;
}
if bytes.ends_with(b")") {
end -= 1;
} else {
working_set.error(ParseError::Unclosed(")".into(), Span::new(end, end)));
2021-09-07 00:02:24 +02:00
}
2021-07-02 09:15:30 +02:00
Protocol: debug_assert!() Span to reflect a valid slice (#6806) Also enforce this by #[non_exhaustive] span such that going forward we cannot, in debug builds (1), construct invalid spans. The motivation for this stems from #6431 where I've seen crashes due to invalid slice indexing. My hope is this will mitigate such senarios 1. https://github.com/nushell/nushell/pull/6431#issuecomment-1278147241 # Description (description of your pull request here) # Tests Make sure you've done the following: - [ ] Add tests that cover your changes, either in the command examples, the crate/tests folder, or in the /tests folder. - [ ] Try to think about corner cases and various ways how your changes could break. Cover them with tests. - [ ] If adding tests is not possible, please document in the PR body a minimal example with steps on how to reproduce so one can verify your change works. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [ ] `cargo clippy --workspace --features=extra -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [ ] `cargo test --workspace --features=extra` to check that all the tests pass # Documentation - [ ] If your PR touches a user-facing nushell feature then make sure that there is an entry in the documentation (https://github.com/nushell/nushell.github.io) for the feature, and update it if necessary.
2022-12-03 10:44:12 +01:00
let span = Span::new(start, end);
2021-07-02 09:15:30 +02:00
2021-09-07 00:02:24 +02:00
let source = working_set.get_span_contents(span);
2021-07-02 09:15:30 +02:00
2021-11-21 19:13:09 +01:00
let (output, err) = lex(source, span.start, &[b'\n', b'\r'], &[], true);
if let Some(err) = err {
working_set.error(err)
}
2021-07-02 09:15:30 +02:00
// Creating a Type scope to parse the new block. This will keep track of
// the previous input type found in that block
Reuse the cached parse results of parsed files (#8949) # Description This does a lookup in the cache of parsed files to see if a span can be found for a file that was previously loaded with the same contents, then uses that span to find the parsed block for that file. The end result should, in theory, be identical but doesn't require any reparsing or creating new blocks/new definitions that aren't needed. This drops the sg.nu benchmark from: ``` ╭───┬───────────────────╮ │ 0 │ 280ms 606µs 208ns │ │ 1 │ 282ms 654µs 416ns │ │ 2 │ 252ms 640µs 541ns │ │ 3 │ 250ms 940µs 41ns │ │ 4 │ 241ms 216µs 375ns │ │ 5 │ 257ms 310µs 583ns │ │ 6 │ 196ms 739µs 416ns │ ╰───┴───────────────────╯ ``` to: ``` ╭───┬───────────────────╮ │ 0 │ 118ms 698µs 125ns │ │ 1 │ 121ms 327µs │ │ 2 │ 121ms 873µs 500ns │ │ 3 │ 124ms 94µs 708ns │ │ 4 │ 113ms 733µs 291ns │ │ 5 │ 108ms 663µs 125ns │ │ 6 │ 63ms 482µs 625ns │ ╰───┴───────────────────╯ ``` I was hoping to also see some startup time improvements, but I didn't notice much there. # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-04-21 21:00:33 +02:00
let output = parse_block(working_set, &output, span, true, true);
Improve type hovers (#9515) # Description This PR does a few things to help improve type hovers and, in the process, fixes a few outstanding issues in the type system. Here's a list of the changes: * `for` now will try to infer the type of the iteration variable based on the expression it's given. This fixes things like `for x in [1, 2, 3] { }` where `x` now properly gets the int type. * Removed old input/output type fields from the signature, focuses on the vec of signatures. Updated a bunch of dataframe commands that hadn't moved over. This helps tie things together a bit better * Fixed inference of types from subexpressions to use the last expression in the block * Fixed handling of explicit types in `let` and `mut` calls, so we now respect that as the authoritative type I also tried to add `def` input/output type inference, but unfortunately we only know the predecl types universally, which means we won't have enough information to properly know what the types of the custom commands are. # User-Facing Changes Script typechecking will get tighter in some cases Hovers should be more accurate in some cases that previously resorted to any. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: Darren Schroeder <343840+fdncred@users.noreply.github.com>
2023-06-28 19:19:48 +02:00
let ty = output.output_type();
Make EngineState clone cheaper with Arc on all of the heavy objects (#12229) # Description This makes many of the larger objects in `EngineState` into `Arc`, and uses `Arc::make_mut` to do clone-on-write if the reference is not unique. This is generally very cheap, giving us the best of both worlds - allowing us to mutate without cloning if we have an exclusive reference, and cloning if we don't. This started as more of a curiosity for me after remembering that `Arc::make_mut` exists and can make using `Arc` for mostly immutable data that sometimes needs to be changed very convenient, and also after hearing someone complain about memory usage on Discord - this is a somewhat significant win for that. The exact objects that were wrapped in `Arc`: - `files`, `file_contents` - the strings and byte buffers - `decls` - the whole `Vec`, but mostly to avoid lots of individual `malloc()` calls on Clone rather than for memory usage - `blocks` - the blocks themselves, rather than the outer Vec - `modules` - the modules themselves, rather than the outer Vec - `env_vars`, `previous_env_vars` - the entire maps - `config` The changes required were relatively minimal, but this is a breaking API change. In particular, blocks are added as Arcs, to allow the parser cache functionality to work. With my normal nu config, running on Linux, this saves me about 15 MiB of process memory usage when running interactively (65 MiB → 50 MiB). This also makes quick command executions cheaper, particularly since every REPL loop now involves a clone of the engine state so that we can recover from a panic. It also reduces memory usage where engine state needs to be cloned and sent to another thread or kept within an iterator. # User-Facing Changes Shouldn't be any, since it's all internal stuff, but it does change some public interfaces so it's a breaking change
2024-03-19 19:07:00 +01:00
let block_id = working_set.add_block(Arc::new(output));
2021-09-09 23:47:20 +02:00
tokens.next();
2021-09-07 00:02:24 +02:00
2021-09-09 23:47:20 +02:00
(
Expression::new(working_set, Expr::Subexpression(block_id), head_span, ty),
2021-09-09 23:47:20 +02:00
true,
)
} else if bytes.starts_with(b"[") {
2022-01-01 22:42:50 +01:00
trace!("parsing: table head of full cell path");
let output = parse_table_expression(working_set, head.span);
tokens.next();
2021-11-11 00:14:00 +01:00
(output, true)
} else if bytes.starts_with(b"{") {
2022-01-01 22:42:50 +01:00
trace!("parsing: record head of full cell path");
let output = parse_record(working_set, head.span);
2021-11-11 00:14:00 +01:00
tokens.next();
(output, true)
2021-09-07 00:02:24 +02:00
} else if bytes.starts_with(b"$") {
2022-01-01 22:42:50 +01:00
trace!("parsing: $variable head of full cell path");
let out = parse_variable_expr(working_set, head.span);
2021-09-07 00:02:24 +02:00
2021-09-09 23:47:20 +02:00
tokens.next();
(out, true)
} else if let Some(var_id) = implicit_head {
trace!("parsing: implicit head of full cell path");
2021-09-09 23:47:20 +02:00
(
Expression::new(working_set, Expr::Var(var_id), head.span, Type::Any),
2021-09-09 23:47:20 +02:00
false,
)
2021-09-07 00:02:24 +02:00
} else {
working_set.error(ParseError::Mismatch(
"variable or subexpression".into(),
String::from_utf8_lossy(bytes).to_string(),
span,
));
return garbage(working_set, span);
2021-09-07 00:02:24 +02:00
};
let tail = parse_cell_path(working_set, tokens, expect_dot);
// FIXME: Get the type of the data at the tail using follow_cell_path() (or something)
let ty = if !tail.is_empty() {
// Until the aforementioned fix is implemented, this is necessary to allow mutable list upserts
// such as $a.1 = 2 to work correctly.
Type::Any
} else {
head.ty.clone()
};
2021-09-07 00:02:24 +02:00
Expression::new(
working_set,
Expr::FullCellPath(Box::new(FullCellPath { head, tail })),
full_cell_span,
ty,
)
2021-09-07 00:02:24 +02:00
} else {
garbage(working_set, span)
2021-09-07 00:02:24 +02:00
}
2021-09-02 10:25:22 +02:00
}
2021-07-02 09:15:30 +02:00
pub fn parse_directory(working_set: &mut StateWorkingSet, span: Span) -> Expression {
let bytes = working_set.get_span_contents(span);
do not attempt to glob expand if the file path is wrapped in quotes (#11569) # Description Fixes: #11455 ### For arguments which is annotated with `:path/:directory/:glob` To fix the issue, we need to have a way to know if a path is originally quoted during runtime. So the information needed to be added at several levels: * parse time (from user input to expression) We need to add quoted information into `Expr::Filepath`, `Expr::Directory`, `Expr::GlobPattern` * eval time When convert from `Expr::Filepath`, `Expr::Directory`, `Expr::GlobPattern` to `Value::String` during runtime, we won't auto expanded the path if it's quoted ### For `ls` It's really special, because it accepts a `String` as a pattern, and it generates `glob` expression inside the command itself. So the idea behind the change is introducing a special SyntaxShape to ls: `SyntaxShape::LsGlobPattern`. So we can track if the pattern is originally quoted easier, and we don't auto expand the path either. Then when constructing a glob pattern inside ls, we check if input pattern is quoted, if so: we escape the input pattern, so we can run `ls a[123]b`, because it's already escaped. Finally, to accomplish the checking process, we also need to introduce a new value type called `Value::QuotedString` to differ from `Value::String`, it's used to generate an enum called `NuPath`, which is finally used in `ls` function. `ls` learned from `NuPath` to know if user input is quoted. # User-Facing Changes Actually it contains several changes ### For arguments which is annotated with `:path/:directory/:glob` #### Before ```nushell > def foo [p: path] { echo $p }; print (foo "~/a"); print (foo '~/a') /home/windsoilder/a /home/windsoilder/a > def foo [p: directory] { echo $p }; print (foo "~/a"); print (foo '~/a') /home/windsoilder/a /home/windsoilder/a > def foo [p: glob] { echo $p }; print (foo "~/a"); print (foo '~/a') /home/windsoilder/a /home/windsoilder/a ``` #### After ```nushell > def foo [p: path] { echo $p }; print (foo "~/a"); print (foo '~/a') ~/a ~/a > def foo [p: directory] { echo $p }; print (foo "~/a"); print (foo '~/a') ~/a ~/a > def foo [p: glob] { echo $p }; print (foo "~/a"); print (foo '~/a') ~/a ~/a ``` ### For ls command `touch '[uwu]'` #### Before ``` ❯ ls -D "[uwu]" Error: × No matches found for [uwu] ╭─[entry #6:1:1] 1 │ ls -D "[uwu]" · ───┬─── · ╰── Pattern, file or folder not found ╰──── help: no matches found ``` #### After ``` ❯ ls -D "[uwu]" ╭───┬───────┬──────┬──────┬──────────╮ │ # │ name │ type │ size │ modified │ ├───┼───────┼──────┼──────┼──────────┤ │ 0 │ [uwu] │ file │ 0 B │ now │ ╰───┴───────┴──────┴──────┴──────────╯ ``` # Tests + Formatting Done # After Submitting NaN
2024-01-21 16:22:25 +01:00
let quoted = is_quoted(bytes);
let (token, err) = unescape_unquote_string(bytes, span);
trace!("parsing: directory");
if err.is_none() {
trace!("-- found {}", token);
Expression::new(
working_set,
Expr::Directory(token, quoted),
span,
Type::String,
)
} else {
working_set.error(ParseError::Expected("directory", span));
garbage(working_set, span)
}
}
pub fn parse_filepath(working_set: &mut StateWorkingSet, span: Span) -> Expression {
2021-10-04 21:21:31 +02:00
let bytes = working_set.get_span_contents(span);
do not attempt to glob expand if the file path is wrapped in quotes (#11569) # Description Fixes: #11455 ### For arguments which is annotated with `:path/:directory/:glob` To fix the issue, we need to have a way to know if a path is originally quoted during runtime. So the information needed to be added at several levels: * parse time (from user input to expression) We need to add quoted information into `Expr::Filepath`, `Expr::Directory`, `Expr::GlobPattern` * eval time When convert from `Expr::Filepath`, `Expr::Directory`, `Expr::GlobPattern` to `Value::String` during runtime, we won't auto expanded the path if it's quoted ### For `ls` It's really special, because it accepts a `String` as a pattern, and it generates `glob` expression inside the command itself. So the idea behind the change is introducing a special SyntaxShape to ls: `SyntaxShape::LsGlobPattern`. So we can track if the pattern is originally quoted easier, and we don't auto expand the path either. Then when constructing a glob pattern inside ls, we check if input pattern is quoted, if so: we escape the input pattern, so we can run `ls a[123]b`, because it's already escaped. Finally, to accomplish the checking process, we also need to introduce a new value type called `Value::QuotedString` to differ from `Value::String`, it's used to generate an enum called `NuPath`, which is finally used in `ls` function. `ls` learned from `NuPath` to know if user input is quoted. # User-Facing Changes Actually it contains several changes ### For arguments which is annotated with `:path/:directory/:glob` #### Before ```nushell > def foo [p: path] { echo $p }; print (foo "~/a"); print (foo '~/a') /home/windsoilder/a /home/windsoilder/a > def foo [p: directory] { echo $p }; print (foo "~/a"); print (foo '~/a') /home/windsoilder/a /home/windsoilder/a > def foo [p: glob] { echo $p }; print (foo "~/a"); print (foo '~/a') /home/windsoilder/a /home/windsoilder/a ``` #### After ```nushell > def foo [p: path] { echo $p }; print (foo "~/a"); print (foo '~/a') ~/a ~/a > def foo [p: directory] { echo $p }; print (foo "~/a"); print (foo '~/a') ~/a ~/a > def foo [p: glob] { echo $p }; print (foo "~/a"); print (foo '~/a') ~/a ~/a ``` ### For ls command `touch '[uwu]'` #### Before ``` ❯ ls -D "[uwu]" Error: × No matches found for [uwu] ╭─[entry #6:1:1] 1 │ ls -D "[uwu]" · ───┬─── · ╰── Pattern, file or folder not found ╰──── help: no matches found ``` #### After ``` ❯ ls -D "[uwu]" ╭───┬───────┬──────┬──────┬──────────╮ │ # │ name │ type │ size │ modified │ ├───┼───────┼──────┼──────┼──────────┤ │ 0 │ [uwu] │ file │ 0 B │ now │ ╰───┴───────┴──────┴──────┴──────────╯ ``` # Tests + Formatting Done # After Submitting NaN
2024-01-21 16:22:25 +01:00
let quoted = is_quoted(bytes);
let (token, err) = unescape_unquote_string(bytes, span);
2022-01-01 22:42:50 +01:00
trace!("parsing: filepath");
2021-10-04 21:21:31 +02:00
if err.is_none() {
trace!("-- found {}", token);
Expression::new(
working_set,
Expr::Filepath(token, quoted),
span,
Type::String,
)
2021-10-04 21:21:31 +02:00
} else {
working_set.error(ParseError::Expected("filepath", span));
garbage(working_set, span)
2021-10-04 21:21:31 +02:00
}
}
/// Parse a datetime type, eg '2022-02-02'
pub fn parse_datetime(working_set: &mut StateWorkingSet, span: Span) -> Expression {
trace!("parsing: datetime");
let bytes = working_set.get_span_contents(span);
if bytes.len() < 6
|| !bytes[0].is_ascii_digit()
|| !bytes[1].is_ascii_digit()
|| !bytes[2].is_ascii_digit()
|| !bytes[3].is_ascii_digit()
|| bytes[4] != b'-'
{
working_set.error(ParseError::Expected("datetime", span));
return garbage(working_set, span);
}
let token = String::from_utf8_lossy(bytes).to_string();
if let Ok(datetime) = chrono::DateTime::parse_from_rfc3339(&token) {
return Expression::new(working_set, Expr::DateTime(datetime), span, Type::Date);
}
// Just the date
let just_date = token.clone() + "T00:00:00+00:00";
if let Ok(datetime) = chrono::DateTime::parse_from_rfc3339(&just_date) {
return Expression::new(working_set, Expr::DateTime(datetime), span, Type::Date);
}
// Date and time, assume UTC
let datetime = token + "+00:00";
if let Ok(datetime) = chrono::DateTime::parse_from_rfc3339(&datetime) {
return Expression::new(working_set, Expr::DateTime(datetime), span, Type::Date);
}
working_set.error(ParseError::Expected("datetime", span));
garbage(working_set, span)
}
2021-10-05 04:27:39 +02:00
/// Parse a duration type, eg '10day'
pub fn parse_duration(working_set: &mut StateWorkingSet, span: Span) -> Expression {
2022-01-01 22:42:50 +01:00
trace!("parsing: duration");
let bytes = working_set.get_span_contents(span);
match parse_unit_value(bytes, span, DURATION_UNIT_GROUPS, Type::Duration, |x| x) {
Some(Ok(expr)) => {
let span_id = working_set.add_span(span);
expr.with_span_id(span_id)
}
Some(Err(mk_err_for)) => {
working_set.error(mk_err_for("duration"));
garbage(working_set, span)
}
None => {
working_set.error(ParseError::Expected("duration with valid units", span));
garbage(working_set, span)
}
}
}
/// Parse a unit type, eg '10kb'
pub fn parse_filesize(working_set: &mut StateWorkingSet, span: Span) -> Expression {
trace!("parsing: filesize");
let bytes = working_set.get_span_contents(span);
// the hex digit `b` might be mistaken for the unit `b`, so check that first
if bytes.starts_with(b"0x") {
working_set.error(ParseError::Expected("filesize with valid units", span));
return garbage(working_set, span);
}
match parse_unit_value(bytes, span, FILESIZE_UNIT_GROUPS, Type::Filesize, |x| {
x.to_ascii_uppercase()
}) {
Some(Ok(expr)) => {
let span_id = working_set.add_span(span);
expr.with_span_id(span_id)
}
Some(Err(mk_err_for)) => {
working_set.error(mk_err_for("filesize"));
garbage(working_set, span)
}
None => {
working_set.error(ParseError::Expected("filesize with valid units", span));
garbage(working_set, span)
}
}
}
type ParseUnitResult<'res> = Result<Expression, Box<dyn Fn(&'res str) -> ParseError>>;
type UnitGroup<'unit> = (Unit, &'unit str, Option<(Unit, i64)>);
pub fn parse_unit_value<'res>(
bytes: &[u8],
span: Span,
unit_groups: &[UnitGroup],
ty: Type,
transform: fn(String) -> String,
) -> Option<ParseUnitResult<'res>> {
if bytes.len() < 2
|| !(bytes[0].is_ascii_digit() || (bytes[0] == b'-' && bytes[1].is_ascii_digit()))
{
return None;
}
let value = transform(String::from_utf8_lossy(bytes).into());
if let Some((unit, name, convert)) = unit_groups.iter().find(|x| value.ends_with(x.1)) {
let lhs_len = value.len() - name.len();
let lhs = strip_underscores(value[..lhs_len].as_bytes());
let lhs_span = Span::new(span.start, span.start + lhs_len);
let unit_span = Span::new(span.start + lhs_len, span.end);
if lhs.ends_with('$') {
// If `parse_unit_value` has higher precedence over `parse_range`,
// a variable with the name of a unit could otherwise not be used as the end of a range.
return None;
}
let (decimal_part, number_part) = modf(match lhs.parse::<f64>() {
Ok(it) => it,
Err(_) => {
let mk_err = move |name| {
ParseError::LabeledError(
format!("{name} value must be a number"),
"not a number".into(),
lhs_span,
)
};
return Some(Err(Box::new(mk_err)));
}
});
let (num, unit) = match convert {
Some(convert_to) => (
((number_part * convert_to.1 as f64) + (decimal_part * convert_to.1 as f64)) as i64,
convert_to.0,
),
None => (number_part as i64, *unit),
};
trace!("-- found {} {:?}", num, unit);
let value = ValueWithUnit {
expr: Expression::new_unknown(Expr::Int(num), lhs_span, Type::Number),
unit: Spanned {
item: unit,
span: unit_span,
},
};
let expr = Expression::new_unknown(Expr::ValueWithUnit(Box::new(value)), span, ty);
Some(Ok(expr))
} else {
None
}
}
pub const FILESIZE_UNIT_GROUPS: &[UnitGroup] = &[
(Unit::Kilobyte, "KB", Some((Unit::Byte, 1000))),
(Unit::Megabyte, "MB", Some((Unit::Kilobyte, 1000))),
(Unit::Gigabyte, "GB", Some((Unit::Megabyte, 1000))),
(Unit::Terabyte, "TB", Some((Unit::Gigabyte, 1000))),
(Unit::Petabyte, "PB", Some((Unit::Terabyte, 1000))),
(Unit::Exabyte, "EB", Some((Unit::Petabyte, 1000))),
(Unit::Kibibyte, "KIB", Some((Unit::Byte, 1024))),
(Unit::Mebibyte, "MIB", Some((Unit::Kibibyte, 1024))),
(Unit::Gibibyte, "GIB", Some((Unit::Mebibyte, 1024))),
(Unit::Tebibyte, "TIB", Some((Unit::Gibibyte, 1024))),
(Unit::Pebibyte, "PIB", Some((Unit::Tebibyte, 1024))),
(Unit::Exbibyte, "EIB", Some((Unit::Pebibyte, 1024))),
(Unit::Byte, "B", None),
];
pub const DURATION_UNIT_GROUPS: &[UnitGroup] = &[
(Unit::Nanosecond, "ns", None),
Fix duration type to not report months or years (#9632) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> This PR should close #8036, #9028 (in the negative) and #9118. Fix for #9118 is a bit pedantic. As reported, the issue is: ``` > 2023-05-07T04:08:45+12:00 - 2019-05-10T09:59:12+12:00 3yr 12month 2day 18hr 9min 33sec ``` with this PR, you now get: ``` > 2023-05-07T04:08:45+12:00 - 2019-05-10T09:59:12+12:00 208wk 1day 18hr 9min 33sec ``` Which is strictly correct, but could still fairly be called "weird date arithmetic". # Description * [x] Abide by constraint that Value::Duration remains a number of nanoseconds with no additional fields. * [x] `to_string()` only displays weeks .. nanoseconds. Duration doesn't have base date to compute months or years from. * [x] `duration | into record` likewise only has fields for weeks .. nanoseconds. * [x] `string | into duration` now accepts compound form of duration to_string() (e.g '2day 3hr`, not just '2day') * [x] `duration | into string` now works (and produces the same representation as to_string(), which may be compound). # User-Facing Changes ## duration -> string -> duration Now you can "round trip" an arbitrary duration value: convert it to a string that may include multiple time units (a "compound" value), then convert that string back into a duration. This required changes to `string | into duration` and the addition of `duration | into string'. ``` > 2day + 3hr 2day 3hr # the "to_string()" representation (in this case, a compound value) > 2day + 3hr | into string 2day 3hr # string value > 2day + 3hr | into string | into duration 2day 3hr # round-trip duration -> string -> duration ``` Note that `to nuon` and `from nuon` already round-tripped durations, but use a different string representation. ## potentially breaking changes * string rendering of a duration no longer has 'yr' or 'month' phrases. * record from `duration | into record` no longer has 'year' or 'month' fields. The excess duration is all lumped into the `week` field, which is the largest time unit you can convert to without knowing the datetime from which the duration was calculated. Scripts that depended on month or year time units on output will need to be changed. ### Examples ``` > 365day 52wk 1day ## Used to be: ## 1yr > 365day | into record ╭──────┬────╮ │ week │ 52 │ │ day │ 1 │ │ sign │ + │ ╰──────┴────╯ ## used to be: ##╭──────┬───╮ ##│ year │ 1 │ ##│ sign │ + │ ##╰──────┴───╯ > (365day + 4wk + 5day + 6hr + 7min + 8sec + 9ms + 10us + 11ns) 56wk 6day 6hr 7min 8sec 9ms 10µs 11ns ## used to be: ## 1yr 1month 3day 6hr 7min 8sec 9ms 10µs 11ns ## which looks reasonable, but was actually only correct in 75% of the years and 25% of the months in the last 4 years. > (365day + 4wk + 5day + 6hr + 7min + 8sec + 9ms + 10us + 11ns) | into record ╭─────────────┬────╮ │ week │ 56 │ │ day │ 6 │ │ hour │ 6 │ │ minute │ 7 │ │ second │ 8 │ │ millisecond │ 9 │ │ microsecond │ 10 │ │ nanosecond │ 11 │ │ sign │ + │ ╰─────────────┴────╯ ``` Strictly speaking, these changes could break an existing user script. Losing years and months as time units is arguably a regression in behavior. Also, the corrected duration calculation could break an existing script that was calibrated using the old algorithm. # Tests + Formatting ``` > toolkit check pr ``` - :green_circle: `toolkit fmt` - :green_circle: `toolkit clippy` - :green_circle: `toolkit test` - :green_circle: `toolkit test stdlib` # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: Bob Hyman <bobhy@localhost.localdomain>
2023-08-08 13:24:09 +02:00
// todo start adding aliases for duration units here
(Unit::Microsecond, "us", Some((Unit::Nanosecond, 1000))),
(
// µ Micro Sign
Unit::Microsecond,
"\u{00B5}s",
Some((Unit::Nanosecond, 1000)),
),
(
// μ Greek small letter Mu
Unit::Microsecond,
"\u{03BC}s",
Some((Unit::Nanosecond, 1000)),
),
(Unit::Millisecond, "ms", Some((Unit::Microsecond, 1000))),
(Unit::Second, "sec", Some((Unit::Millisecond, 1000))),
(Unit::Minute, "min", Some((Unit::Second, 60))),
(Unit::Hour, "hr", Some((Unit::Minute, 60))),
(Unit::Day, "day", Some((Unit::Minute, 1440))),
(Unit::Week, "wk", Some((Unit::Day, 7))),
];
// Borrowed from libm at https://github.com/rust-lang/libm/blob/master/src/math/modf.rs
fn modf(x: f64) -> (f64, f64) {
let rv2: f64;
let mut u = x.to_bits();
let e = ((u >> 52 & 0x7ff) as i32) - 0x3ff;
/* no fractional part */
if e >= 52 {
rv2 = x;
if e == 0x400 && (u << 12) != 0 {
/* nan */
return (x, rv2);
2021-10-05 04:27:39 +02:00
}
u &= 1 << 63;
return (f64::from_bits(u), rv2);
2021-10-05 04:27:39 +02:00
}
/* no integral part*/
if e < 0 {
u &= 1 << 63;
rv2 = f64::from_bits(u);
return (x, rv2);
}
let mask = ((!0) >> 12) >> e;
if (u & mask) == 0 {
rv2 = x;
u &= 1 << 63;
return (f64::from_bits(u), rv2);
}
u &= !mask;
rv2 = f64::from_bits(u);
(x - rv2, rv2)
}
2021-10-05 04:27:39 +02:00
pub fn parse_glob_pattern(working_set: &mut StateWorkingSet, span: Span) -> Expression {
2021-10-04 21:21:31 +02:00
let bytes = working_set.get_span_contents(span);
do not attempt to glob expand if the file path is wrapped in quotes (#11569) # Description Fixes: #11455 ### For arguments which is annotated with `:path/:directory/:glob` To fix the issue, we need to have a way to know if a path is originally quoted during runtime. So the information needed to be added at several levels: * parse time (from user input to expression) We need to add quoted information into `Expr::Filepath`, `Expr::Directory`, `Expr::GlobPattern` * eval time When convert from `Expr::Filepath`, `Expr::Directory`, `Expr::GlobPattern` to `Value::String` during runtime, we won't auto expanded the path if it's quoted ### For `ls` It's really special, because it accepts a `String` as a pattern, and it generates `glob` expression inside the command itself. So the idea behind the change is introducing a special SyntaxShape to ls: `SyntaxShape::LsGlobPattern`. So we can track if the pattern is originally quoted easier, and we don't auto expand the path either. Then when constructing a glob pattern inside ls, we check if input pattern is quoted, if so: we escape the input pattern, so we can run `ls a[123]b`, because it's already escaped. Finally, to accomplish the checking process, we also need to introduce a new value type called `Value::QuotedString` to differ from `Value::String`, it's used to generate an enum called `NuPath`, which is finally used in `ls` function. `ls` learned from `NuPath` to know if user input is quoted. # User-Facing Changes Actually it contains several changes ### For arguments which is annotated with `:path/:directory/:glob` #### Before ```nushell > def foo [p: path] { echo $p }; print (foo "~/a"); print (foo '~/a') /home/windsoilder/a /home/windsoilder/a > def foo [p: directory] { echo $p }; print (foo "~/a"); print (foo '~/a') /home/windsoilder/a /home/windsoilder/a > def foo [p: glob] { echo $p }; print (foo "~/a"); print (foo '~/a') /home/windsoilder/a /home/windsoilder/a ``` #### After ```nushell > def foo [p: path] { echo $p }; print (foo "~/a"); print (foo '~/a') ~/a ~/a > def foo [p: directory] { echo $p }; print (foo "~/a"); print (foo '~/a') ~/a ~/a > def foo [p: glob] { echo $p }; print (foo "~/a"); print (foo '~/a') ~/a ~/a ``` ### For ls command `touch '[uwu]'` #### Before ``` ❯ ls -D "[uwu]" Error: × No matches found for [uwu] ╭─[entry #6:1:1] 1 │ ls -D "[uwu]" · ───┬─── · ╰── Pattern, file or folder not found ╰──── help: no matches found ``` #### After ``` ❯ ls -D "[uwu]" ╭───┬───────┬──────┬──────┬──────────╮ │ # │ name │ type │ size │ modified │ ├───┼───────┼──────┼──────┼──────────┤ │ 0 │ [uwu] │ file │ 0 B │ now │ ╰───┴───────┴──────┴──────┴──────────╯ ``` # Tests + Formatting Done # After Submitting NaN
2024-01-21 16:22:25 +01:00
let quoted = is_quoted(bytes);
let (token, err) = unescape_unquote_string(bytes, span);
trace!("parsing: glob pattern");
2021-10-04 21:21:31 +02:00
if err.is_none() {
2022-01-01 22:42:50 +01:00
trace!("-- found {}", token);
Expression::new(
working_set,
Expr::GlobPattern(token, quoted),
do not attempt to glob expand if the file path is wrapped in quotes (#11569) # Description Fixes: #11455 ### For arguments which is annotated with `:path/:directory/:glob` To fix the issue, we need to have a way to know if a path is originally quoted during runtime. So the information needed to be added at several levels: * parse time (from user input to expression) We need to add quoted information into `Expr::Filepath`, `Expr::Directory`, `Expr::GlobPattern` * eval time When convert from `Expr::Filepath`, `Expr::Directory`, `Expr::GlobPattern` to `Value::String` during runtime, we won't auto expanded the path if it's quoted ### For `ls` It's really special, because it accepts a `String` as a pattern, and it generates `glob` expression inside the command itself. So the idea behind the change is introducing a special SyntaxShape to ls: `SyntaxShape::LsGlobPattern`. So we can track if the pattern is originally quoted easier, and we don't auto expand the path either. Then when constructing a glob pattern inside ls, we check if input pattern is quoted, if so: we escape the input pattern, so we can run `ls a[123]b`, because it's already escaped. Finally, to accomplish the checking process, we also need to introduce a new value type called `Value::QuotedString` to differ from `Value::String`, it's used to generate an enum called `NuPath`, which is finally used in `ls` function. `ls` learned from `NuPath` to know if user input is quoted. # User-Facing Changes Actually it contains several changes ### For arguments which is annotated with `:path/:directory/:glob` #### Before ```nushell > def foo [p: path] { echo $p }; print (foo "~/a"); print (foo '~/a') /home/windsoilder/a /home/windsoilder/a > def foo [p: directory] { echo $p }; print (foo "~/a"); print (foo '~/a') /home/windsoilder/a /home/windsoilder/a > def foo [p: glob] { echo $p }; print (foo "~/a"); print (foo '~/a') /home/windsoilder/a /home/windsoilder/a ``` #### After ```nushell > def foo [p: path] { echo $p }; print (foo "~/a"); print (foo '~/a') ~/a ~/a > def foo [p: directory] { echo $p }; print (foo "~/a"); print (foo '~/a') ~/a ~/a > def foo [p: glob] { echo $p }; print (foo "~/a"); print (foo '~/a') ~/a ~/a ``` ### For ls command `touch '[uwu]'` #### Before ``` ❯ ls -D "[uwu]" Error: × No matches found for [uwu] ╭─[entry #6:1:1] 1 │ ls -D "[uwu]" · ───┬─── · ╰── Pattern, file or folder not found ╰──── help: no matches found ``` #### After ``` ❯ ls -D "[uwu]" ╭───┬───────┬──────┬──────┬──────────╮ │ # │ name │ type │ size │ modified │ ├───┼───────┼──────┼──────┼──────────┤ │ 0 │ [uwu] │ file │ 0 B │ now │ ╰───┴───────┴──────┴──────┴──────────╯ ``` # Tests + Formatting Done # After Submitting NaN
2024-01-21 16:22:25 +01:00
span,
Type::Glob,
)
do not attempt to glob expand if the file path is wrapped in quotes (#11569) # Description Fixes: #11455 ### For arguments which is annotated with `:path/:directory/:glob` To fix the issue, we need to have a way to know if a path is originally quoted during runtime. So the information needed to be added at several levels: * parse time (from user input to expression) We need to add quoted information into `Expr::Filepath`, `Expr::Directory`, `Expr::GlobPattern` * eval time When convert from `Expr::Filepath`, `Expr::Directory`, `Expr::GlobPattern` to `Value::String` during runtime, we won't auto expanded the path if it's quoted ### For `ls` It's really special, because it accepts a `String` as a pattern, and it generates `glob` expression inside the command itself. So the idea behind the change is introducing a special SyntaxShape to ls: `SyntaxShape::LsGlobPattern`. So we can track if the pattern is originally quoted easier, and we don't auto expand the path either. Then when constructing a glob pattern inside ls, we check if input pattern is quoted, if so: we escape the input pattern, so we can run `ls a[123]b`, because it's already escaped. Finally, to accomplish the checking process, we also need to introduce a new value type called `Value::QuotedString` to differ from `Value::String`, it's used to generate an enum called `NuPath`, which is finally used in `ls` function. `ls` learned from `NuPath` to know if user input is quoted. # User-Facing Changes Actually it contains several changes ### For arguments which is annotated with `:path/:directory/:glob` #### Before ```nushell > def foo [p: path] { echo $p }; print (foo "~/a"); print (foo '~/a') /home/windsoilder/a /home/windsoilder/a > def foo [p: directory] { echo $p }; print (foo "~/a"); print (foo '~/a') /home/windsoilder/a /home/windsoilder/a > def foo [p: glob] { echo $p }; print (foo "~/a"); print (foo '~/a') /home/windsoilder/a /home/windsoilder/a ``` #### After ```nushell > def foo [p: path] { echo $p }; print (foo "~/a"); print (foo '~/a') ~/a ~/a > def foo [p: directory] { echo $p }; print (foo "~/a"); print (foo '~/a') ~/a ~/a > def foo [p: glob] { echo $p }; print (foo "~/a"); print (foo '~/a') ~/a ~/a ``` ### For ls command `touch '[uwu]'` #### Before ``` ❯ ls -D "[uwu]" Error: × No matches found for [uwu] ╭─[entry #6:1:1] 1 │ ls -D "[uwu]" · ───┬─── · ╰── Pattern, file or folder not found ╰──── help: no matches found ``` #### After ``` ❯ ls -D "[uwu]" ╭───┬───────┬──────┬──────┬──────────╮ │ # │ name │ type │ size │ modified │ ├───┼───────┼──────┼──────┼──────────┤ │ 0 │ [uwu] │ file │ 0 B │ now │ ╰───┴───────┴──────┴──────┴──────────╯ ``` # Tests + Formatting Done # After Submitting NaN
2024-01-21 16:22:25 +01:00
} else {
working_set.error(ParseError::Expected("glob pattern string", span));
garbage(working_set, span)
do not attempt to glob expand if the file path is wrapped in quotes (#11569) # Description Fixes: #11455 ### For arguments which is annotated with `:path/:directory/:glob` To fix the issue, we need to have a way to know if a path is originally quoted during runtime. So the information needed to be added at several levels: * parse time (from user input to expression) We need to add quoted information into `Expr::Filepath`, `Expr::Directory`, `Expr::GlobPattern` * eval time When convert from `Expr::Filepath`, `Expr::Directory`, `Expr::GlobPattern` to `Value::String` during runtime, we won't auto expanded the path if it's quoted ### For `ls` It's really special, because it accepts a `String` as a pattern, and it generates `glob` expression inside the command itself. So the idea behind the change is introducing a special SyntaxShape to ls: `SyntaxShape::LsGlobPattern`. So we can track if the pattern is originally quoted easier, and we don't auto expand the path either. Then when constructing a glob pattern inside ls, we check if input pattern is quoted, if so: we escape the input pattern, so we can run `ls a[123]b`, because it's already escaped. Finally, to accomplish the checking process, we also need to introduce a new value type called `Value::QuotedString` to differ from `Value::String`, it's used to generate an enum called `NuPath`, which is finally used in `ls` function. `ls` learned from `NuPath` to know if user input is quoted. # User-Facing Changes Actually it contains several changes ### For arguments which is annotated with `:path/:directory/:glob` #### Before ```nushell > def foo [p: path] { echo $p }; print (foo "~/a"); print (foo '~/a') /home/windsoilder/a /home/windsoilder/a > def foo [p: directory] { echo $p }; print (foo "~/a"); print (foo '~/a') /home/windsoilder/a /home/windsoilder/a > def foo [p: glob] { echo $p }; print (foo "~/a"); print (foo '~/a') /home/windsoilder/a /home/windsoilder/a ``` #### After ```nushell > def foo [p: path] { echo $p }; print (foo "~/a"); print (foo '~/a') ~/a ~/a > def foo [p: directory] { echo $p }; print (foo "~/a"); print (foo '~/a') ~/a ~/a > def foo [p: glob] { echo $p }; print (foo "~/a"); print (foo '~/a') ~/a ~/a ``` ### For ls command `touch '[uwu]'` #### Before ``` ❯ ls -D "[uwu]" Error: × No matches found for [uwu] ╭─[entry #6:1:1] 1 │ ls -D "[uwu]" · ───┬─── · ╰── Pattern, file or folder not found ╰──── help: no matches found ``` #### After ``` ❯ ls -D "[uwu]" ╭───┬───────┬──────┬──────┬──────────╮ │ # │ name │ type │ size │ modified │ ├───┼───────┼──────┼──────┼──────────┤ │ 0 │ [uwu] │ file │ 0 B │ now │ ╰───┴───────┴──────┴──────┴──────────╯ ``` # Tests + Formatting Done # After Submitting NaN
2024-01-21 16:22:25 +01:00
}
}
pub fn unescape_string(bytes: &[u8], span: Span) -> (Vec<u8>, Option<ParseError>) {
let mut output = Vec::new();
let mut error = None;
let mut idx = 0;
if !bytes.contains(&b'\\') {
return (bytes.to_vec(), None);
}
'us_loop: while idx < bytes.len() {
if bytes[idx] == b'\\' {
// We're in an escape
idx += 1;
match bytes.get(idx) {
Some(b'"') => {
output.push(b'"');
idx += 1;
}
Some(b'\'') => {
output.push(b'\'');
idx += 1;
}
Some(b'\\') => {
output.push(b'\\');
idx += 1;
}
Some(b'/') => {
output.push(b'/');
idx += 1;
}
Some(b'(') => {
output.push(b'(');
idx += 1;
}
Some(b')') => {
output.push(b')');
idx += 1;
}
Some(b'{') => {
output.push(b'{');
idx += 1;
}
Some(b'}') => {
output.push(b'}');
idx += 1;
}
Some(b'$') => {
output.push(b'$');
idx += 1;
}
Some(b'^') => {
output.push(b'^');
idx += 1;
}
Some(b'#') => {
output.push(b'#');
idx += 1;
}
Some(b'|') => {
output.push(b'|');
idx += 1;
}
Some(b'~') => {
output.push(b'~');
idx += 1;
}
Some(b'a') => {
output.push(0x7);
idx += 1;
}
Some(b'b') => {
output.push(0x8);
idx += 1;
}
Some(b'e') => {
output.push(0x1b);
idx += 1;
}
Some(b'f') => {
output.push(0xc);
idx += 1;
}
Some(b'n') => {
output.push(b'\n');
idx += 1;
}
Some(b'r') => {
output.push(b'\r');
idx += 1;
}
Some(b't') => {
output.push(b'\t');
idx += 1;
}
Some(b'u') => {
let mut digits = String::with_capacity(10);
let mut cur_idx = idx + 1; // index of first beyond current end of token
if let Some(b'{') = bytes.get(idx + 1) {
cur_idx = idx + 2;
loop {
match bytes.get(cur_idx) {
Some(b'}') => {
cur_idx += 1;
break;
}
Some(c) => {
digits.push(*c as char);
cur_idx += 1;
}
_ => {
error = error.or(Some(ParseError::InvalidLiteral(
Syntax errors for string and int (#7952) # Description Added a few syntax errors in ints and strings, changed parser to stop and show that error rather than continue trying to parse those tokens as some other shape. However, I don't see how to push this direction much further, and most of the classic confusing errors can't be changed. Flagged as WIP for the moment, but passes all checks and works better than current release: 1. I have yet to figure out how to make these errors refer back to the book, as I see some other errors do. 2. How to give syntax error when malformed int is first token in line? Currently parsed as external command, user gets confusing error message. 3. Would like to be more strict with *decimal* int literals (lacking, e.g, `0x' prefix). Need to tinker more with the order of parse shape calls, currently, float is tried after int, so '1.4' has to be passed. _(Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience.)_ ```bash 〉"\z" Error: ╭─[entry #3:1:1] 1 │ "\z" · ─┬─ · ╰── Syntax error in string, unrecognized character after escape '\'. ╰──── ``` Canonic presentation of a syntax error. ```bash 〉" \u{01ffbogus}" Error: × Invalid syntax ╭─[entry #2:1:1] 1 │ " \u{01ffbogus}" · ───────┬────── · ╰── Syntax error in string, expecting 1 to 6 hex digits in unicode escape '\u{X...}', max value 10FFFF. ╰──── ``` Malformed unicode escape in string, flagged as error. String parse can be opinionated, it's the last shape tried. ```bash 〉0x22bogus Error: nu::shell::external_command (link) × External command failed ╭─[entry #4:1:1] 1 │ 0x22bogus · ────┬──── · ╰── executable was not found ╰──── help: No such file or directory (os error 2) ``` A *correct* number in first token would be evaluated, but an *incorrect* one is treated as external command? Confusing to users. ```bash 〉0 + 0x22bogus Error: × Invalid syntax ╭─[entry #5:1:1] 1 │ 0 + 0x22bogus · ────┬──── · ╰── Syntax error in int, invalid digits in radix 16 int. ╰──── ``` Can give syntax error if token is unambiguously int literal. e.g has 0b or 0x prefix, could not be a float. ```bash 〉0 + 098bogus Error: nu::parser::unsupported_operation (link) × Types mismatched for operation. ╭─[entry #6:1:1] 1 │ 0 + 098bogus · ┬ ┬ ────┬─── · │ │ ╰── string · │ ╰── doesn't support these values. · ╰── int ╰──── help: Change int or string to be the right types and try again. ``` But *decimal* literal (no prefix) can't be too strict. Parser is going to try float later. So '1.4' must be passed. # User-Facing Changes First and foremost, more specific error messages for typos in string and int literals. Probably improves interactive user experience. But a script that was causing and then checking for specific error might notice a different error message. _(List of all changes that impact the user experience here. This helps us keep track of breaking changes.)_ # Tests + Formatting Added (positive and negative unit tests in `cargo test -p nu-parser`. Didn't add integration tests. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [x] `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [x] `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --------- Co-authored-by: Stefan Holderbach <sholderbach@users.noreply.github.com>
2023-02-13 17:09:50 +01:00
"missing '}' for unicode escape '\\u{X...}'".into(),
"string".into(),
Span::new(span.start + idx, span.end),
)));
break 'us_loop;
}
}
}
}
if (1..=6).contains(&digits.len()) {
let int = u32::from_str_radix(&digits, 16);
if let Ok(int) = int {
if int <= 0x10ffff {
let result = char::from_u32(int);
if let Some(result) = result {
let mut buffer = vec![0; 4];
let result = result.encode_utf8(&mut buffer);
for elem in result.bytes() {
output.push(elem);
}
idx = cur_idx;
continue 'us_loop;
}
}
}
}
// fall through -- escape not accepted above, must be error.
error = error.or(Some(ParseError::InvalidLiteral(
2023-03-10 22:20:31 +01:00
"invalid unicode escape '\\u{X...}', must be 1-6 hex digits, max value 10FFFF".into(),
"string".into(),
Span::new(span.start + idx, span.end),
)));
break 'us_loop;
}
_ => {
error = error.or(Some(ParseError::InvalidLiteral(
Syntax errors for string and int (#7952) # Description Added a few syntax errors in ints and strings, changed parser to stop and show that error rather than continue trying to parse those tokens as some other shape. However, I don't see how to push this direction much further, and most of the classic confusing errors can't be changed. Flagged as WIP for the moment, but passes all checks and works better than current release: 1. I have yet to figure out how to make these errors refer back to the book, as I see some other errors do. 2. How to give syntax error when malformed int is first token in line? Currently parsed as external command, user gets confusing error message. 3. Would like to be more strict with *decimal* int literals (lacking, e.g, `0x' prefix). Need to tinker more with the order of parse shape calls, currently, float is tried after int, so '1.4' has to be passed. _(Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience.)_ ```bash 〉"\z" Error: ╭─[entry #3:1:1] 1 │ "\z" · ─┬─ · ╰── Syntax error in string, unrecognized character after escape '\'. ╰──── ``` Canonic presentation of a syntax error. ```bash 〉" \u{01ffbogus}" Error: × Invalid syntax ╭─[entry #2:1:1] 1 │ " \u{01ffbogus}" · ───────┬────── · ╰── Syntax error in string, expecting 1 to 6 hex digits in unicode escape '\u{X...}', max value 10FFFF. ╰──── ``` Malformed unicode escape in string, flagged as error. String parse can be opinionated, it's the last shape tried. ```bash 〉0x22bogus Error: nu::shell::external_command (link) × External command failed ╭─[entry #4:1:1] 1 │ 0x22bogus · ────┬──── · ╰── executable was not found ╰──── help: No such file or directory (os error 2) ``` A *correct* number in first token would be evaluated, but an *incorrect* one is treated as external command? Confusing to users. ```bash 〉0 + 0x22bogus Error: × Invalid syntax ╭─[entry #5:1:1] 1 │ 0 + 0x22bogus · ────┬──── · ╰── Syntax error in int, invalid digits in radix 16 int. ╰──── ``` Can give syntax error if token is unambiguously int literal. e.g has 0b or 0x prefix, could not be a float. ```bash 〉0 + 098bogus Error: nu::parser::unsupported_operation (link) × Types mismatched for operation. ╭─[entry #6:1:1] 1 │ 0 + 098bogus · ┬ ┬ ────┬─── · │ │ ╰── string · │ ╰── doesn't support these values. · ╰── int ╰──── help: Change int or string to be the right types and try again. ``` But *decimal* literal (no prefix) can't be too strict. Parser is going to try float later. So '1.4' must be passed. # User-Facing Changes First and foremost, more specific error messages for typos in string and int literals. Probably improves interactive user experience. But a script that was causing and then checking for specific error might notice a different error message. _(List of all changes that impact the user experience here. This helps us keep track of breaking changes.)_ # Tests + Formatting Added (positive and negative unit tests in `cargo test -p nu-parser`. Didn't add integration tests. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [x] `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [x] `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --------- Co-authored-by: Stefan Holderbach <sholderbach@users.noreply.github.com>
2023-02-13 17:09:50 +01:00
"unrecognized escape after '\\'".into(),
"string".into(),
Protocol: debug_assert!() Span to reflect a valid slice (#6806) Also enforce this by #[non_exhaustive] span such that going forward we cannot, in debug builds (1), construct invalid spans. The motivation for this stems from #6431 where I've seen crashes due to invalid slice indexing. My hope is this will mitigate such senarios 1. https://github.com/nushell/nushell/pull/6431#issuecomment-1278147241 # Description (description of your pull request here) # Tests Make sure you've done the following: - [ ] Add tests that cover your changes, either in the command examples, the crate/tests folder, or in the /tests folder. - [ ] Try to think about corner cases and various ways how your changes could break. Cover them with tests. - [ ] If adding tests is not possible, please document in the PR body a minimal example with steps on how to reproduce so one can verify your change works. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [ ] `cargo clippy --workspace --features=extra -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [ ] `cargo test --workspace --features=extra` to check that all the tests pass # Documentation - [ ] If your PR touches a user-facing nushell feature then make sure that there is an entry in the documentation (https://github.com/nushell/nushell.github.io) for the feature, and update it if necessary.
2022-12-03 10:44:12 +01:00
Span::new(span.start + idx, span.end),
)));
break 'us_loop;
}
}
} else {
output.push(bytes[idx]);
idx += 1;
}
}
(output, error)
}
pub fn unescape_unquote_string(bytes: &[u8], span: Span) -> (String, Option<ParseError>) {
if bytes.starts_with(b"\"") {
// Needs unescaping
let bytes = trim_quotes(bytes);
let (bytes, err) = unescape_string(bytes, span);
if let Ok(token) = String::from_utf8(bytes) {
(token, err)
} else {
(String::new(), Some(ParseError::Expected("string", span)))
}
} else {
let bytes = trim_quotes(bytes);
if let Ok(token) = String::from_utf8(bytes.into()) {
(token, None)
} else {
(String::new(), Some(ParseError::Expected("string", span)))
}
}
}
/// XXX: This is here temporarily as a patch, but we should replace this with properly representing
/// the quoted state of a string in the AST
fn unescape_string_preserving_quotes(bytes: &[u8], span: Span) -> (String, Option<ParseError>) {
let (bytes, err) = if bytes.starts_with(b"\"") {
let (bytes, err) = unescape_string(bytes, span);
(Cow::Owned(bytes), err)
} else {
(Cow::Borrowed(bytes), None)
};
// The original code for args used lossy conversion here, even though that's not what we
// typically use for strings. Revisit whether that's actually desirable later, but don't
// want to introduce a breaking change for this patch.
let token = String::from_utf8_lossy(&bytes).into_owned();
(token, err)
}
pub fn parse_string(working_set: &mut StateWorkingSet, span: Span) -> Expression {
2022-01-01 22:42:50 +01:00
trace!("parsing: string");
2021-09-02 10:25:22 +02:00
let bytes = working_set.get_span_contents(span);
2022-01-19 15:58:12 +01:00
if bytes.is_empty() {
working_set.error(ParseError::Expected("String", span));
return Expression::garbage(working_set, span);
}
// Check for bare word interpolation
if bytes[0] != b'\'' && bytes[0] != b'"' && bytes[0] != b'`' && bytes.contains(&b'(') {
return parse_string_interpolation(working_set, span);
}
let (s, err) = unescape_unquote_string(bytes, span);
if let Some(err) = err {
working_set.error(err);
}
2021-07-16 22:26:40 +02:00
Expression::new(working_set, Expr::String(s), span, Type::String)
2021-09-02 10:25:22 +02:00
}
2021-07-02 09:15:30 +02:00
do not attempt to glob expand if the file path is wrapped in quotes (#11569) # Description Fixes: #11455 ### For arguments which is annotated with `:path/:directory/:glob` To fix the issue, we need to have a way to know if a path is originally quoted during runtime. So the information needed to be added at several levels: * parse time (from user input to expression) We need to add quoted information into `Expr::Filepath`, `Expr::Directory`, `Expr::GlobPattern` * eval time When convert from `Expr::Filepath`, `Expr::Directory`, `Expr::GlobPattern` to `Value::String` during runtime, we won't auto expanded the path if it's quoted ### For `ls` It's really special, because it accepts a `String` as a pattern, and it generates `glob` expression inside the command itself. So the idea behind the change is introducing a special SyntaxShape to ls: `SyntaxShape::LsGlobPattern`. So we can track if the pattern is originally quoted easier, and we don't auto expand the path either. Then when constructing a glob pattern inside ls, we check if input pattern is quoted, if so: we escape the input pattern, so we can run `ls a[123]b`, because it's already escaped. Finally, to accomplish the checking process, we also need to introduce a new value type called `Value::QuotedString` to differ from `Value::String`, it's used to generate an enum called `NuPath`, which is finally used in `ls` function. `ls` learned from `NuPath` to know if user input is quoted. # User-Facing Changes Actually it contains several changes ### For arguments which is annotated with `:path/:directory/:glob` #### Before ```nushell > def foo [p: path] { echo $p }; print (foo "~/a"); print (foo '~/a') /home/windsoilder/a /home/windsoilder/a > def foo [p: directory] { echo $p }; print (foo "~/a"); print (foo '~/a') /home/windsoilder/a /home/windsoilder/a > def foo [p: glob] { echo $p }; print (foo "~/a"); print (foo '~/a') /home/windsoilder/a /home/windsoilder/a ``` #### After ```nushell > def foo [p: path] { echo $p }; print (foo "~/a"); print (foo '~/a') ~/a ~/a > def foo [p: directory] { echo $p }; print (foo "~/a"); print (foo '~/a') ~/a ~/a > def foo [p: glob] { echo $p }; print (foo "~/a"); print (foo '~/a') ~/a ~/a ``` ### For ls command `touch '[uwu]'` #### Before ``` ❯ ls -D "[uwu]" Error: × No matches found for [uwu] ╭─[entry #6:1:1] 1 │ ls -D "[uwu]" · ───┬─── · ╰── Pattern, file or folder not found ╰──── help: no matches found ``` #### After ``` ❯ ls -D "[uwu]" ╭───┬───────┬──────┬──────┬──────────╮ │ # │ name │ type │ size │ modified │ ├───┼───────┼──────┼──────┼──────────┤ │ 0 │ [uwu] │ file │ 0 B │ now │ ╰───┴───────┴──────┴──────┴──────────╯ ``` # Tests + Formatting Done # After Submitting NaN
2024-01-21 16:22:25 +01:00
fn is_quoted(bytes: &[u8]) -> bool {
(bytes.starts_with(b"\"") && bytes.ends_with(b"\"") && bytes.len() > 1)
|| (bytes.starts_with(b"\'") && bytes.ends_with(b"\'") && bytes.len() > 1)
}
pub fn parse_string_strict(working_set: &mut StateWorkingSet, span: Span) -> Expression {
2022-01-01 22:42:50 +01:00
trace!("parsing: string, with required delimiters");
2021-11-04 03:32:35 +01:00
let bytes = working_set.get_span_contents(span);
2022-01-19 15:58:12 +01:00
// Check for unbalanced quotes:
{
let bytes = if bytes.starts_with(b"$") {
&bytes[1..]
} else {
bytes
};
if bytes.starts_with(b"\"") && (bytes.len() == 1 || !bytes.ends_with(b"\"")) {
working_set.error(ParseError::Unclosed("\"".into(), span));
return garbage(working_set, span);
}
if bytes.starts_with(b"\'") && (bytes.len() == 1 || !bytes.ends_with(b"\'")) {
working_set.error(ParseError::Unclosed("\'".into(), span));
return garbage(working_set, span);
}
2022-01-19 15:58:12 +01:00
}
2021-11-04 03:32:35 +01:00
let (bytes, quoted) = if (bytes.starts_with(b"\"") && bytes.ends_with(b"\"") && bytes.len() > 1)
|| (bytes.starts_with(b"\'") && bytes.ends_with(b"\'") && bytes.len() > 1)
{
(&bytes[1..(bytes.len() - 1)], true)
} else if (bytes.starts_with(b"$\"") && bytes.ends_with(b"\"") && bytes.len() > 2)
|| (bytes.starts_with(b"$\'") && bytes.ends_with(b"\'") && bytes.len() > 2)
{
(&bytes[2..(bytes.len() - 1)], true)
2021-11-04 03:32:35 +01:00
} else {
(bytes, false)
};
if let Ok(token) = String::from_utf8(bytes.into()) {
2022-01-01 22:42:50 +01:00
trace!("-- found {}", token);
2021-11-04 03:32:35 +01:00
if quoted {
Expression::new(working_set, Expr::String(token), span, Type::String)
2021-11-04 03:32:35 +01:00
} else if token.contains(' ') {
working_set.error(ParseError::Expected("string", span));
garbage(working_set, span)
2021-11-04 03:32:35 +01:00
} else {
Expression::new(working_set, Expr::String(token), span, Type::String)
2021-11-04 03:32:35 +01:00
}
} else {
working_set.error(ParseError::Expected("string", span));
garbage(working_set, span)
2021-11-04 03:32:35 +01:00
}
}
pub fn parse_import_pattern(working_set: &mut StateWorkingSet, spans: &[Span]) -> Expression {
Apply nightly clippy fixes (#11083) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> # Description <!-- Thank you for improving Nushell. Please, check our [contributing guide](../CONTRIBUTING.md) and talk to the core team before making major changes. Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience. --> Clippy fixes for rust 1.76.0-nightly # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> N/A # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass (on Windows make sure to [enable developer mode](https://learn.microsoft.com/en-us/windows/apps/get-started/developer-mode-features-and-debugging)) - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-11-17 16:15:55 +01:00
let Some(head_span) = spans.first() else {
working_set.error(ParseError::WrongImportPattern(
"needs at least one component of import pattern".to_string(),
Span::concat(spans),
));
return garbage(working_set, Span::concat(spans));
};
2021-09-26 20:39:19 +02:00
let head_expr = parse_value(working_set, *head_span, &SyntaxShape::Any);
let (maybe_module_id, head_name) = match eval_constant(working_set, &head_expr) {
Name the `Value` conversion functions more clearly (#11851) # Description This PR renames the conversion functions on `Value` to be more consistent. It follows the Rust [API guidelines](https://rust-lang.github.io/api-guidelines/naming.html#ad-hoc-conversions-follow-as_-to_-into_-conventions-c-conv) for ad-hoc conversions. The conversion functions on `Value` now come in a few forms: - `coerce_{type}` takes a `&Value` and attempts to convert the value to `type` (e.g., `i64` are converted to `f64`). This is the old behavior of some of the `as_{type}` functions -- these functions have simply been renamed to better reflect what they do. - The new `as_{type}` functions take a `&Value` and returns an `Ok` result only if the value is of `type` (no conversion is attempted). The returned value will be borrowed if `type` is non-`Copy`, otherwise an owned value is returned. - `into_{type}` exists for non-`Copy` types, but otherwise does not attempt conversion just like `as_type`. It takes an owned `Value` and always returns an owned result. - `coerce_into_{type}` has the same relationship with `coerce_{type}` as `into_{type}` does with `as_{type}`. - `to_{kind}_string`: conversion to different string formats (debug, abbreviated, etc.). Only two of the old string conversion functions were removed, the rest have been renamed only. - `to_{type}`: other conversion functions. Currently, only `to_path` exists. (And `to_string` through `Display`.) This table summaries the above: | Form | Cost | Input Ownership | Output Ownership | Converts `Value` case/`type` | | ---------------------------- | ----- | --------------- | ---------------- | -------- | | `as_{type}` | Cheap | Borrowed | Borrowed/Owned | No | | `into_{type}` | Cheap | Owned | Owned | No | | `coerce_{type}` | Cheap | Borrowed | Borrowed/Owned | Yes | | `coerce_into_{type}` | Cheap | Owned | Owned | Yes | | `to_{kind}_string` | Expensive | Borrowed | Owned | Yes | | `to_{type}` | Expensive | Borrowed | Owned | Yes | # User-Facing Changes Breaking API change for `Value` in `nu-protocol` which is exposed as part of the plugin API.
2024-02-17 19:14:16 +01:00
Ok(val) => match val.coerce_into_string() {
Ok(s) => (working_set.find_module(s.as_bytes()), s.into_bytes()),
Err(err) => {
working_set.error(err.wrap(working_set, Span::concat(spans)));
return garbage(working_set, Span::concat(spans));
}
},
Err(err) => {
working_set.error(err.wrap(working_set, Span::concat(spans)));
return garbage(working_set, Span::concat(spans));
}
};
let mut import_pattern = ImportPattern {
head: ImportPatternHead {
name: head_name,
id: maybe_module_id,
span: *head_span,
},
members: vec![],
hidden: HashSet::new(),
Recursively export constants from modules (#10049) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> # Description <!-- Thank you for improving Nushell. Please, check our [contributing guide](../CONTRIBUTING.md) and talk to the core team before making major changes. Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience. --> https://github.com/nushell/nushell/pull/9773 introduced constants to modules and allowed to export them, but only within one level. This PR: * allows recursive exporting of constants from all submodules * fixes submodule imports in a list import pattern * makes sure exported constants are actual constants Should unblock https://github.com/nushell/nushell/pull/9678 ### Example: ```nushell module spam { export module eggs { export module bacon { export const viking = 'eats' } } } use spam print $spam.eggs.bacon.viking # prints 'eats' use spam [eggs] print $eggs.bacon.viking # prints 'eats' use spam eggs bacon viking print $viking # prints 'eats' ``` ### Limitation 1: Considering the above `spam` module, attempting to get `eggs bacon` from `spam` module doesn't work directly: ```nushell use spam [ eggs bacon ] # attempts to load `eggs`, then `bacon` use spam [ "eggs bacon" ] # obviously wrong name for a constant, but doesn't work also for commands ``` Workaround (for example): ```nushell use spam eggs use eggs [ bacon ] print $bacon.viking # prints 'eats' ``` I'm thinking I'll just leave it in, as you can easily work around this. It is also a limitation of the import pattern in general, not just constants. ### Limitation 2: `overlay use` successfully imports the constants, but `overlay hide` does not hide them, even though it seems to hide normal variables successfully. This needs more investigation. # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> Allows recursive constant exports from submodules. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-08-20 14:51:35 +02:00
constants: vec![],
};
2021-09-27 02:23:22 +02:00
if spans.len() > 1 {
let mut leaf_member_span = None;
2021-09-27 02:23:22 +02:00
for tail_span in spans[1..].iter() {
if let Some(prev_span) = leaf_member_span {
let what = if working_set.get_span_contents(prev_span) == b"*" {
"glob"
} else {
"list"
};
working_set.error(ParseError::WrongImportPattern(
format!(
"{} member can be only at the end of an import pattern",
what
),
prev_span,
));
return Expression::new(
working_set,
Expr::ImportPattern(Box::new(import_pattern)),
prev_span,
Type::List(Box::new(Type::String)),
);
}
let tail = working_set.get_span_contents(*tail_span);
if tail == b"*" {
import_pattern
.members
.push(ImportPatternMember::Glob { span: *tail_span });
leaf_member_span = Some(*tail_span);
} else if tail.starts_with(b"[") {
let result = parse_list_expression(working_set, *tail_span, &SyntaxShape::String);
let mut output = vec![];
if let Expression {
2021-09-27 02:23:22 +02:00
expr: Expr::List(list),
..
} = result
{
for item in list {
match item {
ListItem::Item(expr) => {
let contents = working_set.get_span_contents(expr.span);
output.push((trim_quotes(contents).to_vec(), expr.span));
}
ListItem::Spread(_, spread) => {
working_set.error(ParseError::WrongImportPattern(
"cannot spread in an import pattern".into(),
spread.span,
))
}
}
2021-09-27 02:23:22 +02:00
}
import_pattern
.members
.push(ImportPatternMember::List { names: output });
} else {
working_set.error(ParseError::ExportNotFound(result.span));
return Expression::new(
working_set,
Expr::ImportPattern(Box::new(import_pattern)),
Span::concat(spans),
Type::List(Box::new(Type::String)),
);
2021-09-27 02:23:22 +02:00
}
leaf_member_span = Some(*tail_span);
} else {
let tail = trim_quotes(tail);
import_pattern.members.push(ImportPatternMember::Name {
name: tail.to_vec(),
span: *tail_span,
});
2021-09-27 02:23:22 +02:00
}
2021-09-26 20:39:19 +02:00
}
}
Expression::new(
working_set,
Expr::ImportPattern(Box::new(import_pattern)),
Span::concat(&spans[1..]),
Type::List(Box::new(Type::String)),
)
2021-09-26 20:39:19 +02:00
}
Fix: missing parse error when extra tokens are given to let bindings (#12238) Manual checks are added to `parse_let`, `parse_mut`, and `parse_const`. `parse_var_with_opt_type` is also fixed to update `spans_idx` correctly. Fixes #12125. It's technically a fix, but I'd rather not merge this directly. I'm making this PR to bring into attention the code quality of the parser code. For example: * Inconsistent usage of `spans_idx`. What is its purpose, and which parsing functions need it? I suspect it's possible to remove the usage of `spans_idx` entirely. * Lacking documentation for top-level functions. What does `mutable` mean for `parse_var_with_opt_type()`? * Inconsistent error reporting. Usage of both `working_set.error()` and `working_set.parse_errors.push()`. Using `ParseError::Expected` for an invalid variable name when there's `ParseError::VariableNotValid` (from `parser.rs:5237`). Checking variable names manually when there's `is_variable()` (from `parser.rs:2905`). * `span()` is a terrible name for a function that flattens a bunch of spans into one (from `nu-protocal/src/span.rs:92`). The top-level comment (`Used when you have a slice of spans of at least size 1`) doesn't help either. I've only looked at a small portion of the parser code; I expect there are a lot more. These issues made it much harder to fix a simple bug like #12125. I believe we should invest some effort to cleanup the parser code, which will ease maintainance in the future. I'll willing to help if there is interest.
2024-03-21 16:37:52 +01:00
/// Parse `spans[spans_idx..]` into a variable, with optional type annotation.
/// If the name of the variable ends with a colon (no space in-between allowed), then a type annotation
/// can appear after the variable, in which case the colon is stripped from the name of the variable.
/// `spans_idx` is updated to point to the last span that has been parsed.
2021-09-02 10:25:22 +02:00
pub fn parse_var_with_opt_type(
working_set: &mut StateWorkingSet,
spans: &[Span],
spans_idx: &mut usize,
mutable: bool,
Improve type hovers (#9515) # Description This PR does a few things to help improve type hovers and, in the process, fixes a few outstanding issues in the type system. Here's a list of the changes: * `for` now will try to infer the type of the iteration variable based on the expression it's given. This fixes things like `for x in [1, 2, 3] { }` where `x` now properly gets the int type. * Removed old input/output type fields from the signature, focuses on the vec of signatures. Updated a bunch of dataframe commands that hadn't moved over. This helps tie things together a bit better * Fixed inference of types from subexpressions to use the last expression in the block * Fixed handling of explicit types in `let` and `mut` calls, so we now respect that as the authoritative type I also tried to add `def` input/output type inference, but unfortunately we only know the predecl types universally, which means we won't have enough information to properly know what the types of the custom commands are. # User-Facing Changes Script typechecking will get tighter in some cases Hovers should be more accurate in some cases that previously resorted to any. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: Darren Schroeder <343840+fdncred@users.noreply.github.com>
2023-06-28 19:19:48 +02:00
) -> (Expression, Option<Type>) {
2021-09-02 10:25:22 +02:00
let bytes = working_set.get_span_contents(spans[*spans_idx]).to_vec();
2021-07-16 08:24:46 +02:00
if bytes.contains(&b' ')
|| bytes.contains(&b'"')
|| bytes.contains(&b'\'')
|| bytes.contains(&b'`')
{
working_set.error(ParseError::VariableNotValid(spans[*spans_idx]));
return (garbage(working_set, spans[*spans_idx]), None);
2021-10-12 07:08:55 +02:00
}
2021-09-02 10:25:22 +02:00
if bytes.ends_with(b":") {
// We end with colon, so the next span should be the type
if *spans_idx + 1 < spans.len() {
let span_beginning = *spans_idx;
2021-09-02 10:25:22 +02:00
*spans_idx += 1;
// signature like record<a: int b: int> is broken into multiple spans due to
// whitespaces. Collect the rest into one span and work on it
let full_span = Span::concat(&spans[*spans_idx..]);
let type_bytes = working_set.get_span_contents(full_span).to_vec();
2021-07-16 08:24:46 +02:00
let (tokens, parse_error) =
lex_signature(&type_bytes, full_span.start, &[b','], &[], true);
if let Some(parse_error) = parse_error {
working_set.error(parse_error);
}
let ty = parse_type(working_set, &type_bytes, tokens[0].span);
Fix: missing parse error when extra tokens are given to let bindings (#12238) Manual checks are added to `parse_let`, `parse_mut`, and `parse_const`. `parse_var_with_opt_type` is also fixed to update `spans_idx` correctly. Fixes #12125. It's technically a fix, but I'd rather not merge this directly. I'm making this PR to bring into attention the code quality of the parser code. For example: * Inconsistent usage of `spans_idx`. What is its purpose, and which parsing functions need it? I suspect it's possible to remove the usage of `spans_idx` entirely. * Lacking documentation for top-level functions. What does `mutable` mean for `parse_var_with_opt_type()`? * Inconsistent error reporting. Usage of both `working_set.error()` and `working_set.parse_errors.push()`. Using `ParseError::Expected` for an invalid variable name when there's `ParseError::VariableNotValid` (from `parser.rs:5237`). Checking variable names manually when there's `is_variable()` (from `parser.rs:2905`). * `span()` is a terrible name for a function that flattens a bunch of spans into one (from `nu-protocal/src/span.rs:92`). The top-level comment (`Used when you have a slice of spans of at least size 1`) doesn't help either. I've only looked at a small portion of the parser code; I expect there are a lot more. These issues made it much harder to fix a simple bug like #12125. I believe we should invest some effort to cleanup the parser code, which will ease maintainance in the future. I'll willing to help if there is interest.
2024-03-21 16:37:52 +01:00
*spans_idx = spans.len() - 1;
2021-07-16 08:24:46 +02:00
let var_name = bytes[0..(bytes.len() - 1)].to_vec();
if !is_variable(&var_name) {
working_set.error(ParseError::Expected(
"valid variable name",
disallow blocks as first-class values (#9587) # Description This PR disallows blocks as first-class values by removing the ability to create them using the `block` syntax shape or type. Now, the parser will only ever be able to create closures as first-class values. This means that `{ 3 }` will always be treated as a closure, unless used in the specifically supported use case of the literal being given as an arg to `for`, `if` and other built-in block users. Note: first-class value here means "value that can be passed into commands and held in variables" # User-Facing Changes This may break some user scripts as `: block` is no longer allows as a type annotation. Note: these cases were not actually supported before, as, for example, passing a block that accessed a variable would have errored when the block was later evaluated. Closures do not have the restriction mentioned above and are the much safer value to pass as first-class, so now they are the only block-like value to be allowed to be passed. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-02 21:40:56 +02:00
spans[*spans_idx - 1],
));
return (garbage(working_set, spans[*spans_idx - 1]), None);
}
let id = working_set.add_variable(var_name, spans[*spans_idx - 1], ty.clone(), mutable);
2021-07-16 08:24:46 +02:00
Improve type hovers (#9515) # Description This PR does a few things to help improve type hovers and, in the process, fixes a few outstanding issues in the type system. Here's a list of the changes: * `for` now will try to infer the type of the iteration variable based on the expression it's given. This fixes things like `for x in [1, 2, 3] { }` where `x` now properly gets the int type. * Removed old input/output type fields from the signature, focuses on the vec of signatures. Updated a bunch of dataframe commands that hadn't moved over. This helps tie things together a bit better * Fixed inference of types from subexpressions to use the last expression in the block * Fixed handling of explicit types in `let` and `mut` calls, so we now respect that as the authoritative type I also tried to add `def` input/output type inference, but unfortunately we only know the predecl types universally, which means we won't have enough information to properly know what the types of the custom commands are. # User-Facing Changes Script typechecking will get tighter in some cases Hovers should be more accurate in some cases that previously resorted to any. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: Darren Schroeder <343840+fdncred@users.noreply.github.com>
2023-06-28 19:19:48 +02:00
(
Expression::new(
working_set,
Expr::VarDecl(id),
Span::concat(&spans[span_beginning..*spans_idx + 1]),
ty.clone(),
),
Improve type hovers (#9515) # Description This PR does a few things to help improve type hovers and, in the process, fixes a few outstanding issues in the type system. Here's a list of the changes: * `for` now will try to infer the type of the iteration variable based on the expression it's given. This fixes things like `for x in [1, 2, 3] { }` where `x` now properly gets the int type. * Removed old input/output type fields from the signature, focuses on the vec of signatures. Updated a bunch of dataframe commands that hadn't moved over. This helps tie things together a bit better * Fixed inference of types from subexpressions to use the last expression in the block * Fixed handling of explicit types in `let` and `mut` calls, so we now respect that as the authoritative type I also tried to add `def` input/output type inference, but unfortunately we only know the predecl types universally, which means we won't have enough information to properly know what the types of the custom commands are. # User-Facing Changes Script typechecking will get tighter in some cases Hovers should be more accurate in some cases that previously resorted to any. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: Darren Schroeder <343840+fdncred@users.noreply.github.com>
2023-06-28 19:19:48 +02:00
Some(ty),
)
2021-07-16 08:24:46 +02:00
} else {
let var_name = bytes[0..(bytes.len() - 1)].to_vec();
if !is_variable(&var_name) {
working_set.error(ParseError::Expected(
"valid variable name",
spans[*spans_idx],
));
return (garbage(working_set, spans[*spans_idx]), None);
}
let id = working_set.add_variable(var_name, spans[*spans_idx], Type::Any, mutable);
working_set.error(ParseError::MissingType(spans[*spans_idx]));
Improve type hovers (#9515) # Description This PR does a few things to help improve type hovers and, in the process, fixes a few outstanding issues in the type system. Here's a list of the changes: * `for` now will try to infer the type of the iteration variable based on the expression it's given. This fixes things like `for x in [1, 2, 3] { }` where `x` now properly gets the int type. * Removed old input/output type fields from the signature, focuses on the vec of signatures. Updated a bunch of dataframe commands that hadn't moved over. This helps tie things together a bit better * Fixed inference of types from subexpressions to use the last expression in the block * Fixed handling of explicit types in `let` and `mut` calls, so we now respect that as the authoritative type I also tried to add `def` input/output type inference, but unfortunately we only know the predecl types universally, which means we won't have enough information to properly know what the types of the custom commands are. # User-Facing Changes Script typechecking will get tighter in some cases Hovers should be more accurate in some cases that previously resorted to any. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: Darren Schroeder <343840+fdncred@users.noreply.github.com>
2023-06-28 19:19:48 +02:00
(
Expression::new(working_set, Expr::VarDecl(id), spans[*spans_idx], Type::Any),
Improve type hovers (#9515) # Description This PR does a few things to help improve type hovers and, in the process, fixes a few outstanding issues in the type system. Here's a list of the changes: * `for` now will try to infer the type of the iteration variable based on the expression it's given. This fixes things like `for x in [1, 2, 3] { }` where `x` now properly gets the int type. * Removed old input/output type fields from the signature, focuses on the vec of signatures. Updated a bunch of dataframe commands that hadn't moved over. This helps tie things together a bit better * Fixed inference of types from subexpressions to use the last expression in the block * Fixed handling of explicit types in `let` and `mut` calls, so we now respect that as the authoritative type I also tried to add `def` input/output type inference, but unfortunately we only know the predecl types universally, which means we won't have enough information to properly know what the types of the custom commands are. # User-Facing Changes Script typechecking will get tighter in some cases Hovers should be more accurate in some cases that previously resorted to any. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: Darren Schroeder <343840+fdncred@users.noreply.github.com>
2023-06-28 19:19:48 +02:00
None,
)
2021-07-16 08:24:46 +02:00
}
2021-09-02 10:25:22 +02:00
} else {
let var_name = bytes;
if !is_variable(&var_name) {
working_set.error(ParseError::Expected(
"valid variable name",
spans[*spans_idx],
));
return (garbage(working_set, spans[*spans_idx]), None);
}
let id = working_set.add_variable(
var_name,
Span::concat(&spans[*spans_idx..*spans_idx + 1]),
Type::Any,
mutable,
);
2021-07-08 00:55:46 +02:00
Improve type hovers (#9515) # Description This PR does a few things to help improve type hovers and, in the process, fixes a few outstanding issues in the type system. Here's a list of the changes: * `for` now will try to infer the type of the iteration variable based on the expression it's given. This fixes things like `for x in [1, 2, 3] { }` where `x` now properly gets the int type. * Removed old input/output type fields from the signature, focuses on the vec of signatures. Updated a bunch of dataframe commands that hadn't moved over. This helps tie things together a bit better * Fixed inference of types from subexpressions to use the last expression in the block * Fixed handling of explicit types in `let` and `mut` calls, so we now respect that as the authoritative type I also tried to add `def` input/output type inference, but unfortunately we only know the predecl types universally, which means we won't have enough information to properly know what the types of the custom commands are. # User-Facing Changes Script typechecking will get tighter in some cases Hovers should be more accurate in some cases that previously resorted to any. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: Darren Schroeder <343840+fdncred@users.noreply.github.com>
2023-06-28 19:19:48 +02:00
(
Expression::new(working_set, Expr::VarDecl(id), spans[*spans_idx], Type::Any),
Improve type hovers (#9515) # Description This PR does a few things to help improve type hovers and, in the process, fixes a few outstanding issues in the type system. Here's a list of the changes: * `for` now will try to infer the type of the iteration variable based on the expression it's given. This fixes things like `for x in [1, 2, 3] { }` where `x` now properly gets the int type. * Removed old input/output type fields from the signature, focuses on the vec of signatures. Updated a bunch of dataframe commands that hadn't moved over. This helps tie things together a bit better * Fixed inference of types from subexpressions to use the last expression in the block * Fixed handling of explicit types in `let` and `mut` calls, so we now respect that as the authoritative type I also tried to add `def` input/output type inference, but unfortunately we only know the predecl types universally, which means we won't have enough information to properly know what the types of the custom commands are. # User-Facing Changes Script typechecking will get tighter in some cases Hovers should be more accurate in some cases that previously resorted to any. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: Darren Schroeder <343840+fdncred@users.noreply.github.com>
2023-06-28 19:19:48 +02:00
None,
)
2021-09-02 10:25:22 +02:00
}
}
2021-09-09 23:47:20 +02:00
pub fn expand_to_cell_path(
working_set: &mut StateWorkingSet,
expression: &mut Expression,
var_id: VarId,
) {
trace!("parsing: expanding to cell path");
2021-09-09 23:47:20 +02:00
if let Expression {
expr: Expr::String(_),
span,
..
} = expression
{
// Re-parse the string as if it were a cell-path
let new_expression = parse_full_cell_path(working_set, Some(var_id), *span);
2021-09-09 23:47:20 +02:00
*expression = new_expression;
}
Fix precedence of 'not' operator (#11672) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> # Description A bit hackish but this fixes the precedence of the `not` operator. Before: `not false and false` => true Now: `not false and false` => false Fixes #11633 # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass (on Windows make sure to [enable developer mode](https://learn.microsoft.com/en-us/windows/apps/get-started/developer-mode-features-and-debugging)) - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: Jakub Žádník <kubouch@gmail.com>
2024-01-29 20:42:27 +01:00
if let Expression {
expr: Expr::UnaryNot(inner),
..
} = expression
{
expand_to_cell_path(working_set, inner, var_id);
}
2021-09-09 23:47:20 +02:00
}
Custom command input/output types (#9690) # Description This adds input/output types to custom commands. These are input/output pairs that related an input type to an output type. For example (a single int-to-int input/output pair): ``` def foo []: int -> int { ... } ``` You can also have multiple input/output pairs: ``` def bar []: [int -> string, string -> list<string>] { ... } ``` These types are checked during definition time in the parser. If the block does not match the type, the user will get a parser error. This `:` to begin the input/output signatures should immediately follow the argument signature as shown above. The PR also improves type parsing by re-using the shape parser. The shape parser is now the canonical way to parse types/shapes in user code. This PR also splits `extern` into `extern`/`extern-wrapped` because of the parser limitation that a multi-span argument (which Signature now is) can't precede an optional argument. `extern-wrapped` now takes the required block that was previously optional. # User-Facing Changes The change to `extern` to split into `extern` and `extern-wrapped` is a breaking change. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-14 23:51:28 +02:00
pub fn parse_input_output_types(
working_set: &mut StateWorkingSet,
spans: &[Span],
) -> Vec<(Type, Type)> {
let mut full_span = Span::concat(spans);
Custom command input/output types (#9690) # Description This adds input/output types to custom commands. These are input/output pairs that related an input type to an output type. For example (a single int-to-int input/output pair): ``` def foo []: int -> int { ... } ``` You can also have multiple input/output pairs: ``` def bar []: [int -> string, string -> list<string>] { ... } ``` These types are checked during definition time in the parser. If the block does not match the type, the user will get a parser error. This `:` to begin the input/output signatures should immediately follow the argument signature as shown above. The PR also improves type parsing by re-using the shape parser. The shape parser is now the canonical way to parse types/shapes in user code. This PR also splits `extern` into `extern`/`extern-wrapped` because of the parser limitation that a multi-span argument (which Signature now is) can't precede an optional argument. `extern-wrapped` now takes the required block that was previously optional. # User-Facing Changes The change to `extern` to split into `extern` and `extern-wrapped` is a breaking change. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-14 23:51:28 +02:00
let mut bytes = working_set.get_span_contents(full_span);
if bytes.starts_with(b"[") {
bytes = &bytes[1..];
full_span.start += 1;
}
if bytes.ends_with(b"]") {
bytes = &bytes[..(bytes.len() - 1)];
full_span.end -= 1;
}
let (tokens, parse_error) =
lex_signature(bytes, full_span.start, &[b'\n', b'\r', b','], &[], true);
Custom command input/output types (#9690) # Description This adds input/output types to custom commands. These are input/output pairs that related an input type to an output type. For example (a single int-to-int input/output pair): ``` def foo []: int -> int { ... } ``` You can also have multiple input/output pairs: ``` def bar []: [int -> string, string -> list<string>] { ... } ``` These types are checked during definition time in the parser. If the block does not match the type, the user will get a parser error. This `:` to begin the input/output signatures should immediately follow the argument signature as shown above. The PR also improves type parsing by re-using the shape parser. The shape parser is now the canonical way to parse types/shapes in user code. This PR also splits `extern` into `extern`/`extern-wrapped` because of the parser limitation that a multi-span argument (which Signature now is) can't precede an optional argument. `extern-wrapped` now takes the required block that was previously optional. # User-Facing Changes The change to `extern` to split into `extern` and `extern-wrapped` is a breaking change. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-14 23:51:28 +02:00
if let Some(parse_error) = parse_error {
working_set.error(parse_error);
Custom command input/output types (#9690) # Description This adds input/output types to custom commands. These are input/output pairs that related an input type to an output type. For example (a single int-to-int input/output pair): ``` def foo []: int -> int { ... } ``` You can also have multiple input/output pairs: ``` def bar []: [int -> string, string -> list<string>] { ... } ``` These types are checked during definition time in the parser. If the block does not match the type, the user will get a parser error. This `:` to begin the input/output signatures should immediately follow the argument signature as shown above. The PR also improves type parsing by re-using the shape parser. The shape parser is now the canonical way to parse types/shapes in user code. This PR also splits `extern` into `extern`/`extern-wrapped` because of the parser limitation that a multi-span argument (which Signature now is) can't precede an optional argument. `extern-wrapped` now takes the required block that was previously optional. # User-Facing Changes The change to `extern` to split into `extern` and `extern-wrapped` is a breaking change. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-14 23:51:28 +02:00
}
let mut output = vec![];
let mut idx = 0;
while idx < tokens.len() {
let type_bytes = working_set.get_span_contents(tokens[idx].span).to_vec();
let input_type = parse_type(working_set, &type_bytes, tokens[idx].span);
idx += 1;
if idx >= tokens.len() {
working_set.error(ParseError::Expected(
"arrow (->)",
Span::new(tokens[idx - 1].span.end, tokens[idx - 1].span.end),
));
break;
}
let arrow = working_set.get_span_contents(tokens[idx].span);
if arrow != b"->" {
working_set.error(ParseError::Expected("arrow (->)", tokens[idx].span));
}
idx += 1;
if idx >= tokens.len() {
working_set.error(ParseError::MissingType(Span::new(
tokens[idx - 1].span.end,
tokens[idx - 1].span.end,
)));
break;
}
let type_bytes = working_set.get_span_contents(tokens[idx].span).to_vec();
let output_type = parse_type(working_set, &type_bytes, tokens[idx].span);
output.push((input_type, output_type));
idx += 1;
}
output
}
pub fn parse_full_signature(working_set: &mut StateWorkingSet, spans: &[Span]) -> Expression {
let arg_signature = working_set.get_span_contents(spans[0]);
if arg_signature.ends_with(b":") {
let mut arg_signature =
parse_signature(working_set, Span::new(spans[0].start, spans[0].end - 1));
let input_output_types = parse_input_output_types(working_set, &spans[1..]);
if let Expression {
expr: Expr::Signature(sig),
span: expr_span,
..
} = &mut arg_signature
{
sig.input_output_types = input_output_types;
expr_span.end = Span::concat(&spans[1..]).end;
Custom command input/output types (#9690) # Description This adds input/output types to custom commands. These are input/output pairs that related an input type to an output type. For example (a single int-to-int input/output pair): ``` def foo []: int -> int { ... } ``` You can also have multiple input/output pairs: ``` def bar []: [int -> string, string -> list<string>] { ... } ``` These types are checked during definition time in the parser. If the block does not match the type, the user will get a parser error. This `:` to begin the input/output signatures should immediately follow the argument signature as shown above. The PR also improves type parsing by re-using the shape parser. The shape parser is now the canonical way to parse types/shapes in user code. This PR also splits `extern` into `extern`/`extern-wrapped` because of the parser limitation that a multi-span argument (which Signature now is) can't precede an optional argument. `extern-wrapped` now takes the required block that was previously optional. # User-Facing Changes The change to `extern` to split into `extern` and `extern-wrapped` is a breaking change. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-14 23:51:28 +02:00
}
arg_signature
} else {
parse_signature(working_set, spans[0])
}
}
pub fn parse_row_condition(working_set: &mut StateWorkingSet, spans: &[Span]) -> Expression {
let var_id = working_set.add_variable(b"$it".to_vec(), Span::concat(spans), Type::Any, false);
let expression = parse_math_expression(working_set, spans, Some(var_id));
let span = Span::concat(spans);
2021-11-26 04:49:03 +01:00
let block_id = match expression.expr {
Expr::Block(block_id) => block_id,
Expr::Closure(block_id) => block_id,
2021-11-26 04:49:03 +01:00
_ => {
// We have an expression, so let's convert this into a block.
let mut block = Block::new();
let mut pipeline = Pipeline::new();
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
pipeline.elements.push(PipelineElement {
pipe: None,
expr: expression,
redirection: None,
});
2021-11-26 04:49:03 +01:00
block.pipelines.push(pipeline);
2021-11-26 04:49:03 +01:00
block.signature.required_positional.push(PositionalArg {
name: "$it".into(),
desc: "row condition".into(),
shape: SyntaxShape::Any,
var_id: Some(var_id),
2022-03-07 21:08:56 +01:00
default_value: None,
2021-11-26 04:49:03 +01:00
});
Make EngineState clone cheaper with Arc on all of the heavy objects (#12229) # Description This makes many of the larger objects in `EngineState` into `Arc`, and uses `Arc::make_mut` to do clone-on-write if the reference is not unique. This is generally very cheap, giving us the best of both worlds - allowing us to mutate without cloning if we have an exclusive reference, and cloning if we don't. This started as more of a curiosity for me after remembering that `Arc::make_mut` exists and can make using `Arc` for mostly immutable data that sometimes needs to be changed very convenient, and also after hearing someone complain about memory usage on Discord - this is a somewhat significant win for that. The exact objects that were wrapped in `Arc`: - `files`, `file_contents` - the strings and byte buffers - `decls` - the whole `Vec`, but mostly to avoid lots of individual `malloc()` calls on Clone rather than for memory usage - `blocks` - the blocks themselves, rather than the outer Vec - `modules` - the modules themselves, rather than the outer Vec - `env_vars`, `previous_env_vars` - the entire maps - `config` The changes required were relatively minimal, but this is a breaking API change. In particular, blocks are added as Arcs, to allow the parser cache functionality to work. With my normal nu config, running on Linux, this saves me about 15 MiB of process memory usage when running interactively (65 MiB → 50 MiB). This also makes quick command executions cheaper, particularly since every REPL loop now involves a clone of the engine state so that we can recover from a panic. It also reduces memory usage where engine state needs to be cloned and sent to another thread or kept within an iterator. # User-Facing Changes Shouldn't be any, since it's all internal stuff, but it does change some public interfaces so it's a breaking change
2024-03-19 19:07:00 +01:00
working_set.add_block(Arc::new(block))
2021-11-26 04:49:03 +01:00
}
};
Expression::new(working_set, Expr::RowCondition(block_id), span, Type::Bool)
2021-09-02 10:25:22 +02:00
}
2021-07-16 23:55:12 +02:00
pub fn parse_signature(working_set: &mut StateWorkingSet, span: Span) -> Expression {
2021-09-02 10:25:22 +02:00
let bytes = working_set.get_span_contents(span);
2021-07-16 23:55:12 +02:00
2021-09-02 10:25:22 +02:00
let mut start = span.start;
let mut end = span.end;
2021-07-16 23:55:12 +02:00
let mut has_paren = false;
2021-09-02 10:25:22 +02:00
if bytes.starts_with(b"[") {
start += 1;
} else if bytes.starts_with(b"(") {
has_paren = true;
start += 1;
2021-10-11 22:58:38 +02:00
} else {
working_set.error(ParseError::Expected("[ or (", Span::new(start, start + 1)));
return garbage(working_set, span);
2021-09-02 10:25:22 +02:00
}
2021-10-11 22:58:38 +02:00
if (has_paren && bytes.ends_with(b")")) || (!has_paren && bytes.ends_with(b"]")) {
2021-09-02 10:25:22 +02:00
end -= 1;
} else {
working_set.error(ParseError::Unclosed("] or )".into(), Span::new(end, end)));
2021-09-02 10:25:22 +02:00
}
2021-07-16 23:55:12 +02:00
let sig = parse_signature_helper(working_set, Span::new(start, end));
2021-09-06 01:16:27 +02:00
Expression::new(working_set, Expr::Signature(sig), span, Type::Signature)
2021-09-06 01:16:27 +02:00
}
pub fn parse_signature_helper(working_set: &mut StateWorkingSet, span: Span) -> Box<Signature> {
2021-09-06 01:16:27 +02:00
enum ParseMode {
Arg,
AfterCommaArg,
Type,
DefaultValue,
2021-09-06 01:16:27 +02:00
}
2022-03-07 21:08:56 +01:00
#[derive(Debug)]
2021-09-06 01:16:27 +02:00
enum Arg {
Positional {
arg: PositionalArg,
required: bool,
type_annotated: bool,
},
RestPositional(PositionalArg),
Flag {
flag: Flag,
type_annotated: bool,
},
2021-09-06 01:16:27 +02:00
}
2021-09-02 10:25:22 +02:00
let source = working_set.get_span_contents(span);
2021-07-16 23:55:12 +02:00
allow lists to have type annotations (#8529) this pr refines #8270 and closes #8109 # description examples: the original syntax is okay ```nu def okay [nums: list] {} # the type of list will be list<any> ``` empty annotations are allowed in any variation the last two may be caught by a future formatter, but do not affect `nu` code currently ```nu def okay [nums: list<>] {} # okay def okay [nums: list< >] {} # weird but also okay def okay [nums: list< >] {} # also weird but okay ``` types are allowed (See [notes](#notes) below) ```nu def okay [nums: list<int>] {} # `test [a b c]` will throw an error def okay [nums: list< int > {} # any amount of space within the angle brackets is okay def err [nums: list <int>] {} # this is not okay, `nums` and `<int>` will be parsed as # two separate params, ``` nested annotations are allowed in many variations ```nu def okay [items: list<list<int>>] {} def okay [items: list<list>] {} ``` any unterminated annotation is caught ```nu Error: nu::parser::unexpected_eof × Unexpected end of code. ╭─[source:1:1] 1 │ def err [nums: list<int] {} · ▲ · ╰── expected closing > ╰──── ``` unknown types are flagged ```nu Error: nu::parser::unknown_type × Unknown type. ╭─[source:1:1] 1 │ def err [nums: list<str>] {} · ─┬─ · ╰── unknown type ╰──── Error: nu::parser::unknown_type × Unknown type. ╭─[source:1:1] 1 │ def err [nums: list<int, string>] {} · ─────┬───── · ╰── unknown type ╰──── ``` # notes the error message for mismatched types in not as intuitive ```nu Error: nu::parser::parse_mismatch × Parse mismatch during operation. ╭─[source:1:1] 1 │ def err [nums: list<int>] {}; err [a b c] · ┬ · ╰── expected int ╰──── ``` it should be something like this ```nu Error: nu::parser::parse_mismatch × Parse mismatch during operation. ╭─[source:1:1] 1 │ def err [nums: list<int>] {}; err [a b c] · ──┬── · ╰── expected list<int> ╰──── ``` this is currently not implemented
2023-03-24 12:54:06 +01:00
let (output, err) = lex_signature(
2022-03-07 21:08:56 +01:00
source,
span.start,
&[b'\n', b'\r'],
&[b':', b'=', b','],
2022-03-07 21:08:56 +01:00
false,
);
if let Some(err) = err {
working_set.error(err);
}
2021-07-16 23:55:12 +02:00
2021-09-02 10:25:22 +02:00
let mut args: Vec<Arg> = vec![];
let mut parse_mode = ParseMode::Arg;
2021-07-16 23:55:12 +02:00
2021-09-02 10:25:22 +02:00
for token in &output {
match token {
Token {
contents: crate::TokenContents::Item,
span,
} => {
let span = *span;
let contents = working_set.get_span_contents(span).to_vec();
2021-09-02 10:25:22 +02:00
// The : symbol separates types
2021-09-02 10:25:22 +02:00
if contents == b":" {
match parse_mode {
ParseMode::Arg => {
parse_mode = ParseMode::Type;
2021-07-16 23:55:12 +02:00
}
ParseMode::AfterCommaArg => {
working_set.error(ParseError::Expected("parameter or flag", span));
}
ParseMode::Type | ParseMode::DefaultValue => {
2021-09-02 10:25:22 +02:00
// We're seeing two types for the same thing for some reason, error
working_set.error(ParseError::Expected("type", span));
2021-09-02 10:25:22 +02:00
}
}
}
// The = symbol separates a variable from its default value
else if contents == b"=" {
2022-03-07 21:08:56 +01:00
match parse_mode {
ParseMode::Type | ParseMode::Arg => {
parse_mode = ParseMode::DefaultValue;
2022-03-07 21:08:56 +01:00
}
ParseMode::AfterCommaArg => {
working_set.error(ParseError::Expected("parameter or flag", span));
}
ParseMode::DefaultValue => {
2022-03-07 21:08:56 +01:00
// We're seeing two default values for some reason, error
working_set.error(ParseError::Expected("default value", span));
2022-03-07 21:08:56 +01:00
}
}
}
// The , symbol separates params only
else if contents == b"," {
match parse_mode {
ParseMode::Arg => parse_mode = ParseMode::AfterCommaArg,
ParseMode::AfterCommaArg => {
working_set.error(ParseError::Expected("parameter or flag", span));
}
ParseMode::Type => {
working_set.error(ParseError::Expected("type", span));
}
ParseMode::DefaultValue => {
working_set.error(ParseError::Expected("default value", span));
}
}
2021-09-02 10:25:22 +02:00
} else {
match parse_mode {
ParseMode::Arg | ParseMode::AfterCommaArg => {
// Long flag with optional short form following with no whitespace, e.g. --output, --age(-a)
2021-09-02 10:25:22 +02:00
if contents.starts_with(b"--") && contents.len() > 2 {
// Split the long flag from the short flag with the ( character as delimiter.
// The trailing ) is removed further down.
2021-09-02 10:25:22 +02:00
let flags: Vec<_> =
contents.split(|x| x == &b'(').map(|x| x.to_vec()).collect();
2021-10-13 19:53:27 +02:00
let long = String::from_utf8_lossy(&flags[0][2..]).to_string();
let mut variable_name = flags[0][2..].to_vec();
// Replace the '-' in a variable name with '_'
(0..variable_name.len()).for_each(|idx| {
if variable_name[idx] == b'-' {
variable_name[idx] = b'_';
}
});
if !is_variable(&variable_name) {
working_set.error(ParseError::Expected(
"valid variable name for this long flag",
span,
))
}
2022-03-09 10:42:19 +01:00
let var_id =
working_set.add_variable(variable_name, span, Type::Any, false);
2021-09-02 10:25:22 +02:00
// If there's no short flag, exit now. Otherwise, parse it.
2021-09-02 10:25:22 +02:00
if flags.len() == 1 {
args.push(Arg::Flag {
flag: Flag {
arg: None,
desc: String::new(),
long,
short: None,
required: false,
var_id: Some(var_id),
default_value: None,
},
type_annotated: false,
});
} else if flags.len() >= 3 {
working_set.error(ParseError::Expected(
"only one short flag alternative",
span,
));
2021-09-02 10:25:22 +02:00
} else {
let short_flag = &flags[1];
let short_flag = if !short_flag.starts_with(b"-")
|| !short_flag.ends_with(b")")
{
working_set.error(ParseError::Expected(
"short flag alternative for the long flag",
span,
));
2021-09-02 10:25:22 +02:00
short_flag
2021-07-17 20:52:50 +02:00
} else {
// Obtain the flag's name by removing the starting - and trailing )
2021-09-02 10:25:22 +02:00
&short_flag[1..(short_flag.len() - 1)]
};
// Note that it is currently possible to make a short flag with non-alphanumeric characters,
// like -).
2021-07-17 00:39:30 +02:00
let short_flag =
String::from_utf8_lossy(short_flag).to_string();
let chars: Vec<char> = short_flag.chars().collect();
2021-10-13 19:53:27 +02:00
let long = String::from_utf8_lossy(&flags[0][2..]).to_string();
let mut variable_name = flags[0][2..].to_vec();
(0..variable_name.len()).for_each(|idx| {
if variable_name[idx] == b'-' {
variable_name[idx] = b'_';
}
});
if !is_variable(&variable_name) {
working_set.error(ParseError::Expected(
"valid variable name for this short flag",
span,
))
}
let var_id = working_set.add_variable(
variable_name,
span,
Type::Any,
false,
);
2021-07-17 00:39:30 +02:00
2021-09-02 10:25:22 +02:00
if chars.len() == 1 {
args.push(Arg::Flag {
flag: Flag {
arg: None,
desc: String::new(),
long,
short: Some(chars[0]),
required: false,
var_id: Some(var_id),
default_value: None,
},
type_annotated: false,
});
2021-09-02 10:25:22 +02:00
} else {
working_set.error(ParseError::Expected("short flag", span));
2021-09-02 10:25:22 +02:00
}
}
parse_mode = ParseMode::Arg;
}
// Mandatory short flag, e.g. -e (must be one character)
else if contents.starts_with(b"-") && contents.len() > 1 {
2021-09-02 10:25:22 +02:00
let short_flag = &contents[1..];
let short_flag = String::from_utf8_lossy(short_flag).to_string();
let chars: Vec<char> = short_flag.chars().collect();
if chars.len() > 1 {
working_set.error(ParseError::Expected("short flag", span));
2022-01-06 22:06:54 +01:00
}
2021-09-02 10:25:22 +02:00
2022-01-06 22:06:54 +01:00
let mut encoded_var_name = vec![0u8; 4];
let len = chars[0].encode_utf8(&mut encoded_var_name).len();
let variable_name = encoded_var_name[0..len].to_vec();
if !is_variable(&variable_name) {
working_set.error(ParseError::Expected(
"valid variable name for this short flag",
span,
))
}
2022-03-09 10:42:19 +01:00
let var_id =
working_set.add_variable(variable_name, span, Type::Any, false);
2021-09-02 10:25:22 +02:00
args.push(Arg::Flag {
flag: Flag {
arg: None,
desc: String::new(),
long: String::new(),
short: Some(chars[0]),
required: false,
var_id: Some(var_id),
default_value: None,
},
type_annotated: false,
});
parse_mode = ParseMode::Arg;
}
// Short flag alias for long flag, e.g. --b (-a)
// This is the same as the short flag in --b(-a)
else if contents.starts_with(b"(-") {
if matches!(parse_mode, ParseMode::AfterCommaArg) {
working_set
.error(ParseError::Expected("parameter or flag", span));
}
2021-09-02 10:25:22 +02:00
let short_flag = &contents[2..];
let short_flag = if !short_flag.ends_with(b")") {
working_set.error(ParseError::Expected("short flag", span));
2021-09-02 10:25:22 +02:00
short_flag
} else {
&short_flag[..(short_flag.len() - 1)]
};
2021-07-17 20:52:50 +02:00
2021-09-02 10:25:22 +02:00
let short_flag = String::from_utf8_lossy(short_flag).to_string();
let chars: Vec<char> = short_flag.chars().collect();
2021-07-17 20:52:50 +02:00
2021-09-02 10:25:22 +02:00
if chars.len() == 1 {
match args.last_mut() {
Some(Arg::Flag { flag, .. }) => {
2021-09-02 10:25:22 +02:00
if flag.short.is_some() {
working_set.error(ParseError::Expected(
"one short flag",
span,
));
2021-09-02 10:25:22 +02:00
} else {
flag.short = Some(chars[0]);
2021-07-17 20:52:50 +02:00
}
}
2021-09-02 10:25:22 +02:00
_ => {
working_set
.error(ParseError::Expected("unknown flag", span));
2021-09-02 10:25:22 +02:00
}
2021-07-17 20:52:50 +02:00
}
2021-07-30 00:56:51 +02:00
} else {
working_set.error(ParseError::Expected("short flag", span));
2021-07-16 23:55:12 +02:00
}
}
// Positional arg, optional
else if contents.ends_with(b"?") {
2021-09-02 10:25:22 +02:00
let contents: Vec<_> = contents[..(contents.len() - 1)].into();
let name = String::from_utf8_lossy(&contents).to_string();
if !is_variable(&contents) {
working_set.error(ParseError::Expected(
"valid variable name for this optional parameter",
span,
))
}
let var_id =
working_set.add_variable(contents, span, Type::Any, false);
2021-09-02 10:25:22 +02:00
args.push(Arg::Positional {
arg: PositionalArg {
2021-09-02 10:25:22 +02:00
desc: String::new(),
name,
shape: SyntaxShape::Any,
var_id: Some(var_id),
2022-03-07 21:08:56 +01:00
default_value: None,
2021-09-02 10:25:22 +02:00
},
required: false,
type_annotated: false,
});
parse_mode = ParseMode::Arg;
}
// Rest param
else if let Some(contents) = contents.strip_prefix(b"...") {
2021-09-07 05:37:02 +02:00
let name = String::from_utf8_lossy(contents).to_string();
let contents_vec: Vec<u8> = contents.to_vec();
if !is_variable(&contents_vec) {
working_set.error(ParseError::Expected(
"valid variable name for this rest parameter",
span,
))
}
2021-09-07 05:37:02 +02:00
2022-03-09 10:42:19 +01:00
let var_id =
working_set.add_variable(contents_vec, span, Type::Any, false);
2021-09-07 05:37:02 +02:00
args.push(Arg::RestPositional(PositionalArg {
desc: String::new(),
name,
shape: SyntaxShape::Any,
var_id: Some(var_id),
2022-03-07 21:08:56 +01:00
default_value: None,
}));
parse_mode = ParseMode::Arg;
}
// Normal param
else {
let name = String::from_utf8_lossy(&contents).to_string();
2021-09-02 10:25:22 +02:00
let contents_vec = contents.to_vec();
if !is_variable(&contents_vec) {
working_set.error(ParseError::Expected(
"valid variable name for this parameter",
span,
))
}
2022-03-09 10:42:19 +01:00
let var_id =
working_set.add_variable(contents_vec, span, Type::Any, false);
2021-09-02 10:25:22 +02:00
// Positional arg, required
args.push(Arg::Positional {
arg: PositionalArg {
2021-09-02 10:25:22 +02:00
desc: String::new(),
name,
shape: SyntaxShape::Any,
var_id: Some(var_id),
2022-03-07 21:08:56 +01:00
default_value: None,
2021-09-02 10:25:22 +02:00
},
required: true,
type_annotated: false,
});
parse_mode = ParseMode::Arg;
2021-07-16 23:55:12 +02:00
}
2021-09-02 10:25:22 +02:00
}
ParseMode::Type => {
2021-09-02 10:25:22 +02:00
if let Some(last) = args.last_mut() {
let syntax_shape = parse_shape_name(
working_set,
&contents,
span,
ShapeDescriptorUse::Argument,
);
2021-10-12 19:44:23 +02:00
//TODO check if we're replacing a custom parameter already
2021-09-02 10:25:22 +02:00
match last {
Arg::Positional {
arg: PositionalArg { shape, var_id, .. },
required: _,
type_annotated,
} => {
2021-09-02 10:25:22 +02:00
working_set.set_variable_type(var_id.expect("internal error: all custom parameters must have var_ids"), syntax_shape.to_type());
*shape = syntax_shape;
*type_annotated = true;
2021-09-02 10:25:22 +02:00
}
Arg::RestPositional(PositionalArg {
shape, var_id, ..
}) => {
working_set.set_variable_type(var_id.expect("internal error: all custom parameters must have var_ids"), Type::List(Box::new(syntax_shape.to_type())));
*shape = syntax_shape;
}
Arg::Flag {
flag: Flag { arg, var_id, .. },
type_annotated,
} => {
working_set.set_variable_type(var_id.expect("internal error: all custom parameters must have var_ids"), syntax_shape.to_type());
if syntax_shape == SyntaxShape::Boolean {
working_set.error(ParseError::LabeledError(
"Type annotations are not allowed for boolean switches.".to_string(),
"Remove the `: bool` type annotation.".to_string(),
span,
));
}
*arg = Some(syntax_shape);
*type_annotated = true;
2021-07-16 23:55:12 +02:00
}
}
}
parse_mode = ParseMode::Arg;
2021-07-16 23:55:12 +02:00
}
ParseMode::DefaultValue => {
2022-03-07 21:08:56 +01:00
if let Some(last) = args.last_mut() {
let expression = parse_value(working_set, span, &SyntaxShape::Any);
2022-03-07 21:08:56 +01:00
//TODO check if we're replacing a custom parameter already
match last {
Arg::Positional {
arg:
PositionalArg {
shape,
var_id,
default_value,
..
},
2022-03-07 21:08:56 +01:00
required,
type_annotated,
} => {
2022-03-07 21:08:56 +01:00
let var_id = var_id.expect("internal error: all custom parameters must have var_ids");
2022-03-09 10:42:19 +01:00
let var_type = &working_set.get_variable(var_id).ty;
2022-03-07 21:08:56 +01:00
match var_type {
Type::Any => {
if !*type_annotated {
working_set.set_variable_type(
var_id,
expression.ty.clone(),
);
}
2022-03-07 21:08:56 +01:00
}
allow records to have type annotations (#8914) # Description follow up to #8529 cleaned up version of #8892 - the original syntax is okay ```nu def okay [rec: record] {} ``` - you can now add type annotations for fields if you know them before hand ```nu def okay [rec: record<name: string>] {} ``` - you can specify multiple fields ```nu def okay [person: record<name: string age: int>] {} # an optional comma is allowed def okay [person: record<name: string, age: int>] {} ``` - if annotations are specified, any use of the command will be type checked against the specified type ```nu def unwrap [result: record<ok: bool, value: any>] {} unwrap {ok: 2, value: "value"} # errors with Error: nu::parser::type_mismatch × Type mismatch. ╭─[entry #4:1:1] 1 │ unwrap {ok: 2, value: "value"} · ───────┬───── · ╰── expected record<ok: bool, value: any>, found record<ok: int, value: string> ╰──── ``` > here the error is in the `ok` field, since `any` is coerced into any type > as a result `unwrap {ok: true, value: "value"}` is okay - the key must be a string, either quoted or unquoted ```nu def err [rec: record<{}: list>] {} # errors with Error: × `record` type annotations key not string ╭─[entry #7:1:1] 1 │ def unwrap [result: record<{}: bool, value: any>] {} · ─┬ · ╰── must be a string ╰──── ``` - a key doesn't have to have a type in which case it is assumed to be `any` ```nu def okay [person: record<name age>] {} def okay [person: record<name: string age>] {} ``` - however, if you put a colon, you have to specify a type ```nu def err [person: record<name: >] {} # errors with Error: nu::parser::parse_mismatch × Parse mismatch during operation. ╭─[entry #12:1:1] 1 │ def unwrap [res: record<name: >] { $res } · ┬ · ╰── expected type after colon ╰──── ``` # User-Facing Changes **[BREAKING CHANGES]** - this change adds a field to `SyntaxShape::Record` so any plugins that used it will have to update and include the field. though if you are unsure of the type the record expects, `SyntaxShape::Record(vec![])` will suffice
2023-04-26 15:16:55 +02:00
_ => {
if !type_compatible(var_type, &expression.ty) {
working_set.error(
ParseError::AssignmentMismatch(
2022-03-07 21:08:56 +01:00
"Default value wrong type".into(),
format!(
allow records to have type annotations (#8914) # Description follow up to #8529 cleaned up version of #8892 - the original syntax is okay ```nu def okay [rec: record] {} ``` - you can now add type annotations for fields if you know them before hand ```nu def okay [rec: record<name: string>] {} ``` - you can specify multiple fields ```nu def okay [person: record<name: string age: int>] {} # an optional comma is allowed def okay [person: record<name: string, age: int>] {} ``` - if annotations are specified, any use of the command will be type checked against the specified type ```nu def unwrap [result: record<ok: bool, value: any>] {} unwrap {ok: 2, value: "value"} # errors with Error: nu::parser::type_mismatch × Type mismatch. ╭─[entry #4:1:1] 1 │ unwrap {ok: 2, value: "value"} · ───────┬───── · ╰── expected record<ok: bool, value: any>, found record<ok: int, value: string> ╰──── ``` > here the error is in the `ok` field, since `any` is coerced into any type > as a result `unwrap {ok: true, value: "value"}` is okay - the key must be a string, either quoted or unquoted ```nu def err [rec: record<{}: list>] {} # errors with Error: × `record` type annotations key not string ╭─[entry #7:1:1] 1 │ def unwrap [result: record<{}: bool, value: any>] {} · ─┬ · ╰── must be a string ╰──── ``` - a key doesn't have to have a type in which case it is assumed to be `any` ```nu def okay [person: record<name age>] {} def okay [person: record<name: string age>] {} ``` - however, if you put a colon, you have to specify a type ```nu def err [person: record<name: >] {} # errors with Error: nu::parser::parse_mismatch × Parse mismatch during operation. ╭─[entry #12:1:1] 1 │ def unwrap [res: record<name: >] { $res } · ┬ · ╰── expected type after colon ╰──── ``` # User-Facing Changes **[BREAKING CHANGES]** - this change adds a field to `SyntaxShape::Record` so any plugins that used it will have to update and include the field. though if you are unsure of the type the record expects, `SyntaxShape::Record(vec![])` will suffice
2023-04-26 15:16:55 +02:00
"expected default value to be `{var_type}`"
),
2022-03-07 21:08:56 +01:00
expression.span,
),
)
2022-03-07 21:08:56 +01:00
}
}
}
*default_value = if let Ok(constant) =
eval_constant(working_set, &expression)
{
Some(constant)
} else {
working_set.error(ParseError::NonConstantDefaultValue(
expression.span,
));
None
};
if !*type_annotated {
*shape = expression.ty.to_shape();
}
2022-03-07 21:08:56 +01:00
*required = false;
}
Arg::RestPositional(..) => {
working_set.error(ParseError::AssignmentMismatch(
"Rest parameter was given a default value".into(),
"can't have default value".into(),
expression.span,
))
2022-03-07 21:08:56 +01:00
}
Arg::Flag {
flag:
Flag {
arg,
var_id,
default_value,
..
},
type_annotated,
} => {
let expression_span = expression.span;
*default_value = if let Ok(value) =
eval_constant(working_set, &expression)
{
Some(value)
} else {
working_set.error(ParseError::NonConstantDefaultValue(
expression_span,
));
None
};
2022-03-07 21:08:56 +01:00
let var_id = var_id.expect("internal error: all custom parameters must have var_ids");
2022-03-09 10:42:19 +01:00
let var_type = &working_set.get_variable(var_id).ty;
2022-03-07 21:08:56 +01:00
let expression_ty = expression.ty.clone();
// Flags with no TypeMode are just present/not-present switches
// in the case, `var_type` is any.
match var_type {
Type::Any => {
if !*type_annotated {
*arg = Some(expression_ty.to_shape());
working_set
.set_variable_type(var_id, expression_ty);
2022-03-07 21:08:56 +01:00
}
}
t => {
if !type_compatible(t, &expression_ty) {
working_set.error(
ParseError::AssignmentMismatch(
"Default value is the wrong type"
.into(),
format!(
allow records to have type annotations (#8914) # Description follow up to #8529 cleaned up version of #8892 - the original syntax is okay ```nu def okay [rec: record] {} ``` - you can now add type annotations for fields if you know them before hand ```nu def okay [rec: record<name: string>] {} ``` - you can specify multiple fields ```nu def okay [person: record<name: string age: int>] {} # an optional comma is allowed def okay [person: record<name: string, age: int>] {} ``` - if annotations are specified, any use of the command will be type checked against the specified type ```nu def unwrap [result: record<ok: bool, value: any>] {} unwrap {ok: 2, value: "value"} # errors with Error: nu::parser::type_mismatch × Type mismatch. ╭─[entry #4:1:1] 1 │ unwrap {ok: 2, value: "value"} · ───────┬───── · ╰── expected record<ok: bool, value: any>, found record<ok: int, value: string> ╰──── ``` > here the error is in the `ok` field, since `any` is coerced into any type > as a result `unwrap {ok: true, value: "value"}` is okay - the key must be a string, either quoted or unquoted ```nu def err [rec: record<{}: list>] {} # errors with Error: × `record` type annotations key not string ╭─[entry #7:1:1] 1 │ def unwrap [result: record<{}: bool, value: any>] {} · ─┬ · ╰── must be a string ╰──── ``` - a key doesn't have to have a type in which case it is assumed to be `any` ```nu def okay [person: record<name age>] {} def okay [person: record<name: string age>] {} ``` - however, if you put a colon, you have to specify a type ```nu def err [person: record<name: >] {} # errors with Error: nu::parser::parse_mismatch × Parse mismatch during operation. ╭─[entry #12:1:1] 1 │ def unwrap [res: record<name: >] { $res } · ┬ · ╰── expected type after colon ╰──── ``` # User-Facing Changes **[BREAKING CHANGES]** - this change adds a field to `SyntaxShape::Record` so any plugins that used it will have to update and include the field. though if you are unsure of the type the record expects, `SyntaxShape::Record(vec![])` will suffice
2023-04-26 15:16:55 +02:00
"expected default value to be `{t}`"
),
expression_span,
),
)
2022-03-07 21:08:56 +01:00
}
}
}
}
}
}
parse_mode = ParseMode::Arg;
2022-03-07 21:08:56 +01:00
}
2021-07-16 23:55:12 +02:00
}
}
2021-09-02 10:25:22 +02:00
}
Token {
contents: crate::TokenContents::Comment,
span,
} => {
Protocol: debug_assert!() Span to reflect a valid slice (#6806) Also enforce this by #[non_exhaustive] span such that going forward we cannot, in debug builds (1), construct invalid spans. The motivation for this stems from #6431 where I've seen crashes due to invalid slice indexing. My hope is this will mitigate such senarios 1. https://github.com/nushell/nushell/pull/6431#issuecomment-1278147241 # Description (description of your pull request here) # Tests Make sure you've done the following: - [ ] Add tests that cover your changes, either in the command examples, the crate/tests folder, or in the /tests folder. - [ ] Try to think about corner cases and various ways how your changes could break. Cover them with tests. - [ ] If adding tests is not possible, please document in the PR body a minimal example with steps on how to reproduce so one can verify your change works. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [ ] `cargo clippy --workspace --features=extra -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [ ] `cargo test --workspace --features=extra` to check that all the tests pass # Documentation - [ ] If your PR touches a user-facing nushell feature then make sure that there is an entry in the documentation (https://github.com/nushell/nushell.github.io) for the feature, and update it if necessary.
2022-12-03 10:44:12 +01:00
let contents = working_set.get_span_contents(Span::new(span.start + 1, span.end));
2021-09-02 10:25:22 +02:00
let mut contents = String::from_utf8_lossy(contents).to_string();
contents = contents.trim().into();
if let Some(last) = args.last_mut() {
match last {
Arg::Flag { flag, .. } => {
2021-09-02 10:25:22 +02:00
if !flag.desc.is_empty() {
flag.desc.push('\n');
2021-07-17 00:31:36 +02:00
}
2021-09-02 10:25:22 +02:00
flag.desc.push_str(&contents);
}
Arg::Positional {
arg: positional, ..
} => {
2021-09-02 10:25:22 +02:00
if !positional.desc.is_empty() {
positional.desc.push('\n');
2021-07-17 00:31:36 +02:00
}
2021-09-02 10:25:22 +02:00
positional.desc.push_str(&contents);
2021-07-17 00:31:36 +02:00
}
Arg::RestPositional(positional) => {
if !positional.desc.is_empty() {
positional.desc.push('\n');
}
positional.desc.push_str(&contents);
}
2021-07-17 00:31:36 +02:00
}
}
2021-07-16 23:55:12 +02:00
}
2021-09-02 10:25:22 +02:00
_ => {}
2021-07-16 23:55:12 +02:00
}
2021-09-02 10:25:22 +02:00
}
2021-07-16 23:55:12 +02:00
2021-09-02 10:25:22 +02:00
let mut sig = Signature::new(String::new());
for arg in args {
match arg {
Arg::Positional {
arg: positional,
required,
..
} => {
2021-09-07 05:37:02 +02:00
if required {
2022-03-07 21:08:56 +01:00
if !sig.optional_positional.is_empty() {
working_set.error(ParseError::RequiredAfterOptional(
positional.name.clone(),
span,
))
2022-03-07 21:08:56 +01:00
}
2021-09-02 10:25:22 +02:00
sig.required_positional.push(positional)
} else {
sig.optional_positional.push(positional)
2021-07-17 00:53:45 +02:00
}
2021-07-16 23:55:12 +02:00
}
Arg::Flag { flag, .. } => sig.named.push(flag),
Arg::RestPositional(positional) => {
if positional.name.is_empty() {
working_set.error(ParseError::RestNeedsName(span))
} else if sig.rest_positional.is_none() {
sig.rest_positional = Some(PositionalArg {
name: positional.name,
..positional
})
} else {
// Too many rest params
working_set.error(ParseError::MultipleRestParams(span))
}
}
2021-07-16 23:55:12 +02:00
}
}
Box::new(sig)
2021-09-02 10:25:22 +02:00
}
2021-07-08 09:49:17 +02:00
2021-09-02 10:25:22 +02:00
pub fn parse_list_expression(
working_set: &mut StateWorkingSet,
span: Span,
element_shape: &SyntaxShape,
) -> Expression {
2021-09-02 10:25:22 +02:00
let bytes = working_set.get_span_contents(span);
2021-07-08 09:49:17 +02:00
2021-09-02 10:25:22 +02:00
let mut start = span.start;
let mut end = span.end;
2021-07-08 09:49:17 +02:00
2021-09-02 10:25:22 +02:00
if bytes.starts_with(b"[") {
start += 1;
}
if bytes.ends_with(b"]") {
end -= 1;
} else {
working_set.error(ParseError::Unclosed("]".into(), Span::new(end, end)));
2021-09-02 10:25:22 +02:00
}
2021-07-08 09:49:17 +02:00
Protocol: debug_assert!() Span to reflect a valid slice (#6806) Also enforce this by #[non_exhaustive] span such that going forward we cannot, in debug builds (1), construct invalid spans. The motivation for this stems from #6431 where I've seen crashes due to invalid slice indexing. My hope is this will mitigate such senarios 1. https://github.com/nushell/nushell/pull/6431#issuecomment-1278147241 # Description (description of your pull request here) # Tests Make sure you've done the following: - [ ] Add tests that cover your changes, either in the command examples, the crate/tests folder, or in the /tests folder. - [ ] Try to think about corner cases and various ways how your changes could break. Cover them with tests. - [ ] If adding tests is not possible, please document in the PR body a minimal example with steps on how to reproduce so one can verify your change works. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [ ] `cargo clippy --workspace --features=extra -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [ ] `cargo test --workspace --features=extra` to check that all the tests pass # Documentation - [ ] If your PR touches a user-facing nushell feature then make sure that there is an entry in the documentation (https://github.com/nushell/nushell.github.io) for the feature, and update it if necessary.
2022-12-03 10:44:12 +01:00
let inner_span = Span::new(start, end);
let source = working_set.get_span_contents(inner_span);
2021-07-08 09:49:17 +02:00
let (output, err) = lex(source, inner_span.start, &[b'\n', b'\r', b','], &[], true);
if let Some(err) = err {
working_set.error(err)
}
2021-07-08 09:49:17 +02:00
let (mut output, err) = lite_parse(&output);
if let Some(err) = err {
working_set.error(err)
}
2021-07-08 23:45:56 +02:00
2021-09-02 10:25:22 +02:00
let mut args = vec![];
2021-08-17 02:26:05 +02:00
2021-09-02 10:25:22 +02:00
let mut contained_type: Option<Type> = None;
2021-07-08 09:49:17 +02:00
2021-09-02 10:25:22 +02:00
if !output.block.is_empty() {
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
for mut command in output.block.remove(0).commands {
2021-09-02 10:25:22 +02:00
let mut spans_idx = 0;
2021-07-08 23:45:56 +02:00
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
while spans_idx < command.parts.len() {
let curr_span = command.parts[spans_idx];
let curr_tok = working_set.get_span_contents(curr_span);
let (arg, ty) = if curr_tok.starts_with(b"...")
&& curr_tok.len() > 3
&& (curr_tok[3] == b'$' || curr_tok[3] == b'[' || curr_tok[3] == b'(')
{
// Parse the spread operator
// Remove "..." before parsing argument to spread operator
command.parts[spans_idx] = Span::new(curr_span.start + 3, curr_span.end);
let spread_arg = parse_multispan_value(
working_set,
&command.parts,
&mut spans_idx,
&SyntaxShape::List(Box::new(element_shape.clone())),
);
let elem_ty = match &spread_arg.ty {
Type::List(elem_ty) => *elem_ty.clone(),
_ => Type::Any,
};
let span = Span::new(curr_span.start, curr_span.start + 3);
(ListItem::Spread(span, spread_arg), elem_ty)
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
} else {
let arg = parse_multispan_value(
working_set,
&command.parts,
&mut spans_idx,
element_shape,
);
let ty = arg.ty.clone();
(ListItem::Item(arg), ty)
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
};
2021-09-02 10:25:22 +02:00
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
if let Some(ref ctype) = contained_type {
if *ctype != ty {
contained_type = Some(Type::Any);
2021-08-17 02:26:05 +02:00
}
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
} else {
contained_type = Some(ty);
}
2021-08-17 02:26:05 +02:00
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
args.push(arg);
2021-07-16 08:24:46 +02:00
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
spans_idx += 1;
2021-07-08 09:49:17 +02:00
}
}
}
Expression::new(
working_set,
Expr::List(args),
span,
Type::List(Box::new(if let Some(ty) = contained_type {
ty
} else {
Type::Any
})),
)
2021-09-02 10:25:22 +02:00
}
fn parse_table_row(
working_set: &mut StateWorkingSet,
span: Span,
) -> Result<(Vec<Expression>, Span), Span> {
let list = parse_list_expression(working_set, span, &SyntaxShape::Any);
let Expression {
expr: Expr::List(list),
span,
..
} = list
else {
unreachable!("the item must be a list")
};
list.into_iter()
.map(|item| match item {
ListItem::Item(expr) => Ok(expr),
ListItem::Spread(_, spread) => Err(spread.span),
})
.collect::<Result<_, _>>()
.map(|exprs| (exprs, span))
}
fn parse_table_expression(working_set: &mut StateWorkingSet, span: Span) -> Expression {
let bytes = working_set.get_span_contents(span);
let inner_span = {
let start = if bytes.starts_with(b"[") {
span.start + 1
} else {
span.start
};
2021-07-06 00:58:56 +02:00
let end = if bytes.ends_with(b"]") {
span.end - 1
} else {
let end = span.end;
working_set.error(ParseError::Unclosed("]".into(), Span::new(end, end)));
span.end
};
2021-07-06 00:58:56 +02:00
Span::new(start, end)
};
2021-07-06 00:58:56 +02:00
let source = working_set.get_span_contents(inner_span);
let (tokens, err) = lex(source, inner_span.start, &[b'\n', b'\r', b','], &[], true);
if let Some(err) = err {
working_set.error(err);
}
2021-07-06 00:58:56 +02:00
Fix exponential parser time on sequence of [[[[ (#10439) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> # Description <!-- Thank you for improving Nushell. Please, check our [contributing guide](../CONTRIBUTING.md) and talk to the core team before making major changes. Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience. --> Before this change, parsing `[[[[[[[[[[[[[[[[[[[[[[` would cause nushell to consume several gigabytes of memory, now it should be linear in time. The old code first tried parsing the head of the table as a list and then after that it checked if it got more arguments. If it didn't, it throws away the previous result and tries to parse the whole thing as a list, which means we call `parse_list_expression` twice for each call to `parse_table_expression`, resulting in the exponential growth The fix is to simply check that we have all the arguments we need before parsing the head of the table, so we know that we will either call parse_list_expression only on sub-expressions or on the whole thing, never both. Fixes #10438 # User-Facing Changes Should give a noticable speedup when typing a sequence of `[[[[[[` open brackets <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting I would like to add tests, but I'm not sure how to do that without crashing CI with OOM on regression - [x] Don't forget to add tests that cover your changes. - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [x] `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used` to check that you're using the standard code style - [x] `cargo test --workspace` to check that all tests pass (on Windows make sure to [enable developer mode](https://learn.microsoft.com/en-us/windows/apps/get-started/developer-mode-features-and-debugging)) - [x] `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library <!-- > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-09-20 17:53:48 +02:00
// Check that we have all arguments first, before trying to parse the first
// in order to avoid exponential parsing time
let [first, second, rest @ ..] = &tokens[..] else {
return parse_list_expression(working_set, span, &SyntaxShape::Any);
};
Fix exponential parser time on sequence of [[[[ (#10439) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> # Description <!-- Thank you for improving Nushell. Please, check our [contributing guide](../CONTRIBUTING.md) and talk to the core team before making major changes. Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience. --> Before this change, parsing `[[[[[[[[[[[[[[[[[[[[[[` would cause nushell to consume several gigabytes of memory, now it should be linear in time. The old code first tried parsing the head of the table as a list and then after that it checked if it got more arguments. If it didn't, it throws away the previous result and tries to parse the whole thing as a list, which means we call `parse_list_expression` twice for each call to `parse_table_expression`, resulting in the exponential growth The fix is to simply check that we have all the arguments we need before parsing the head of the table, so we know that we will either call parse_list_expression only on sub-expressions or on the whole thing, never both. Fixes #10438 # User-Facing Changes Should give a noticable speedup when typing a sequence of `[[[[[[` open brackets <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting I would like to add tests, but I'm not sure how to do that without crashing CI with OOM on regression - [x] Don't forget to add tests that cover your changes. - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [x] `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used` to check that you're using the standard code style - [x] `cargo test --workspace` to check that all tests pass (on Windows make sure to [enable developer mode](https://learn.microsoft.com/en-us/windows/apps/get-started/developer-mode-features-and-debugging)) - [x] `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library <!-- > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-09-20 17:53:48 +02:00
if !working_set.get_span_contents(first.span).starts_with(b"[")
|| second.contents != TokenContents::Semicolon
|| rest.is_empty()
{
return parse_list_expression(working_set, span, &SyntaxShape::Any);
};
let head = parse_table_row(working_set, first.span);
2021-09-02 10:25:22 +02:00
let errors = working_set.parse_errors.len();
let (head, rows) = match head {
Ok((head, _)) => {
let rows = rest
.iter()
.filter_map(|it| {
use std::cmp::Ordering;
match working_set.get_span_contents(it.span) {
b"," => None,
text if !text.starts_with(b"[") => {
let err = ParseError::LabeledErrorWithHelp {
error: String::from("Table item not list"),
label: String::from("not a list"),
span: it.span,
help: String::from("All table items must be lists"),
};
working_set.error(err);
None
2021-09-08 20:54:27 +02:00
}
_ => match parse_table_row(working_set, it.span) {
Ok((list, span)) => {
match list.len().cmp(&head.len()) {
Ordering::Less => {
let err = ParseError::MissingColumns(head.len(), span);
working_set.error(err);
}
Ordering::Greater => {
let span = {
let start = list[head.len()].span.start;
let end = span.end;
Span::new(start, end)
};
let err = ParseError::ExtraColumns(head.len(), span);
working_set.error(err);
}
Ordering::Equal => {}
}
Some(list)
}
Err(span) => {
let err = ParseError::LabeledError(
String::from("Cannot spread in a table row"),
String::from("invalid spread here"),
span,
);
working_set.error(err);
None
}
},
2021-09-08 20:54:27 +02:00
}
})
.collect();
(head, rows)
}
Err(span) => {
let err = ParseError::LabeledError(
String::from("Cannot spread in a table row"),
String::from("invalid spread here"),
span,
);
working_set.error(err);
(Vec::new(), Vec::new())
}
};
let ty = if working_set.parse_errors.len() == errors {
let (ty, errs) = table_type(&head, &rows);
working_set.parse_errors.extend(errs);
ty
} else {
Type::table()
};
let table = Table {
columns: head.into(),
rows: rows.into_iter().map(Into::into).collect(),
};
Expression::new(working_set, Expr::Table(table), span, ty)
2021-09-02 10:25:22 +02:00
}
2021-07-02 09:32:30 +02:00
fn table_type(head: &[Expression], rows: &[Vec<Expression>]) -> (Type, Vec<ParseError>) {
let mut errors = vec![];
let mut rows = rows.to_vec();
let mut mk_ty = || -> Type {
rows.iter_mut()
.map(|row| row.pop().map(|x| x.ty).unwrap_or_default())
.reduce(|acc, ty| -> Type {
if type_compatible(&acc, &ty) {
ty
} else {
Type::Any
}
})
.unwrap_or_default()
};
let mk_error = |span| ParseError::LabeledErrorWithHelp {
error: "Table column name not string".into(),
label: "must be a string".into(),
help: "Table column names should be able to be converted into strings".into(),
span,
};
let mut ty = head
.iter()
.rev()
.map(|expr| {
if let Some(str) = expr.as_string() {
str
} else {
errors.push(mk_error(expr.span));
String::from("{ column }")
}
})
.map(|title| (title, mk_ty()))
.collect_vec();
ty.reverse();
(Type::Table(ty.into()), errors)
}
pub fn parse_block_expression(working_set: &mut StateWorkingSet, span: Span) -> Expression {
2022-01-01 22:42:50 +01:00
trace!("parsing: block expression");
2021-09-02 10:25:22 +02:00
let bytes = working_set.get_span_contents(span);
2021-07-02 09:32:30 +02:00
2021-09-02 10:25:22 +02:00
let mut start = span.start;
let mut end = span.end;
2021-07-02 09:32:30 +02:00
2021-09-02 10:25:22 +02:00
if bytes.starts_with(b"{") {
start += 1;
} else {
working_set.error(ParseError::Expected("block", span));
return garbage(working_set, span);
2021-09-02 10:25:22 +02:00
}
if bytes.ends_with(b"}") {
end -= 1;
} else {
working_set.error(ParseError::Unclosed("}".into(), Span::new(end, end)));
2021-09-02 10:25:22 +02:00
}
2021-07-02 09:32:30 +02:00
Protocol: debug_assert!() Span to reflect a valid slice (#6806) Also enforce this by #[non_exhaustive] span such that going forward we cannot, in debug builds (1), construct invalid spans. The motivation for this stems from #6431 where I've seen crashes due to invalid slice indexing. My hope is this will mitigate such senarios 1. https://github.com/nushell/nushell/pull/6431#issuecomment-1278147241 # Description (description of your pull request here) # Tests Make sure you've done the following: - [ ] Add tests that cover your changes, either in the command examples, the crate/tests folder, or in the /tests folder. - [ ] Try to think about corner cases and various ways how your changes could break. Cover them with tests. - [ ] If adding tests is not possible, please document in the PR body a minimal example with steps on how to reproduce so one can verify your change works. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [ ] `cargo clippy --workspace --features=extra -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [ ] `cargo test --workspace --features=extra` to check that all the tests pass # Documentation - [ ] If your PR touches a user-facing nushell feature then make sure that there is an entry in the documentation (https://github.com/nushell/nushell.github.io) for the feature, and update it if necessary.
2022-12-03 10:44:12 +01:00
let inner_span = Span::new(start, end);
2021-09-02 10:25:22 +02:00
2022-01-03 06:21:26 +01:00
let source = working_set.get_span_contents(inner_span);
2021-09-02 10:25:22 +02:00
let (output, err) = lex(source, start, &[], &[], false);
if let Some(err) = err {
working_set.error(err);
}
2021-09-02 10:25:22 +02:00
2021-09-06 04:20:02 +02:00
working_set.enter_scope();
2021-09-08 00:00:20 +02:00
// Check to see if we have parameters
let (signature, amt_to_skip): (Option<(Box<Signature>, Span)>, usize) = match output.first() {
Some(Token {
contents: TokenContents::Pipe,
span,
}) => {
working_set.error(ParseError::Expected("block but found closure", *span));
(None, 0)
}
_ => (None, 0),
};
Reuse the cached parse results of parsed files (#8949) # Description This does a lookup in the cache of parsed files to see if a span can be found for a file that was previously loaded with the same contents, then uses that span to find the parsed block for that file. The end result should, in theory, be identical but doesn't require any reparsing or creating new blocks/new definitions that aren't needed. This drops the sg.nu benchmark from: ``` ╭───┬───────────────────╮ │ 0 │ 280ms 606µs 208ns │ │ 1 │ 282ms 654µs 416ns │ │ 2 │ 252ms 640µs 541ns │ │ 3 │ 250ms 940µs 41ns │ │ 4 │ 241ms 216µs 375ns │ │ 5 │ 257ms 310µs 583ns │ │ 6 │ 196ms 739µs 416ns │ ╰───┴───────────────────╯ ``` to: ``` ╭───┬───────────────────╮ │ 0 │ 118ms 698µs 125ns │ │ 1 │ 121ms 327µs │ │ 2 │ 121ms 873µs 500ns │ │ 3 │ 124ms 94µs 708ns │ │ 4 │ 113ms 733µs 291ns │ │ 5 │ 108ms 663µs 125ns │ │ 6 │ 63ms 482µs 625ns │ ╰───┴───────────────────╯ ``` I was hoping to also see some startup time improvements, but I didn't notice much there. # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-04-21 21:00:33 +02:00
let mut output = parse_block(working_set, &output[amt_to_skip..], span, false, false);
if let Some(signature) = signature {
output.signature = signature.0;
}
output.span = Some(span);
working_set.exit_scope();
Make EngineState clone cheaper with Arc on all of the heavy objects (#12229) # Description This makes many of the larger objects in `EngineState` into `Arc`, and uses `Arc::make_mut` to do clone-on-write if the reference is not unique. This is generally very cheap, giving us the best of both worlds - allowing us to mutate without cloning if we have an exclusive reference, and cloning if we don't. This started as more of a curiosity for me after remembering that `Arc::make_mut` exists and can make using `Arc` for mostly immutable data that sometimes needs to be changed very convenient, and also after hearing someone complain about memory usage on Discord - this is a somewhat significant win for that. The exact objects that were wrapped in `Arc`: - `files`, `file_contents` - the strings and byte buffers - `decls` - the whole `Vec`, but mostly to avoid lots of individual `malloc()` calls on Clone rather than for memory usage - `blocks` - the blocks themselves, rather than the outer Vec - `modules` - the modules themselves, rather than the outer Vec - `env_vars`, `previous_env_vars` - the entire maps - `config` The changes required were relatively minimal, but this is a breaking API change. In particular, blocks are added as Arcs, to allow the parser cache functionality to work. With my normal nu config, running on Linux, this saves me about 15 MiB of process memory usage when running interactively (65 MiB → 50 MiB). This also makes quick command executions cheaper, particularly since every REPL loop now involves a clone of the engine state so that we can recover from a panic. It also reduces memory usage where engine state needs to be cloned and sent to another thread or kept within an iterator. # User-Facing Changes Shouldn't be any, since it's all internal stuff, but it does change some public interfaces so it's a breaking change
2024-03-19 19:07:00 +01:00
let block_id = working_set.add_block(Arc::new(output));
Expression::new(working_set, Expr::Block(block_id), span, Type::Block)
}
pub fn parse_match_block_expression(working_set: &mut StateWorkingSet, span: Span) -> Expression {
Add pattern matching (#8590) # Description This adds `match` and basic pattern matching. An example: ``` match $x { 1..10 => { print "Value is between 1 and 10" } { foo: $bar } => { print $"Value has a 'foo' field with value ($bar)" } [$a, $b] => { print $"Value is a list with two items: ($a) and ($b)" } _ => { print "Value is none of the above" } } ``` Like the recent changes to `if` to allow it to be used as an expression, `match` can also be used as an expression. This allows you to assign the result to a variable, eg) `let xyz = match ...` I've also included a short-hand pattern for matching records, as I think it might help when doing a lot of record patterns: `{$foo}` which is equivalent to `{foo: $foo}`. There are still missing components, so consider this the first step in full pattern matching support. Currently missing: * Patterns for strings * Or-patterns (like the `|` in Rust) * Patterns for tables (unclear how we want to match a table, so it'll need some design) * Patterns for binary values * And much more # User-Facing Changes [see above] # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-03-24 02:52:01 +01:00
let bytes = working_set.get_span_contents(span);
let mut start = span.start;
let mut end = span.end;
if bytes.starts_with(b"{") {
start += 1;
} else {
working_set.error(ParseError::Expected("closure", span));
return garbage(working_set, span);
Add pattern matching (#8590) # Description This adds `match` and basic pattern matching. An example: ``` match $x { 1..10 => { print "Value is between 1 and 10" } { foo: $bar } => { print $"Value has a 'foo' field with value ($bar)" } [$a, $b] => { print $"Value is a list with two items: ($a) and ($b)" } _ => { print "Value is none of the above" } } ``` Like the recent changes to `if` to allow it to be used as an expression, `match` can also be used as an expression. This allows you to assign the result to a variable, eg) `let xyz = match ...` I've also included a short-hand pattern for matching records, as I think it might help when doing a lot of record patterns: `{$foo}` which is equivalent to `{foo: $foo}`. There are still missing components, so consider this the first step in full pattern matching support. Currently missing: * Patterns for strings * Or-patterns (like the `|` in Rust) * Patterns for tables (unclear how we want to match a table, so it'll need some design) * Patterns for binary values * And much more # User-Facing Changes [see above] # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-03-24 02:52:01 +01:00
}
if bytes.ends_with(b"}") {
end -= 1;
} else {
working_set.error(ParseError::Unclosed("}".into(), Span::new(end, end)));
Add pattern matching (#8590) # Description This adds `match` and basic pattern matching. An example: ``` match $x { 1..10 => { print "Value is between 1 and 10" } { foo: $bar } => { print $"Value has a 'foo' field with value ($bar)" } [$a, $b] => { print $"Value is a list with two items: ($a) and ($b)" } _ => { print "Value is none of the above" } } ``` Like the recent changes to `if` to allow it to be used as an expression, `match` can also be used as an expression. This allows you to assign the result to a variable, eg) `let xyz = match ...` I've also included a short-hand pattern for matching records, as I think it might help when doing a lot of record patterns: `{$foo}` which is equivalent to `{foo: $foo}`. There are still missing components, so consider this the first step in full pattern matching support. Currently missing: * Patterns for strings * Or-patterns (like the `|` in Rust) * Patterns for tables (unclear how we want to match a table, so it'll need some design) * Patterns for binary values * And much more # User-Facing Changes [see above] # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-03-24 02:52:01 +01:00
}
let inner_span = Span::new(start, end);
let source = working_set.get_span_contents(inner_span);
Allow comments in match blocks (#11717) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> # Description <!-- Thank you for improving Nushell. Please, check our [contributing guide](../CONTRIBUTING.md) and talk to the core team before making major changes. Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience. --> Fix #9878 # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> Writing comments in match blocks will be allowed. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass (on Windows make sure to [enable developer mode](https://learn.microsoft.com/en-us/windows/apps/get-started/developer-mode-features-and-debugging)) - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2024-02-08 00:22:42 +01:00
let (output, err) = lex(source, start, &[b' ', b'\r', b'\n', b',', b'|'], &[], true);
if let Some(err) = err {
working_set.error(err);
}
Add pattern matching (#8590) # Description This adds `match` and basic pattern matching. An example: ``` match $x { 1..10 => { print "Value is between 1 and 10" } { foo: $bar } => { print $"Value has a 'foo' field with value ($bar)" } [$a, $b] => { print $"Value is a list with two items: ($a) and ($b)" } _ => { print "Value is none of the above" } } ``` Like the recent changes to `if` to allow it to be used as an expression, `match` can also be used as an expression. This allows you to assign the result to a variable, eg) `let xyz = match ...` I've also included a short-hand pattern for matching records, as I think it might help when doing a lot of record patterns: `{$foo}` which is equivalent to `{foo: $foo}`. There are still missing components, so consider this the first step in full pattern matching support. Currently missing: * Patterns for strings * Or-patterns (like the `|` in Rust) * Patterns for tables (unclear how we want to match a table, so it'll need some design) * Patterns for binary values * And much more # User-Facing Changes [see above] # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-03-24 02:52:01 +01:00
let mut position = 0;
let mut output_matches = vec![];
while position < output.len() {
// Each match gets its own scope
working_set.enter_scope();
// First parse the pattern
let mut pattern = parse_pattern(working_set, output[position].span);
Add pattern matching (#8590) # Description This adds `match` and basic pattern matching. An example: ``` match $x { 1..10 => { print "Value is between 1 and 10" } { foo: $bar } => { print $"Value has a 'foo' field with value ($bar)" } [$a, $b] => { print $"Value is a list with two items: ($a) and ($b)" } _ => { print "Value is none of the above" } } ``` Like the recent changes to `if` to allow it to be used as an expression, `match` can also be used as an expression. This allows you to assign the result to a variable, eg) `let xyz = match ...` I've also included a short-hand pattern for matching records, as I think it might help when doing a lot of record patterns: `{$foo}` which is equivalent to `{foo: $foo}`. There are still missing components, so consider this the first step in full pattern matching support. Currently missing: * Patterns for strings * Or-patterns (like the `|` in Rust) * Patterns for tables (unclear how we want to match a table, so it'll need some design) * Patterns for binary values * And much more # User-Facing Changes [see above] # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-03-24 02:52:01 +01:00
position += 1;
if position >= output.len() {
working_set.error(ParseError::Mismatch(
Add pattern matching (#8590) # Description This adds `match` and basic pattern matching. An example: ``` match $x { 1..10 => { print "Value is between 1 and 10" } { foo: $bar } => { print $"Value has a 'foo' field with value ($bar)" } [$a, $b] => { print $"Value is a list with two items: ($a) and ($b)" } _ => { print "Value is none of the above" } } ``` Like the recent changes to `if` to allow it to be used as an expression, `match` can also be used as an expression. This allows you to assign the result to a variable, eg) `let xyz = match ...` I've also included a short-hand pattern for matching records, as I think it might help when doing a lot of record patterns: `{$foo}` which is equivalent to `{foo: $foo}`. There are still missing components, so consider this the first step in full pattern matching support. Currently missing: * Patterns for strings * Or-patterns (like the `|` in Rust) * Patterns for tables (unclear how we want to match a table, so it'll need some design) * Patterns for binary values * And much more # User-Facing Changes [see above] # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-03-24 02:52:01 +01:00
"=>".into(),
"end of input".into(),
Span::new(output[position - 1].span.end, output[position - 1].span.end),
));
Add pattern matching (#8590) # Description This adds `match` and basic pattern matching. An example: ``` match $x { 1..10 => { print "Value is between 1 and 10" } { foo: $bar } => { print $"Value has a 'foo' field with value ($bar)" } [$a, $b] => { print $"Value is a list with two items: ($a) and ($b)" } _ => { print "Value is none of the above" } } ``` Like the recent changes to `if` to allow it to be used as an expression, `match` can also be used as an expression. This allows you to assign the result to a variable, eg) `let xyz = match ...` I've also included a short-hand pattern for matching records, as I think it might help when doing a lot of record patterns: `{$foo}` which is equivalent to `{foo: $foo}`. There are still missing components, so consider this the first step in full pattern matching support. Currently missing: * Patterns for strings * Or-patterns (like the `|` in Rust) * Patterns for tables (unclear how we want to match a table, so it'll need some design) * Patterns for binary values * And much more # User-Facing Changes [see above] # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-03-24 02:52:01 +01:00
working_set.exit_scope();
break;
}
let mut connector = working_set.get_span_contents(output[position].span);
// Multiple patterns connected by '|'
if connector == b"|" && position < output.len() {
let mut or_pattern = vec![pattern];
while connector == b"|" && position < output.len() {
connector = b"";
position += 1;
if position >= output.len() {
working_set.error(ParseError::Mismatch(
"pattern".into(),
"end of input".into(),
Span::new(output[position - 1].span.end, output[position - 1].span.end),
));
working_set.exit_scope();
break;
}
let pattern = parse_pattern(working_set, output[position].span);
or_pattern.push(pattern);
position += 1;
if position >= output.len() {
working_set.error(ParseError::Mismatch(
"=>".into(),
"end of input".into(),
Span::new(output[position - 1].span.end, output[position - 1].span.end),
));
working_set.exit_scope();
break;
} else {
connector = working_set.get_span_contents(output[position].span);
}
}
let start = or_pattern
.first()
.expect("internal error: unexpected state of or-pattern")
.span
.start;
let end = or_pattern
.last()
.expect("internal error: unexpected state of or-pattern")
.span
.end;
pattern = MatchPattern {
pattern: Pattern::Or(or_pattern),
guard: None,
span: Span::new(start, end),
}
// A match guard
} else if connector == b"if" {
let if_end = {
let end = output[position].span.end;
Span::new(end, end)
};
position += 1;
let mk_err = || ParseError::LabeledErrorWithHelp {
error: "Match guard without an expression".into(),
label: "expected an expression".into(),
help: "The `if` keyword must be followed with an expression".into(),
span: if_end,
};
if output.get(position).is_none() {
working_set.error(mk_err());
return garbage(working_set, span);
};
let (tokens, found) = if let Some((pos, _)) = output[position..]
.iter()
.find_position(|t| working_set.get_span_contents(t.span) == b"=>")
{
if position + pos == position {
working_set.error(mk_err());
return garbage(working_set, span);
}
(&output[position..position + pos], true)
} else {
(&output[position..], false)
};
let mut start = 0;
let guard = parse_multispan_value(
working_set,
&tokens.iter().map(|tok| tok.span).collect_vec(),
&mut start,
&SyntaxShape::MathExpression,
);
pattern.guard = Some(guard);
position += if found { start + 1 } else { start };
connector = working_set.get_span_contents(output[position].span);
}
// Then the `=>` arrow
if connector != b"=>" {
working_set.error(ParseError::Mismatch(
Add pattern matching (#8590) # Description This adds `match` and basic pattern matching. An example: ``` match $x { 1..10 => { print "Value is between 1 and 10" } { foo: $bar } => { print $"Value has a 'foo' field with value ($bar)" } [$a, $b] => { print $"Value is a list with two items: ($a) and ($b)" } _ => { print "Value is none of the above" } } ``` Like the recent changes to `if` to allow it to be used as an expression, `match` can also be used as an expression. This allows you to assign the result to a variable, eg) `let xyz = match ...` I've also included a short-hand pattern for matching records, as I think it might help when doing a lot of record patterns: `{$foo}` which is equivalent to `{foo: $foo}`. There are still missing components, so consider this the first step in full pattern matching support. Currently missing: * Patterns for strings * Or-patterns (like the `|` in Rust) * Patterns for tables (unclear how we want to match a table, so it'll need some design) * Patterns for binary values * And much more # User-Facing Changes [see above] # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-03-24 02:52:01 +01:00
"=>".into(),
"end of input".into(),
Span::new(output[position - 1].span.end, output[position - 1].span.end),
));
} else {
position += 1;
Add pattern matching (#8590) # Description This adds `match` and basic pattern matching. An example: ``` match $x { 1..10 => { print "Value is between 1 and 10" } { foo: $bar } => { print $"Value has a 'foo' field with value ($bar)" } [$a, $b] => { print $"Value is a list with two items: ($a) and ($b)" } _ => { print "Value is none of the above" } } ``` Like the recent changes to `if` to allow it to be used as an expression, `match` can also be used as an expression. This allows you to assign the result to a variable, eg) `let xyz = match ...` I've also included a short-hand pattern for matching records, as I think it might help when doing a lot of record patterns: `{$foo}` which is equivalent to `{foo: $foo}`. There are still missing components, so consider this the first step in full pattern matching support. Currently missing: * Patterns for strings * Or-patterns (like the `|` in Rust) * Patterns for tables (unclear how we want to match a table, so it'll need some design) * Patterns for binary values * And much more # User-Facing Changes [see above] # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-03-24 02:52:01 +01:00
}
// Finally, the value/expression/block that we will run to produce the result
if position >= output.len() {
working_set.error(ParseError::Mismatch(
Add pattern matching (#8590) # Description This adds `match` and basic pattern matching. An example: ``` match $x { 1..10 => { print "Value is between 1 and 10" } { foo: $bar } => { print $"Value has a 'foo' field with value ($bar)" } [$a, $b] => { print $"Value is a list with two items: ($a) and ($b)" } _ => { print "Value is none of the above" } } ``` Like the recent changes to `if` to allow it to be used as an expression, `match` can also be used as an expression. This allows you to assign the result to a variable, eg) `let xyz = match ...` I've also included a short-hand pattern for matching records, as I think it might help when doing a lot of record patterns: `{$foo}` which is equivalent to `{foo: $foo}`. There are still missing components, so consider this the first step in full pattern matching support. Currently missing: * Patterns for strings * Or-patterns (like the `|` in Rust) * Patterns for tables (unclear how we want to match a table, so it'll need some design) * Patterns for binary values * And much more # User-Facing Changes [see above] # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-03-24 02:52:01 +01:00
"match result".into(),
"end of input".into(),
Span::new(output[position - 1].span.end, output[position - 1].span.end),
));
Add pattern matching (#8590) # Description This adds `match` and basic pattern matching. An example: ``` match $x { 1..10 => { print "Value is between 1 and 10" } { foo: $bar } => { print $"Value has a 'foo' field with value ($bar)" } [$a, $b] => { print $"Value is a list with two items: ($a) and ($b)" } _ => { print "Value is none of the above" } } ``` Like the recent changes to `if` to allow it to be used as an expression, `match` can also be used as an expression. This allows you to assign the result to a variable, eg) `let xyz = match ...` I've also included a short-hand pattern for matching records, as I think it might help when doing a lot of record patterns: `{$foo}` which is equivalent to `{foo: $foo}`. There are still missing components, so consider this the first step in full pattern matching support. Currently missing: * Patterns for strings * Or-patterns (like the `|` in Rust) * Patterns for tables (unclear how we want to match a table, so it'll need some design) * Patterns for binary values * And much more # User-Facing Changes [see above] # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-03-24 02:52:01 +01:00
working_set.exit_scope();
break;
}
let result = parse_multispan_value(
Add pattern matching (#8590) # Description This adds `match` and basic pattern matching. An example: ``` match $x { 1..10 => { print "Value is between 1 and 10" } { foo: $bar } => { print $"Value has a 'foo' field with value ($bar)" } [$a, $b] => { print $"Value is a list with two items: ($a) and ($b)" } _ => { print "Value is none of the above" } } ``` Like the recent changes to `if` to allow it to be used as an expression, `match` can also be used as an expression. This allows you to assign the result to a variable, eg) `let xyz = match ...` I've also included a short-hand pattern for matching records, as I think it might help when doing a lot of record patterns: `{$foo}` which is equivalent to `{foo: $foo}`. There are still missing components, so consider this the first step in full pattern matching support. Currently missing: * Patterns for strings * Or-patterns (like the `|` in Rust) * Patterns for tables (unclear how we want to match a table, so it'll need some design) * Patterns for binary values * And much more # User-Facing Changes [see above] # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-03-24 02:52:01 +01:00
working_set,
&[output[position].span],
&mut 0,
&SyntaxShape::OneOf(vec![SyntaxShape::Block, SyntaxShape::Expression]),
);
position += 1;
working_set.exit_scope();
output_matches.push((pattern, result));
}
Expression::new(
working_set,
Expr::MatchBlock(output_matches),
span,
Type::Any,
)
Add pattern matching (#8590) # Description This adds `match` and basic pattern matching. An example: ``` match $x { 1..10 => { print "Value is between 1 and 10" } { foo: $bar } => { print $"Value has a 'foo' field with value ($bar)" } [$a, $b] => { print $"Value is a list with two items: ($a) and ($b)" } _ => { print "Value is none of the above" } } ``` Like the recent changes to `if` to allow it to be used as an expression, `match` can also be used as an expression. This allows you to assign the result to a variable, eg) `let xyz = match ...` I've also included a short-hand pattern for matching records, as I think it might help when doing a lot of record patterns: `{$foo}` which is equivalent to `{foo: $foo}`. There are still missing components, so consider this the first step in full pattern matching support. Currently missing: * Patterns for strings * Or-patterns (like the `|` in Rust) * Patterns for tables (unclear how we want to match a table, so it'll need some design) * Patterns for binary values * And much more # User-Facing Changes [see above] # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-03-24 02:52:01 +01:00
}
pub fn parse_closure_expression(
working_set: &mut StateWorkingSet,
shape: &SyntaxShape,
span: Span,
) -> Expression {
trace!("parsing: closure expression");
let bytes = working_set.get_span_contents(span);
let mut start = span.start;
let mut end = span.end;
if bytes.starts_with(b"{") {
start += 1;
} else {
working_set.error(ParseError::Expected("closure", span));
return garbage(working_set, span);
}
if bytes.ends_with(b"}") {
end -= 1;
} else {
working_set.error(ParseError::Unclosed("}".into(), Span::new(end, end)));
}
Protocol: debug_assert!() Span to reflect a valid slice (#6806) Also enforce this by #[non_exhaustive] span such that going forward we cannot, in debug builds (1), construct invalid spans. The motivation for this stems from #6431 where I've seen crashes due to invalid slice indexing. My hope is this will mitigate such senarios 1. https://github.com/nushell/nushell/pull/6431#issuecomment-1278147241 # Description (description of your pull request here) # Tests Make sure you've done the following: - [ ] Add tests that cover your changes, either in the command examples, the crate/tests folder, or in the /tests folder. - [ ] Try to think about corner cases and various ways how your changes could break. Cover them with tests. - [ ] If adding tests is not possible, please document in the PR body a minimal example with steps on how to reproduce so one can verify your change works. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [ ] `cargo clippy --workspace --features=extra -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [ ] `cargo test --workspace --features=extra` to check that all the tests pass # Documentation - [ ] If your PR touches a user-facing nushell feature then make sure that there is an entry in the documentation (https://github.com/nushell/nushell.github.io) for the feature, and update it if necessary.
2022-12-03 10:44:12 +01:00
let inner_span = Span::new(start, end);
let source = working_set.get_span_contents(inner_span);
let (output, err) = lex(source, start, &[], &[], false);
if let Some(err) = err {
working_set.error(err);
}
working_set.enter_scope();
2021-09-02 10:25:22 +02:00
// Check to see if we have parameters
let (signature, amt_to_skip): (Option<(Box<Signature>, Span)>, usize) = match output.first() {
2021-09-02 10:25:22 +02:00
Some(Token {
contents: TokenContents::Pipe,
2021-09-06 01:16:27 +02:00
span,
}) => {
// We've found a parameter list
let start_point = span.start;
let mut token_iter = output.iter().enumerate().skip(1);
let mut end_span = None;
let mut amt_to_skip = 1;
for token in &mut token_iter {
if let Token {
2021-09-02 10:25:22 +02:00
contents: TokenContents::Pipe,
2021-09-06 01:16:27 +02:00
span,
} = token.1
{
end_span = Some(span);
amt_to_skip = token.0;
break;
2021-08-25 21:29:36 +02:00
}
}
2021-09-06 01:16:27 +02:00
let end_point = if let Some(span) = end_span {
span.end
} else {
end
};
Protocol: debug_assert!() Span to reflect a valid slice (#6806) Also enforce this by #[non_exhaustive] span such that going forward we cannot, in debug builds (1), construct invalid spans. The motivation for this stems from #6431 where I've seen crashes due to invalid slice indexing. My hope is this will mitigate such senarios 1. https://github.com/nushell/nushell/pull/6431#issuecomment-1278147241 # Description (description of your pull request here) # Tests Make sure you've done the following: - [ ] Add tests that cover your changes, either in the command examples, the crate/tests folder, or in the /tests folder. - [ ] Try to think about corner cases and various ways how your changes could break. Cover them with tests. - [ ] If adding tests is not possible, please document in the PR body a minimal example with steps on how to reproduce so one can verify your change works. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [ ] `cargo clippy --workspace --features=extra -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [ ] `cargo test --workspace --features=extra` to check that all the tests pass # Documentation - [ ] If your PR touches a user-facing nushell feature then make sure that there is an entry in the documentation (https://github.com/nushell/nushell.github.io) for the feature, and update it if necessary.
2022-12-03 10:44:12 +01:00
let signature_span = Span::new(start_point, end_point);
let signature = parse_signature_helper(working_set, signature_span);
2021-09-06 01:16:27 +02:00
(Some((signature, signature_span)), amt_to_skip)
2021-09-02 10:25:22 +02:00
}
Some(Token {
Better errors when bash-like operators are used (#7241) # Description Adds improved errors for when a user uses a bashism that nu doesn't support. fixes #7237 Examples: ``` Error: nu::parser::shell_andand (link) × The '&&' operator is not supported in Nushell ╭─[entry #1:1:1] 1 │ ls && ls · ─┬ · ╰── instead of '&&', use ';' or 'and' ╰──── help: use ';' instead of the shell '&&', or 'and' instead of the boolean '&&' ``` ``` Error: nu::parser::shell_oror (link) × The '||' operator is not supported in Nushell ╭─[entry #8:1:1] 1 │ ls || ls · ─┬ · ╰── instead of '||', use 'try' or 'or' ╰──── help: use 'try' instead of the shell '||', or 'or' instead of the boolean '||' ``` ``` Error: nu::parser::shell_err (link) × The '2>' shell operation is 'err>' in Nushell. ╭─[entry #9:1:1] 1 │ foo 2> bar.txt · ─┬ · ╰── use 'err>' instead of '2>' in Nushell ╰──── ``` ``` Error: nu::parser::shell_outerr (link) × The '2>&1' shell operation is 'out+err>' in Nushell. ╭─[entry #10:1:1] 1 │ foo 2>&1 bar.txt · ──┬─ · ╰── use 'out+err>' instead of '2>&1' in Nushell ╰──── help: Nushell redirection will write all of stdout before stderr. ``` # User-Facing Changes **BREAKING CHANGES** This removes the `&&` and `||` operators. We previously supported by `&&`/`and` and `||`/`or`. With this change, only `and` and `or` are valid boolean operators. # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2022-12-08 00:02:11 +01:00
contents: TokenContents::PipePipe,
span,
Better errors when bash-like operators are used (#7241) # Description Adds improved errors for when a user uses a bashism that nu doesn't support. fixes #7237 Examples: ``` Error: nu::parser::shell_andand (link) × The '&&' operator is not supported in Nushell ╭─[entry #1:1:1] 1 │ ls && ls · ─┬ · ╰── instead of '&&', use ';' or 'and' ╰──── help: use ';' instead of the shell '&&', or 'and' instead of the boolean '&&' ``` ``` Error: nu::parser::shell_oror (link) × The '||' operator is not supported in Nushell ╭─[entry #8:1:1] 1 │ ls || ls · ─┬ · ╰── instead of '||', use 'try' or 'or' ╰──── help: use 'try' instead of the shell '||', or 'or' instead of the boolean '||' ``` ``` Error: nu::parser::shell_err (link) × The '2>' shell operation is 'err>' in Nushell. ╭─[entry #9:1:1] 1 │ foo 2> bar.txt · ─┬ · ╰── use 'err>' instead of '2>' in Nushell ╰──── ``` ``` Error: nu::parser::shell_outerr (link) × The '2>&1' shell operation is 'out+err>' in Nushell. ╭─[entry #10:1:1] 1 │ foo 2>&1 bar.txt · ──┬─ · ╰── use 'out+err>' instead of '2>&1' in Nushell ╰──── help: Nushell redirection will write all of stdout before stderr. ``` # User-Facing Changes **BREAKING CHANGES** This removes the `&&` and `||` operators. We previously supported by `&&`/`and` and `||`/`or`. With this change, only `and` and `or` are valid boolean operators. # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2022-12-08 00:02:11 +01:00
}) => (
Some((Box::new(Signature::new("closure".to_string())), *span)),
1,
),
_ => (None, 0),
2021-09-02 10:25:22 +02:00
};
2021-08-25 21:29:36 +02:00
// TODO: Finish this
if let SyntaxShape::Closure(Some(v)) = shape {
if let Some((sig, sig_span)) = &signature {
if sig.num_positionals() > v.len() {
working_set.error(ParseError::ExpectedWithStringMsg(
format!(
"{} closure parameter{}",
v.len(),
if v.len() > 1 { "s" } else { "" }
),
*sig_span,
));
}
for (expected, PositionalArg { name, shape, .. }) in
v.iter().zip(sig.required_positional.iter())
{
if expected != shape && *shape != SyntaxShape::Any {
working_set.error(ParseError::ParameterMismatchType(
name.to_owned(),
expected.to_string(),
shape.to_string(),
*sig_span,
));
}
}
}
}
Reuse the cached parse results of parsed files (#8949) # Description This does a lookup in the cache of parsed files to see if a span can be found for a file that was previously loaded with the same contents, then uses that span to find the parsed block for that file. The end result should, in theory, be identical but doesn't require any reparsing or creating new blocks/new definitions that aren't needed. This drops the sg.nu benchmark from: ``` ╭───┬───────────────────╮ │ 0 │ 280ms 606µs 208ns │ │ 1 │ 282ms 654µs 416ns │ │ 2 │ 252ms 640µs 541ns │ │ 3 │ 250ms 940µs 41ns │ │ 4 │ 241ms 216µs 375ns │ │ 5 │ 257ms 310µs 583ns │ │ 6 │ 196ms 739µs 416ns │ ╰───┴───────────────────╯ ``` to: ``` ╭───┬───────────────────╮ │ 0 │ 118ms 698µs 125ns │ │ 1 │ 121ms 327µs │ │ 2 │ 121ms 873µs 500ns │ │ 3 │ 124ms 94µs 708ns │ │ 4 │ 113ms 733µs 291ns │ │ 5 │ 108ms 663µs 125ns │ │ 6 │ 63ms 482µs 625ns │ ╰───┴───────────────────╯ ``` I was hoping to also see some startup time improvements, but I didn't notice much there. # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-04-21 21:00:33 +02:00
let mut output = parse_block(working_set, &output[amt_to_skip..], span, false, false);
2021-07-02 09:32:30 +02:00
2021-09-06 01:16:27 +02:00
if let Some(signature) = signature {
output.signature = signature.0;
2021-09-06 01:16:27 +02:00
}
output.span = Some(span);
2021-10-25 22:04:23 +02:00
2021-09-06 04:20:02 +02:00
working_set.exit_scope();
Make EngineState clone cheaper with Arc on all of the heavy objects (#12229) # Description This makes many of the larger objects in `EngineState` into `Arc`, and uses `Arc::make_mut` to do clone-on-write if the reference is not unique. This is generally very cheap, giving us the best of both worlds - allowing us to mutate without cloning if we have an exclusive reference, and cloning if we don't. This started as more of a curiosity for me after remembering that `Arc::make_mut` exists and can make using `Arc` for mostly immutable data that sometimes needs to be changed very convenient, and also after hearing someone complain about memory usage on Discord - this is a somewhat significant win for that. The exact objects that were wrapped in `Arc`: - `files`, `file_contents` - the strings and byte buffers - `decls` - the whole `Vec`, but mostly to avoid lots of individual `malloc()` calls on Clone rather than for memory usage - `blocks` - the blocks themselves, rather than the outer Vec - `modules` - the modules themselves, rather than the outer Vec - `env_vars`, `previous_env_vars` - the entire maps - `config` The changes required were relatively minimal, but this is a breaking API change. In particular, blocks are added as Arcs, to allow the parser cache functionality to work. With my normal nu config, running on Linux, this saves me about 15 MiB of process memory usage when running interactively (65 MiB → 50 MiB). This also makes quick command executions cheaper, particularly since every REPL loop now involves a clone of the engine state so that we can recover from a panic. It also reduces memory usage where engine state needs to be cloned and sent to another thread or kept within an iterator. # User-Facing Changes Shouldn't be any, since it's all internal stuff, but it does change some public interfaces so it's a breaking change
2024-03-19 19:07:00 +01:00
let block_id = working_set.add_block(Arc::new(output));
2021-07-16 22:26:40 +02:00
Expression::new(working_set, Expr::Closure(block_id), span, Type::Closure)
2021-09-02 10:25:22 +02:00
}
2021-07-02 09:32:30 +02:00
2021-09-02 10:25:22 +02:00
pub fn parse_value(
working_set: &mut StateWorkingSet,
span: Span,
shape: &SyntaxShape,
) -> Expression {
trace!("parsing: value: {}", shape);
2021-09-02 10:25:22 +02:00
let bytes = working_set.get_span_contents(span);
if bytes.is_empty() {
working_set.error(ParseError::IncompleteParser(span));
return garbage(working_set, span);
}
// Check for reserved keyword values
match bytes {
b"true" => {
if matches!(shape, SyntaxShape::Boolean) || matches!(shape, SyntaxShape::Any) {
return Expression::new(working_set, Expr::Bool(true), span, Type::Bool);
} else {
working_set.error(ParseError::Expected("non-boolean value", span));
return Expression::garbage(working_set, span);
}
}
b"false" => {
if matches!(shape, SyntaxShape::Boolean) || matches!(shape, SyntaxShape::Any) {
return Expression::new(working_set, Expr::Bool(false), span, Type::Bool);
} else {
working_set.error(ParseError::Expected("non-boolean value", span));
return Expression::garbage(working_set, span);
}
}
b"null" => {
return Expression::new(working_set, Expr::Nothing, span, Type::Nothing);
}
Require that values that look like numbers parse as numberlike (#8635) # Description Require that any value that looks like it might be a number (starts with a digit, or a '-' + digit, or a '+' + digits, or a special form float like `-inf`, `inf`, or `NaN`) must now be treated as a number-like value. Number-like syntax can only parse into number-like values. Number-like values include: durations, ints, floats, ranges, filesizes, binary data, etc. # User-Facing Changes BREAKING CHANGE BREAKING CHANGE BREAKING CHANGE BREAKING CHANGE BREAKING CHANGE BREAKING CHANGE BREAKING CHANGE BREAKING CHANGE Just making sure we see this for release notes 😅 This breaks any and all numberlike values that were treated as strings before. Example, we used to allow `3,` as a bare word. Anything like this would now require quotes or backticks to be treated as a string or bare word, respectively. # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-03-28 08:31:38 +02:00
b"-inf" | b"inf" | b"NaN" => {
return parse_float(working_set, span);
Require that values that look like numbers parse as numberlike (#8635) # Description Require that any value that looks like it might be a number (starts with a digit, or a '-' + digit, or a '+' + digits, or a special form float like `-inf`, `inf`, or `NaN`) must now be treated as a number-like value. Number-like syntax can only parse into number-like values. Number-like values include: durations, ints, floats, ranges, filesizes, binary data, etc. # User-Facing Changes BREAKING CHANGE BREAKING CHANGE BREAKING CHANGE BREAKING CHANGE BREAKING CHANGE BREAKING CHANGE BREAKING CHANGE BREAKING CHANGE Just making sure we see this for release notes 😅 This breaks any and all numberlike values that were treated as strings before. Example, we used to allow `3,` as a bare word. Anything like this would now require quotes or backticks to be treated as a string or bare word, respectively. # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-03-28 08:31:38 +02:00
}
_ => {}
}
match bytes[0] {
b'$' => return parse_dollar_expr(working_set, span),
b'(' => return parse_paren_expr(working_set, span, shape),
b'{' => return parse_brace_expr(working_set, span, shape),
b'[' => match shape {
2021-09-02 10:25:22 +02:00
SyntaxShape::Any
| SyntaxShape::List(_)
| SyntaxShape::Table(_)
treat path contains '?' as pattern (#10142) Fix https://github.com/nushell/nushell/issues/10136 # Description Current nushell only handle path containing '*' as match pattern and treat '?' as just normal path. This pr makes path containing '?' is also processed as pattern. 🔴 **Concerns: Need to design/comfirm a consistent rule to handle dirs/files with '?' in their names.** Currently: - if no dir has exactly same name with pattern, it will print the list of matched directories - if pattern exactly matches an empty dir's name, it will just print the empty dir's content ( i.e. `[]`) - if pattern exactly matches an dir's name, it will perform pattern match and print all the dir contains e.g. ```bash mkdir src ls s?c ``` | name | type | size | modified | | ---- | ---- | ------ | --------------------------------------------- | | src | dir | 1.1 KB | Tue, 29 Aug 2023 07:39:41 +0900 (9 hours ago) | ----------- ```bash mkdir src mkdir scc mkdir scs ls s?c ``` | name | type | size | modified | | ---- | ---- | ------ | ------------------------------------------------ | | scc | dir | 64 B | Tue, 29 Aug 2023 16:55:31 +0900 (14 seconds ago) | | src | dir | 1.1 KB | Tue, 29 Aug 2023 07:39:41 +0900 (9 hours ago) | ----------- ```bash mkdir s?c ls s?c ``` print empty (i.e. ls of dir `s?c`) ----------- ```bash mkdir -p s?c/test ls s?c ``` |name|type|size|modified| |-|-|-|-| |s?c/test|dir|64 B|Tue, 29 Aug 2023 16:47:53 +0900 (2 minutes ago)| |src/bytes|dir|480 B|Fri, 25 Aug 2023 17:43:52 +0900 (3 days ago)| |src/charting|dir|160 B|Fri, 25 Aug 2023 17:43:52 +0900 (3 days ago)| |src/conversions|dir|160 B|Fri, 25 Aug 2023 17:43:52 +0900 (3 days ago)| ----------- # User-Facing Changes User will be able to use '?' to match directory/file. # Tests + Formatting - :green_circle: `toolkit fmt` - :green_circle: `toolkit clippy` - :green_circle: `toolkit test` - :green_circle: `toolkit test stdlib` # After Submitting None --------- Co-authored-by: Horasal <horsal@horsal.dev>
2023-09-04 02:25:00 +02:00
| SyntaxShape::Signature
| SyntaxShape::Filepath
do not attempt to glob expand if the file path is wrapped in quotes (#11569) # Description Fixes: #11455 ### For arguments which is annotated with `:path/:directory/:glob` To fix the issue, we need to have a way to know if a path is originally quoted during runtime. So the information needed to be added at several levels: * parse time (from user input to expression) We need to add quoted information into `Expr::Filepath`, `Expr::Directory`, `Expr::GlobPattern` * eval time When convert from `Expr::Filepath`, `Expr::Directory`, `Expr::GlobPattern` to `Value::String` during runtime, we won't auto expanded the path if it's quoted ### For `ls` It's really special, because it accepts a `String` as a pattern, and it generates `glob` expression inside the command itself. So the idea behind the change is introducing a special SyntaxShape to ls: `SyntaxShape::LsGlobPattern`. So we can track if the pattern is originally quoted easier, and we don't auto expand the path either. Then when constructing a glob pattern inside ls, we check if input pattern is quoted, if so: we escape the input pattern, so we can run `ls a[123]b`, because it's already escaped. Finally, to accomplish the checking process, we also need to introduce a new value type called `Value::QuotedString` to differ from `Value::String`, it's used to generate an enum called `NuPath`, which is finally used in `ls` function. `ls` learned from `NuPath` to know if user input is quoted. # User-Facing Changes Actually it contains several changes ### For arguments which is annotated with `:path/:directory/:glob` #### Before ```nushell > def foo [p: path] { echo $p }; print (foo "~/a"); print (foo '~/a') /home/windsoilder/a /home/windsoilder/a > def foo [p: directory] { echo $p }; print (foo "~/a"); print (foo '~/a') /home/windsoilder/a /home/windsoilder/a > def foo [p: glob] { echo $p }; print (foo "~/a"); print (foo '~/a') /home/windsoilder/a /home/windsoilder/a ``` #### After ```nushell > def foo [p: path] { echo $p }; print (foo "~/a"); print (foo '~/a') ~/a ~/a > def foo [p: directory] { echo $p }; print (foo "~/a"); print (foo '~/a') ~/a ~/a > def foo [p: glob] { echo $p }; print (foo "~/a"); print (foo '~/a') ~/a ~/a ``` ### For ls command `touch '[uwu]'` #### Before ``` ❯ ls -D "[uwu]" Error: × No matches found for [uwu] ╭─[entry #6:1:1] 1 │ ls -D "[uwu]" · ───┬─── · ╰── Pattern, file or folder not found ╰──── help: no matches found ``` #### After ``` ❯ ls -D "[uwu]" ╭───┬───────┬──────┬──────┬──────────╮ │ # │ name │ type │ size │ modified │ ├───┼───────┼──────┼──────┼──────────┤ │ 0 │ [uwu] │ file │ 0 B │ now │ ╰───┴───────┴──────┴──────┴──────────╯ ``` # Tests + Formatting Done # After Submitting NaN
2024-01-21 16:22:25 +01:00
| SyntaxShape::String
Unify glob behavior on `open`, `rm`, `cp-old`, `mv`, `umv`, `cp` and `du` commands (#11621) # Description This pr is a follow up to [#11569](https://github.com/nushell/nushell/pull/11569#issuecomment-1902279587) > Revert the logic in https://github.com/nushell/nushell/pull/10694 and apply the logic in this pr to mv, cp, rv will require a larger change, I need to think how to achieve the bahavior And sorry @bobhy for reverting some of your changes. This pr is going to unify glob behavior on the given commands: * open * rm * cp-old * mv * umv * cp * du So they have the same behavior to `ls`, which is: If given parameter is quoted by single quote(`'`) or double quote(`"`), don't auto-expand the glob pattern. If not quoted, auto-expand the glob pattern. Fixes: #9558 Fixes: #10211 Fixes: #9310 Fixes: #10364 # TODO But there is one thing remains: if we give a variable to the command, it will always auto-expand the glob pattern, e.g: ```nushell let path = "a[123]b" rm $path ``` I don't think it's expected. But I also think user might want to auto-expand the glob pattern in variables. So I'll introduce a new command called `glob escape`, then if user doesn't want to auto-expand the glob pattern, he can just do this: `rm ($path | glob escape)` # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting Done # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> ## NOTE This pr changes the semantic of `GlobPattern`, before this pr, it will `expand path` after evaluated, this makes `nu_engine::glob_from` have no chance to glob things right if a path contains glob pattern. e.g: [#9310 ](https://github.com/nushell/nushell/issues/9310#issuecomment-1886824030) #10211 I think changing the semantic is fine, because it makes glob works if path contains something like '*'. It maybe a breaking change if a custom command's argument are annotated by `: glob`.
2024-01-26 14:57:35 +01:00
| SyntaxShape::GlobPattern => {}
2021-09-02 10:25:22 +02:00
_ => {
working_set.error(ParseError::Expected("non-[] value", span));
return Expression::garbage(working_set, span);
2021-07-08 23:45:56 +02:00
}
},
add raw-string literal support (#9956) # Description This PR adds raw string support by using `r#` at the beginning of single quoted strings and `#` at the end. Notice that escapes do not process, even within single quotes, parentheses don't mean anything, $variables don't mean anything. It's just a string. ```nushell ❯ echo r#'one\ntwo (blah) ($var)'# one\ntwo (blah) ($var) ``` Notice how they work without `echo` or `print` and how they work without carriage returns. ```nushell ❯ r#'adsfa'# adsfa ❯ r##"asdfa'@qpejq'## asdfa'@qpejq ❯ r#'asdfasdfasf ∙ foqwejfqo@'23rfjqf'# ``` They also have a special configurable color in the repl. (use single quotes though) ![image](https://github.com/nushell/nushell/assets/343840/8780e21d-de4c-45b3-9880-2425f5fe10ef) They should work like rust raw literals and allow `r##`, `r###`, `r####`, etc, to help with having one or many `#`'s in the middle of your raw-string. They should work with `let` as well. ```nushell r#'some\nraw\nstring'# | str upcase ``` closes https://github.com/nushell/nushell/issues/5091 # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: WindSoilder <WindSoilder@outlook.com> Co-authored-by: Ian Manske <ian.manske@pm.me>
2024-05-02 15:36:37 +02:00
b'r' if bytes.len() > 1 && bytes[1] == b'#' => {
return parse_raw_string(working_set, span);
}
_ => {}
2021-09-02 10:25:22 +02:00
}
2021-07-01 03:31:02 +02:00
2021-09-02 10:25:22 +02:00
match shape {
SyntaxShape::CompleterWrapper(shape, custom_completion) => {
let mut expression = parse_value(working_set, span, shape);
expression.custom_completion = Some(*custom_completion);
expression
}
SyntaxShape::Number => parse_number(working_set, span),
SyntaxShape::Float => parse_float(working_set, span),
SyntaxShape::Int => parse_int(working_set, span),
SyntaxShape::Duration => parse_duration(working_set, span),
SyntaxShape::DateTime => parse_datetime(working_set, span),
SyntaxShape::Filesize => parse_filesize(working_set, span),
SyntaxShape::Range => parse_range(working_set, span),
2021-10-04 21:21:31 +02:00
SyntaxShape::Filepath => parse_filepath(working_set, span),
SyntaxShape::Directory => parse_directory(working_set, span),
2021-10-04 21:21:31 +02:00
SyntaxShape::GlobPattern => parse_glob_pattern(working_set, span),
SyntaxShape::String => parse_string(working_set, span),
2022-03-01 00:31:53 +01:00
SyntaxShape::Binary => parse_binary(working_set, span),
2021-09-02 10:25:22 +02:00
SyntaxShape::Signature => {
if bytes.starts_with(b"[") {
parse_signature(working_set, span)
2021-09-02 10:25:22 +02:00
} else {
working_set.error(ParseError::Expected("signature", span));
Expression::garbage(working_set, span)
2021-07-09 08:23:20 +02:00
}
2021-09-02 10:25:22 +02:00
}
SyntaxShape::List(elem) => {
if bytes.starts_with(b"[") {
parse_list_expression(working_set, span, elem)
2021-09-02 10:25:22 +02:00
} else {
working_set.error(ParseError::Expected("list", span));
Expression::garbage(working_set, span)
2021-07-08 09:49:17 +02:00
}
2021-09-02 10:25:22 +02:00
}
SyntaxShape::Table(_) => {
2021-09-02 10:25:22 +02:00
if bytes.starts_with(b"[") {
parse_table_expression(working_set, span)
2021-09-02 10:25:22 +02:00
} else {
working_set.error(ParseError::Expected("table", span));
Expression::garbage(working_set, span)
2021-07-02 08:44:37 +02:00
}
2021-07-01 02:01:04 +02:00
}
SyntaxShape::CellPath => parse_simple_cell_path(working_set, span),
2021-10-12 06:49:17 +02:00
SyntaxShape::Boolean => {
// Redundant, though we catch bad boolean parses here
if bytes == b"true" || bytes == b"false" {
Expression::new(working_set, Expr::Bool(true), span, Type::Bool)
2021-10-12 06:49:17 +02:00
} else {
working_set.error(ParseError::Expected("bool", span));
Expression::garbage(working_set, span)
2021-10-12 06:49:17 +02:00
}
}
Syntax errors for string and int (#7952) # Description Added a few syntax errors in ints and strings, changed parser to stop and show that error rather than continue trying to parse those tokens as some other shape. However, I don't see how to push this direction much further, and most of the classic confusing errors can't be changed. Flagged as WIP for the moment, but passes all checks and works better than current release: 1. I have yet to figure out how to make these errors refer back to the book, as I see some other errors do. 2. How to give syntax error when malformed int is first token in line? Currently parsed as external command, user gets confusing error message. 3. Would like to be more strict with *decimal* int literals (lacking, e.g, `0x' prefix). Need to tinker more with the order of parse shape calls, currently, float is tried after int, so '1.4' has to be passed. _(Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience.)_ ```bash 〉"\z" Error: ╭─[entry #3:1:1] 1 │ "\z" · ─┬─ · ╰── Syntax error in string, unrecognized character after escape '\'. ╰──── ``` Canonic presentation of a syntax error. ```bash 〉" \u{01ffbogus}" Error: × Invalid syntax ╭─[entry #2:1:1] 1 │ " \u{01ffbogus}" · ───────┬────── · ╰── Syntax error in string, expecting 1 to 6 hex digits in unicode escape '\u{X...}', max value 10FFFF. ╰──── ``` Malformed unicode escape in string, flagged as error. String parse can be opinionated, it's the last shape tried. ```bash 〉0x22bogus Error: nu::shell::external_command (link) × External command failed ╭─[entry #4:1:1] 1 │ 0x22bogus · ────┬──── · ╰── executable was not found ╰──── help: No such file or directory (os error 2) ``` A *correct* number in first token would be evaluated, but an *incorrect* one is treated as external command? Confusing to users. ```bash 〉0 + 0x22bogus Error: × Invalid syntax ╭─[entry #5:1:1] 1 │ 0 + 0x22bogus · ────┬──── · ╰── Syntax error in int, invalid digits in radix 16 int. ╰──── ``` Can give syntax error if token is unambiguously int literal. e.g has 0b or 0x prefix, could not be a float. ```bash 〉0 + 098bogus Error: nu::parser::unsupported_operation (link) × Types mismatched for operation. ╭─[entry #6:1:1] 1 │ 0 + 098bogus · ┬ ┬ ────┬─── · │ │ ╰── string · │ ╰── doesn't support these values. · ╰── int ╰──── help: Change int or string to be the right types and try again. ``` But *decimal* literal (no prefix) can't be too strict. Parser is going to try float later. So '1.4' must be passed. # User-Facing Changes First and foremost, more specific error messages for typos in string and int literals. Probably improves interactive user experience. But a script that was causing and then checking for specific error might notice a different error message. _(List of all changes that impact the user experience here. This helps us keep track of breaking changes.)_ # Tests + Formatting Added (positive and negative unit tests in `cargo test -p nu-parser`. Didn't add integration tests. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [x] `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [x] `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --------- Co-authored-by: Stefan Holderbach <sholderbach@users.noreply.github.com>
2023-02-13 17:09:50 +01:00
// Be sure to return ParseError::Expected(..) if invoked for one of these shapes, but lex
// stream doesn't start with '{'} -- parsing in SyntaxShape::Any arm depends on this error variant.
allow records to have type annotations (#8914) # Description follow up to #8529 cleaned up version of #8892 - the original syntax is okay ```nu def okay [rec: record] {} ``` - you can now add type annotations for fields if you know them before hand ```nu def okay [rec: record<name: string>] {} ``` - you can specify multiple fields ```nu def okay [person: record<name: string age: int>] {} # an optional comma is allowed def okay [person: record<name: string, age: int>] {} ``` - if annotations are specified, any use of the command will be type checked against the specified type ```nu def unwrap [result: record<ok: bool, value: any>] {} unwrap {ok: 2, value: "value"} # errors with Error: nu::parser::type_mismatch × Type mismatch. ╭─[entry #4:1:1] 1 │ unwrap {ok: 2, value: "value"} · ───────┬───── · ╰── expected record<ok: bool, value: any>, found record<ok: int, value: string> ╰──── ``` > here the error is in the `ok` field, since `any` is coerced into any type > as a result `unwrap {ok: true, value: "value"}` is okay - the key must be a string, either quoted or unquoted ```nu def err [rec: record<{}: list>] {} # errors with Error: × `record` type annotations key not string ╭─[entry #7:1:1] 1 │ def unwrap [result: record<{}: bool, value: any>] {} · ─┬ · ╰── must be a string ╰──── ``` - a key doesn't have to have a type in which case it is assumed to be `any` ```nu def okay [person: record<name age>] {} def okay [person: record<name: string age>] {} ``` - however, if you put a colon, you have to specify a type ```nu def err [person: record<name: >] {} # errors with Error: nu::parser::parse_mismatch × Parse mismatch during operation. ╭─[entry #12:1:1] 1 │ def unwrap [res: record<name: >] { $res } · ┬ · ╰── expected type after colon ╰──── ``` # User-Facing Changes **[BREAKING CHANGES]** - this change adds a field to `SyntaxShape::Record` so any plugins that used it will have to update and include the field. though if you are unsure of the type the record expects, `SyntaxShape::Record(vec![])` will suffice
2023-04-26 15:16:55 +02:00
SyntaxShape::Block | SyntaxShape::Closure(..) | SyntaxShape::Record(_) => {
working_set.error(ParseError::Expected("block, closure or record", span));
Expression::garbage(working_set, span)
}
Syntax errors for string and int (#7952) # Description Added a few syntax errors in ints and strings, changed parser to stop and show that error rather than continue trying to parse those tokens as some other shape. However, I don't see how to push this direction much further, and most of the classic confusing errors can't be changed. Flagged as WIP for the moment, but passes all checks and works better than current release: 1. I have yet to figure out how to make these errors refer back to the book, as I see some other errors do. 2. How to give syntax error when malformed int is first token in line? Currently parsed as external command, user gets confusing error message. 3. Would like to be more strict with *decimal* int literals (lacking, e.g, `0x' prefix). Need to tinker more with the order of parse shape calls, currently, float is tried after int, so '1.4' has to be passed. _(Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience.)_ ```bash 〉"\z" Error: ╭─[entry #3:1:1] 1 │ "\z" · ─┬─ · ╰── Syntax error in string, unrecognized character after escape '\'. ╰──── ``` Canonic presentation of a syntax error. ```bash 〉" \u{01ffbogus}" Error: × Invalid syntax ╭─[entry #2:1:1] 1 │ " \u{01ffbogus}" · ───────┬────── · ╰── Syntax error in string, expecting 1 to 6 hex digits in unicode escape '\u{X...}', max value 10FFFF. ╰──── ``` Malformed unicode escape in string, flagged as error. String parse can be opinionated, it's the last shape tried. ```bash 〉0x22bogus Error: nu::shell::external_command (link) × External command failed ╭─[entry #4:1:1] 1 │ 0x22bogus · ────┬──── · ╰── executable was not found ╰──── help: No such file or directory (os error 2) ``` A *correct* number in first token would be evaluated, but an *incorrect* one is treated as external command? Confusing to users. ```bash 〉0 + 0x22bogus Error: × Invalid syntax ╭─[entry #5:1:1] 1 │ 0 + 0x22bogus · ────┬──── · ╰── Syntax error in int, invalid digits in radix 16 int. ╰──── ``` Can give syntax error if token is unambiguously int literal. e.g has 0b or 0x prefix, could not be a float. ```bash 〉0 + 098bogus Error: nu::parser::unsupported_operation (link) × Types mismatched for operation. ╭─[entry #6:1:1] 1 │ 0 + 098bogus · ┬ ┬ ────┬─── · │ │ ╰── string · │ ╰── doesn't support these values. · ╰── int ╰──── help: Change int or string to be the right types and try again. ``` But *decimal* literal (no prefix) can't be too strict. Parser is going to try float later. So '1.4' must be passed. # User-Facing Changes First and foremost, more specific error messages for typos in string and int literals. Probably improves interactive user experience. But a script that was causing and then checking for specific error might notice a different error message. _(List of all changes that impact the user experience here. This helps us keep track of breaking changes.)_ # Tests + Formatting Added (positive and negative unit tests in `cargo test -p nu-parser`. Didn't add integration tests. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [x] `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [x] `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --------- Co-authored-by: Stefan Holderbach <sholderbach@users.noreply.github.com>
2023-02-13 17:09:50 +01:00
2021-09-02 10:25:22 +02:00
SyntaxShape::Any => {
if bytes.starts_with(b"[") {
//parse_value(working_set, span, &SyntaxShape::Table)
parse_full_cell_path(working_set, None, span)
} else {
Syntax errors for string and int (#7952) # Description Added a few syntax errors in ints and strings, changed parser to stop and show that error rather than continue trying to parse those tokens as some other shape. However, I don't see how to push this direction much further, and most of the classic confusing errors can't be changed. Flagged as WIP for the moment, but passes all checks and works better than current release: 1. I have yet to figure out how to make these errors refer back to the book, as I see some other errors do. 2. How to give syntax error when malformed int is first token in line? Currently parsed as external command, user gets confusing error message. 3. Would like to be more strict with *decimal* int literals (lacking, e.g, `0x' prefix). Need to tinker more with the order of parse shape calls, currently, float is tried after int, so '1.4' has to be passed. _(Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience.)_ ```bash 〉"\z" Error: ╭─[entry #3:1:1] 1 │ "\z" · ─┬─ · ╰── Syntax error in string, unrecognized character after escape '\'. ╰──── ``` Canonic presentation of a syntax error. ```bash 〉" \u{01ffbogus}" Error: × Invalid syntax ╭─[entry #2:1:1] 1 │ " \u{01ffbogus}" · ───────┬────── · ╰── Syntax error in string, expecting 1 to 6 hex digits in unicode escape '\u{X...}', max value 10FFFF. ╰──── ``` Malformed unicode escape in string, flagged as error. String parse can be opinionated, it's the last shape tried. ```bash 〉0x22bogus Error: nu::shell::external_command (link) × External command failed ╭─[entry #4:1:1] 1 │ 0x22bogus · ────┬──── · ╰── executable was not found ╰──── help: No such file or directory (os error 2) ``` A *correct* number in first token would be evaluated, but an *incorrect* one is treated as external command? Confusing to users. ```bash 〉0 + 0x22bogus Error: × Invalid syntax ╭─[entry #5:1:1] 1 │ 0 + 0x22bogus · ────┬──── · ╰── Syntax error in int, invalid digits in radix 16 int. ╰──── ``` Can give syntax error if token is unambiguously int literal. e.g has 0b or 0x prefix, could not be a float. ```bash 〉0 + 098bogus Error: nu::parser::unsupported_operation (link) × Types mismatched for operation. ╭─[entry #6:1:1] 1 │ 0 + 098bogus · ┬ ┬ ────┬─── · │ │ ╰── string · │ ╰── doesn't support these values. · ╰── int ╰──── help: Change int or string to be the right types and try again. ``` But *decimal* literal (no prefix) can't be too strict. Parser is going to try float later. So '1.4' must be passed. # User-Facing Changes First and foremost, more specific error messages for typos in string and int literals. Probably improves interactive user experience. But a script that was causing and then checking for specific error might notice a different error message. _(List of all changes that impact the user experience here. This helps us keep track of breaking changes.)_ # Tests + Formatting Added (positive and negative unit tests in `cargo test -p nu-parser`. Didn't add integration tests. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [x] `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [x] `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --------- Co-authored-by: Stefan Holderbach <sholderbach@users.noreply.github.com>
2023-02-13 17:09:50 +01:00
let shapes = [
SyntaxShape::Binary,
SyntaxShape::Filesize,
SyntaxShape::Duration,
Syntax errors for string and int (#7952) # Description Added a few syntax errors in ints and strings, changed parser to stop and show that error rather than continue trying to parse those tokens as some other shape. However, I don't see how to push this direction much further, and most of the classic confusing errors can't be changed. Flagged as WIP for the moment, but passes all checks and works better than current release: 1. I have yet to figure out how to make these errors refer back to the book, as I see some other errors do. 2. How to give syntax error when malformed int is first token in line? Currently parsed as external command, user gets confusing error message. 3. Would like to be more strict with *decimal* int literals (lacking, e.g, `0x' prefix). Need to tinker more with the order of parse shape calls, currently, float is tried after int, so '1.4' has to be passed. _(Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience.)_ ```bash 〉"\z" Error: ╭─[entry #3:1:1] 1 │ "\z" · ─┬─ · ╰── Syntax error in string, unrecognized character after escape '\'. ╰──── ``` Canonic presentation of a syntax error. ```bash 〉" \u{01ffbogus}" Error: × Invalid syntax ╭─[entry #2:1:1] 1 │ " \u{01ffbogus}" · ───────┬────── · ╰── Syntax error in string, expecting 1 to 6 hex digits in unicode escape '\u{X...}', max value 10FFFF. ╰──── ``` Malformed unicode escape in string, flagged as error. String parse can be opinionated, it's the last shape tried. ```bash 〉0x22bogus Error: nu::shell::external_command (link) × External command failed ╭─[entry #4:1:1] 1 │ 0x22bogus · ────┬──── · ╰── executable was not found ╰──── help: No such file or directory (os error 2) ``` A *correct* number in first token would be evaluated, but an *incorrect* one is treated as external command? Confusing to users. ```bash 〉0 + 0x22bogus Error: × Invalid syntax ╭─[entry #5:1:1] 1 │ 0 + 0x22bogus · ────┬──── · ╰── Syntax error in int, invalid digits in radix 16 int. ╰──── ``` Can give syntax error if token is unambiguously int literal. e.g has 0b or 0x prefix, could not be a float. ```bash 〉0 + 098bogus Error: nu::parser::unsupported_operation (link) × Types mismatched for operation. ╭─[entry #6:1:1] 1 │ 0 + 098bogus · ┬ ┬ ────┬─── · │ │ ╰── string · │ ╰── doesn't support these values. · ╰── int ╰──── help: Change int or string to be the right types and try again. ``` But *decimal* literal (no prefix) can't be too strict. Parser is going to try float later. So '1.4' must be passed. # User-Facing Changes First and foremost, more specific error messages for typos in string and int literals. Probably improves interactive user experience. But a script that was causing and then checking for specific error might notice a different error message. _(List of all changes that impact the user experience here. This helps us keep track of breaking changes.)_ # Tests + Formatting Added (positive and negative unit tests in `cargo test -p nu-parser`. Didn't add integration tests. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [x] `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [x] `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --------- Co-authored-by: Stefan Holderbach <sholderbach@users.noreply.github.com>
2023-02-13 17:09:50 +01:00
SyntaxShape::Range,
SyntaxShape::DateTime,
SyntaxShape::Int,
SyntaxShape::Number,
Syntax errors for string and int (#7952) # Description Added a few syntax errors in ints and strings, changed parser to stop and show that error rather than continue trying to parse those tokens as some other shape. However, I don't see how to push this direction much further, and most of the classic confusing errors can't be changed. Flagged as WIP for the moment, but passes all checks and works better than current release: 1. I have yet to figure out how to make these errors refer back to the book, as I see some other errors do. 2. How to give syntax error when malformed int is first token in line? Currently parsed as external command, user gets confusing error message. 3. Would like to be more strict with *decimal* int literals (lacking, e.g, `0x' prefix). Need to tinker more with the order of parse shape calls, currently, float is tried after int, so '1.4' has to be passed. _(Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience.)_ ```bash 〉"\z" Error: ╭─[entry #3:1:1] 1 │ "\z" · ─┬─ · ╰── Syntax error in string, unrecognized character after escape '\'. ╰──── ``` Canonic presentation of a syntax error. ```bash 〉" \u{01ffbogus}" Error: × Invalid syntax ╭─[entry #2:1:1] 1 │ " \u{01ffbogus}" · ───────┬────── · ╰── Syntax error in string, expecting 1 to 6 hex digits in unicode escape '\u{X...}', max value 10FFFF. ╰──── ``` Malformed unicode escape in string, flagged as error. String parse can be opinionated, it's the last shape tried. ```bash 〉0x22bogus Error: nu::shell::external_command (link) × External command failed ╭─[entry #4:1:1] 1 │ 0x22bogus · ────┬──── · ╰── executable was not found ╰──── help: No such file or directory (os error 2) ``` A *correct* number in first token would be evaluated, but an *incorrect* one is treated as external command? Confusing to users. ```bash 〉0 + 0x22bogus Error: × Invalid syntax ╭─[entry #5:1:1] 1 │ 0 + 0x22bogus · ────┬──── · ╰── Syntax error in int, invalid digits in radix 16 int. ╰──── ``` Can give syntax error if token is unambiguously int literal. e.g has 0b or 0x prefix, could not be a float. ```bash 〉0 + 098bogus Error: nu::parser::unsupported_operation (link) × Types mismatched for operation. ╭─[entry #6:1:1] 1 │ 0 + 098bogus · ┬ ┬ ────┬─── · │ │ ╰── string · │ ╰── doesn't support these values. · ╰── int ╰──── help: Change int or string to be the right types and try again. ``` But *decimal* literal (no prefix) can't be too strict. Parser is going to try float later. So '1.4' must be passed. # User-Facing Changes First and foremost, more specific error messages for typos in string and int literals. Probably improves interactive user experience. But a script that was causing and then checking for specific error might notice a different error message. _(List of all changes that impact the user experience here. This helps us keep track of breaking changes.)_ # Tests + Formatting Added (positive and negative unit tests in `cargo test -p nu-parser`. Didn't add integration tests. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [x] `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [x] `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --------- Co-authored-by: Stefan Holderbach <sholderbach@users.noreply.github.com>
2023-02-13 17:09:50 +01:00
SyntaxShape::String,
];
for shape in shapes.iter() {
let starting_error_count = working_set.parse_errors.len();
let s = parse_value(working_set, span, shape);
if starting_error_count == working_set.parse_errors.len() {
return s;
} else {
match working_set.parse_errors.get(starting_error_count) {
Some(
ParseError::Expected(_, _)
| ParseError::ExpectedWithStringMsg(_, _),
) => {
working_set.parse_errors.truncate(starting_error_count);
continue;
}
_ => {
return s;
}
Syntax errors for string and int (#7952) # Description Added a few syntax errors in ints and strings, changed parser to stop and show that error rather than continue trying to parse those tokens as some other shape. However, I don't see how to push this direction much further, and most of the classic confusing errors can't be changed. Flagged as WIP for the moment, but passes all checks and works better than current release: 1. I have yet to figure out how to make these errors refer back to the book, as I see some other errors do. 2. How to give syntax error when malformed int is first token in line? Currently parsed as external command, user gets confusing error message. 3. Would like to be more strict with *decimal* int literals (lacking, e.g, `0x' prefix). Need to tinker more with the order of parse shape calls, currently, float is tried after int, so '1.4' has to be passed. _(Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience.)_ ```bash 〉"\z" Error: ╭─[entry #3:1:1] 1 │ "\z" · ─┬─ · ╰── Syntax error in string, unrecognized character after escape '\'. ╰──── ``` Canonic presentation of a syntax error. ```bash 〉" \u{01ffbogus}" Error: × Invalid syntax ╭─[entry #2:1:1] 1 │ " \u{01ffbogus}" · ───────┬────── · ╰── Syntax error in string, expecting 1 to 6 hex digits in unicode escape '\u{X...}', max value 10FFFF. ╰──── ``` Malformed unicode escape in string, flagged as error. String parse can be opinionated, it's the last shape tried. ```bash 〉0x22bogus Error: nu::shell::external_command (link) × External command failed ╭─[entry #4:1:1] 1 │ 0x22bogus · ────┬──── · ╰── executable was not found ╰──── help: No such file or directory (os error 2) ``` A *correct* number in first token would be evaluated, but an *incorrect* one is treated as external command? Confusing to users. ```bash 〉0 + 0x22bogus Error: × Invalid syntax ╭─[entry #5:1:1] 1 │ 0 + 0x22bogus · ────┬──── · ╰── Syntax error in int, invalid digits in radix 16 int. ╰──── ``` Can give syntax error if token is unambiguously int literal. e.g has 0b or 0x prefix, could not be a float. ```bash 〉0 + 098bogus Error: nu::parser::unsupported_operation (link) × Types mismatched for operation. ╭─[entry #6:1:1] 1 │ 0 + 098bogus · ┬ ┬ ────┬─── · │ │ ╰── string · │ ╰── doesn't support these values. · ╰── int ╰──── help: Change int or string to be the right types and try again. ``` But *decimal* literal (no prefix) can't be too strict. Parser is going to try float later. So '1.4' must be passed. # User-Facing Changes First and foremost, more specific error messages for typos in string and int literals. Probably improves interactive user experience. But a script that was causing and then checking for specific error might notice a different error message. _(List of all changes that impact the user experience here. This helps us keep track of breaking changes.)_ # Tests + Formatting Added (positive and negative unit tests in `cargo test -p nu-parser`. Didn't add integration tests. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [x] `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [x] `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --------- Co-authored-by: Stefan Holderbach <sholderbach@users.noreply.github.com>
2023-02-13 17:09:50 +01:00
}
}
}
working_set.error(ParseError::Expected("any shape", span));
garbage(working_set, span)
2021-07-02 08:44:37 +02:00
}
2021-09-02 10:25:22 +02:00
}
x => {
working_set.error(ParseError::ExpectedWithStringMsg(
x.to_type().to_string(),
span,
));
garbage(working_set, span)
}
2021-07-02 08:44:37 +02:00
}
2021-09-02 10:25:22 +02:00
}
2021-07-02 08:44:37 +02:00
pub fn parse_operator(working_set: &mut StateWorkingSet, span: Span) -> Expression {
2021-09-02 10:25:22 +02:00
let contents = working_set.get_span_contents(span);
let operator = match contents {
b"=" => Operator::Assignment(Assignment::Assign),
b"+=" => Operator::Assignment(Assignment::PlusAssign),
b"++=" => Operator::Assignment(Assignment::AppendAssign),
b"-=" => Operator::Assignment(Assignment::MinusAssign),
b"*=" => Operator::Assignment(Assignment::MultiplyAssign),
b"/=" => Operator::Assignment(Assignment::DivideAssign),
b"==" => Operator::Comparison(Comparison::Equal),
b"!=" => Operator::Comparison(Comparison::NotEqual),
b"<" => Operator::Comparison(Comparison::LessThan),
b"<=" => Operator::Comparison(Comparison::LessThanOrEqual),
b">" => Operator::Comparison(Comparison::GreaterThan),
b">=" => Operator::Comparison(Comparison::GreaterThanOrEqual),
b"=~" => Operator::Comparison(Comparison::RegexMatch),
b"!~" => Operator::Comparison(Comparison::NotRegexMatch),
b"+" => Operator::Math(Math::Plus),
b"++" => Operator::Math(Math::Append),
b"-" => Operator::Math(Math::Minus),
b"*" => Operator::Math(Math::Multiply),
b"/" => Operator::Math(Math::Divide),
b"//" => Operator::Math(Math::FloorDivision),
b"in" => Operator::Comparison(Comparison::In),
b"not-in" => Operator::Comparison(Comparison::NotIn),
b"mod" => Operator::Math(Math::Modulo),
b"bit-or" => Operator::Bits(Bits::BitOr),
b"bit-xor" => Operator::Bits(Bits::BitXor),
b"bit-and" => Operator::Bits(Bits::BitAnd),
b"bit-shl" => Operator::Bits(Bits::ShiftLeft),
b"bit-shr" => Operator::Bits(Bits::ShiftRight),
b"starts-with" => Operator::Comparison(Comparison::StartsWith),
b"ends-with" => Operator::Comparison(Comparison::EndsWith),
Better errors when bash-like operators are used (#7241) # Description Adds improved errors for when a user uses a bashism that nu doesn't support. fixes #7237 Examples: ``` Error: nu::parser::shell_andand (link) × The '&&' operator is not supported in Nushell ╭─[entry #1:1:1] 1 │ ls && ls · ─┬ · ╰── instead of '&&', use ';' or 'and' ╰──── help: use ';' instead of the shell '&&', or 'and' instead of the boolean '&&' ``` ``` Error: nu::parser::shell_oror (link) × The '||' operator is not supported in Nushell ╭─[entry #8:1:1] 1 │ ls || ls · ─┬ · ╰── instead of '||', use 'try' or 'or' ╰──── help: use 'try' instead of the shell '||', or 'or' instead of the boolean '||' ``` ``` Error: nu::parser::shell_err (link) × The '2>' shell operation is 'err>' in Nushell. ╭─[entry #9:1:1] 1 │ foo 2> bar.txt · ─┬ · ╰── use 'err>' instead of '2>' in Nushell ╰──── ``` ``` Error: nu::parser::shell_outerr (link) × The '2>&1' shell operation is 'out+err>' in Nushell. ╭─[entry #10:1:1] 1 │ foo 2>&1 bar.txt · ──┬─ · ╰── use 'out+err>' instead of '2>&1' in Nushell ╰──── help: Nushell redirection will write all of stdout before stderr. ``` # User-Facing Changes **BREAKING CHANGES** This removes the `&&` and `||` operators. We previously supported by `&&`/`and` and `||`/`or`. With this change, only `and` and `or` are valid boolean operators. # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2022-12-08 00:02:11 +01:00
b"and" => Operator::Boolean(Boolean::And),
b"or" => Operator::Boolean(Boolean::Or),
b"xor" => Operator::Boolean(Boolean::Xor),
b"**" => Operator::Math(Math::Pow),
// WARNING: not actual operators below! Error handling only
pow @ (b"^" | b"pow") => {
working_set.error(ParseError::UnknownOperator(
match pow {
b"^" => "^",
b"pow" => "pow",
_ => unreachable!(),
},
"Use '**' for exponentiation or 'bit-xor' for bitwise XOR.",
span,
));
return garbage(working_set, span);
}
equality @ (b"is" | b"===") => {
working_set.error(ParseError::UnknownOperator(
match equality {
b"is" => "is",
b"===" => "===",
_ => unreachable!(),
},
"Did you mean '=='?",
span,
));
return garbage(working_set, span);
}
b"contains" => {
working_set.error(ParseError::UnknownOperator(
"contains",
"Did you mean '$string =~ $pattern' or '$element in $container'?",
span,
));
return garbage(working_set, span);
}
b"%" => {
working_set.error(ParseError::UnknownOperator(
"%",
"Did you mean 'mod'?",
span,
));
return garbage(working_set, span);
}
b"&" => {
working_set.error(ParseError::UnknownOperator(
"&",
"Did you mean 'bit-and'?",
span,
));
return garbage(working_set, span);
}
b"<<" => {
working_set.error(ParseError::UnknownOperator(
"<<",
"Did you mean 'bit-shl'?",
span,
));
return garbage(working_set, span);
}
b">>" => {
working_set.error(ParseError::UnknownOperator(
">>",
"Did you mean 'bit-shr'?",
span,
));
return garbage(working_set, span);
}
bits @ (b"bits-and" | b"bits-xor" | b"bits-or" | b"bits-shl" | b"bits-shr") => {
working_set.error(ParseError::UnknownOperator(
match bits {
b"bits-and" => "bits-and",
b"bits-xor" => "bits-xor",
b"bits-or" => "bits-or",
b"bits-shl" => "bits-shl",
b"bits-shr" => "bits-shr",
_ => unreachable!(),
},
match bits {
b"bits-and" => "Did you mean 'bit-and'?",
b"bits-xor" => "Did you mean 'bit-xor'?",
b"bits-or" => "Did you mean 'bit-or'?",
b"bits-shl" => "Did you mean 'bit-shl'?",
b"bits-shr" => "Did you mean 'bit-shr'?",
_ => unreachable!(),
},
span,
));
return garbage(working_set, span);
}
2021-09-02 10:25:22 +02:00
_ => {
working_set.error(ParseError::Expected("operator", span));
return garbage(working_set, span);
2021-09-02 10:25:22 +02:00
}
};
2021-07-02 08:44:37 +02:00
Expression::new(working_set, Expr::Operator(operator), span, Type::Any)
2021-09-02 10:25:22 +02:00
}
2021-07-02 08:44:37 +02:00
2021-09-02 10:25:22 +02:00
pub fn parse_math_expression(
working_set: &mut StateWorkingSet,
spans: &[Span],
2021-09-09 23:47:20 +02:00
lhs_row_var_id: Option<VarId>,
) -> Expression {
trace!("parsing: math expression");
2021-09-02 10:25:22 +02:00
// As the expr_stack grows, we increase the required precedence to grow larger
// If, at any time, the operator we're looking at is the same or lower precedence
// of what is in the expression stack, we collapse the expression stack.
//
// This leads to an expression stack that grows under increasing precedence and collapses
// under decreasing/sustained precedence
//
// The end result is a stack that we can fold into binary operations as right associations
// safely.
let mut expr_stack: Vec<Expression> = vec![];
let mut idx = 0;
let mut last_prec = 1000000;
2022-04-06 21:10:25 +02:00
let first_span = working_set.get_span_contents(spans[0]);
Fix precedence of 'not' operator (#11672) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> # Description A bit hackish but this fixes the precedence of the `not` operator. Before: `not false and false` => true Now: `not false and false` => false Fixes #11633 # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass (on Windows make sure to [enable developer mode](https://learn.microsoft.com/en-us/windows/apps/get-started/developer-mode-features-and-debugging)) - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: Jakub Žádník <kubouch@gmail.com>
2024-01-29 20:42:27 +01:00
let mut not_start_spans = vec![];
Add pattern matching (#8590) # Description This adds `match` and basic pattern matching. An example: ``` match $x { 1..10 => { print "Value is between 1 and 10" } { foo: $bar } => { print $"Value has a 'foo' field with value ($bar)" } [$a, $b] => { print $"Value is a list with two items: ($a) and ($b)" } _ => { print "Value is none of the above" } } ``` Like the recent changes to `if` to allow it to be used as an expression, `match` can also be used as an expression. This allows you to assign the result to a variable, eg) `let xyz = match ...` I've also included a short-hand pattern for matching records, as I think it might help when doing a lot of record patterns: `{$foo}` which is equivalent to `{foo: $foo}`. There are still missing components, so consider this the first step in full pattern matching support. Currently missing: * Patterns for strings * Or-patterns (like the `|` in Rust) * Patterns for tables (unclear how we want to match a table, so it'll need some design) * Patterns for binary values * And much more # User-Facing Changes [see above] # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-03-24 02:52:01 +01:00
if first_span == b"if" || first_span == b"match" {
// If expression
if spans.len() > 1 {
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
return parse_call(working_set, spans, spans[0]);
} else {
working_set.error(ParseError::Expected(
"expression",
Span::new(spans[0].end, spans[0].end),
));
return garbage(working_set, spans[0]);
}
} else if first_span == b"not" {
Fix precedence of 'not' operator (#11672) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> # Description A bit hackish but this fixes the precedence of the `not` operator. Before: `not false and false` => true Now: `not false and false` => false Fixes #11633 # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass (on Windows make sure to [enable developer mode](https://learn.microsoft.com/en-us/windows/apps/get-started/developer-mode-features-and-debugging)) - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: Jakub Žádník <kubouch@gmail.com>
2024-01-29 20:42:27 +01:00
not_start_spans.push(spans[idx].start);
idx += 1;
while idx < spans.len() {
let next_value = working_set.get_span_contents(spans[idx]);
if next_value == b"not" {
not_start_spans.push(spans[idx].start);
idx += 1;
} else {
break;
}
}
if idx == spans.len() {
working_set.error(ParseError::Expected(
"expression",
Fix precedence of 'not' operator (#11672) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> # Description A bit hackish but this fixes the precedence of the `not` operator. Before: `not false and false` => true Now: `not false and false` => false Fixes #11633 # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass (on Windows make sure to [enable developer mode](https://learn.microsoft.com/en-us/windows/apps/get-started/developer-mode-features-and-debugging)) - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: Jakub Žádník <kubouch@gmail.com>
2024-01-29 20:42:27 +01:00
Span::new(spans[idx - 1].end, spans[idx - 1].end),
));
return garbage(working_set, spans[idx - 1]);
2022-04-06 21:10:25 +02:00
}
}
Fix precedence of 'not' operator (#11672) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> # Description A bit hackish but this fixes the precedence of the `not` operator. Before: `not false and false` => true Now: `not false and false` => false Fixes #11633 # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass (on Windows make sure to [enable developer mode](https://learn.microsoft.com/en-us/windows/apps/get-started/developer-mode-features-and-debugging)) - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: Jakub Žádník <kubouch@gmail.com>
2024-01-29 20:42:27 +01:00
let mut lhs = parse_value(working_set, spans[idx], &SyntaxShape::Any);
for not_start_span in not_start_spans.iter().rev() {
// lhs = Expression {
// expr: Expr::UnaryNot(Box::new(lhs)),
// span: Span::new(*not_start_span, spans[idx].end),
// ty: Type::Bool,
// custom_completion: None,
// };
lhs = Expression::new(
working_set,
Expr::UnaryNot(Box::new(lhs)),
Span::new(*not_start_span, spans[idx].end),
Type::Bool,
);
Fix precedence of 'not' operator (#11672) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> # Description A bit hackish but this fixes the precedence of the `not` operator. Before: `not false and false` => true Now: `not false and false` => false Fixes #11633 # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass (on Windows make sure to [enable developer mode](https://learn.microsoft.com/en-us/windows/apps/get-started/developer-mode-features-and-debugging)) - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: Jakub Žádník <kubouch@gmail.com>
2024-01-29 20:42:27 +01:00
}
not_start_spans.clear();
2021-09-02 10:25:22 +02:00
idx += 1;
if idx >= spans.len() {
// We already found the one part of our expression, so let's expand
if let Some(row_var_id) = lhs_row_var_id {
expand_to_cell_path(working_set, &mut lhs, row_var_id);
}
}
2021-09-02 10:25:22 +02:00
expr_stack.push(lhs);
while idx < spans.len() {
let op = parse_operator(working_set, spans[idx]);
2021-07-02 08:44:37 +02:00
2021-09-02 10:25:22 +02:00
let op_prec = op.precedence();
2021-07-02 08:44:37 +02:00
2021-09-02 10:25:22 +02:00
idx += 1;
2021-07-02 08:44:37 +02:00
2021-09-02 10:25:22 +02:00
if idx == spans.len() {
// Handle broken math expr `1 +` etc
working_set.error(ParseError::IncompleteMathExpression(spans[idx - 1]));
2021-07-02 08:44:37 +02:00
expr_stack.push(Expression::garbage(working_set, spans[idx - 1]));
expr_stack.push(Expression::garbage(working_set, spans[idx - 1]));
2021-07-22 21:50:59 +02:00
2021-09-02 10:25:22 +02:00
break;
}
2021-07-22 21:50:59 +02:00
support env and mut assignment with `if` block and `match` guard (#9650) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> # Description <!-- Thank you for improving Nushell. Please, check our [contributing guide](../CONTRIBUTING.md) and talk to the core team before making major changes. Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience. --> Fixes: https://github.com/nushell/nushell/issues/9595 So we can do the following in nushell: ```nushell mut a = 3 $a = if 4 == 3 { 10 } else {20} ``` or ```nushell $env.BUILD_EXT = match 3 { 1 => { 'yes!' }, _ => { 'no!' } } ``` # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: WindSoilder <windsoilder@DESKTOP-R8GRJ1D.localdomain>
2023-07-13 10:55:41 +02:00
let content = working_set.get_span_contents(spans[idx]);
// allow `if` to be a special value for assignment.
Fix precedence of 'not' operator (#11672) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> # Description A bit hackish but this fixes the precedence of the `not` operator. Before: `not false and false` => true Now: `not false and false` => false Fixes #11633 # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass (on Windows make sure to [enable developer mode](https://learn.microsoft.com/en-us/windows/apps/get-started/developer-mode-features-and-debugging)) - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: Jakub Žádník <kubouch@gmail.com>
2024-01-29 20:42:27 +01:00
support env and mut assignment with `if` block and `match` guard (#9650) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> # Description <!-- Thank you for improving Nushell. Please, check our [contributing guide](../CONTRIBUTING.md) and talk to the core team before making major changes. Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience. --> Fixes: https://github.com/nushell/nushell/issues/9595 So we can do the following in nushell: ```nushell mut a = 3 $a = if 4 == 3 { 10 } else {20} ``` or ```nushell $env.BUILD_EXT = match 3 { 1 => { 'yes!' }, _ => { 'no!' } } ``` # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: WindSoilder <windsoilder@DESKTOP-R8GRJ1D.localdomain>
2023-07-13 10:55:41 +02:00
if content == b"if" || content == b"match" {
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
let rhs = parse_call(working_set, &spans[idx..], spans[0]);
support env and mut assignment with `if` block and `match` guard (#9650) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> # Description <!-- Thank you for improving Nushell. Please, check our [contributing guide](../CONTRIBUTING.md) and talk to the core team before making major changes. Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience. --> Fixes: https://github.com/nushell/nushell/issues/9595 So we can do the following in nushell: ```nushell mut a = 3 $a = if 4 == 3 { 10 } else {20} ``` or ```nushell $env.BUILD_EXT = match 3 { 1 => { 'yes!' }, _ => { 'no!' } } ``` # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: WindSoilder <windsoilder@DESKTOP-R8GRJ1D.localdomain>
2023-07-13 10:55:41 +02:00
expr_stack.push(op);
expr_stack.push(rhs);
break;
Fix precedence of 'not' operator (#11672) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> # Description A bit hackish but this fixes the precedence of the `not` operator. Before: `not false and false` => true Now: `not false and false` => false Fixes #11633 # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass (on Windows make sure to [enable developer mode](https://learn.microsoft.com/en-us/windows/apps/get-started/developer-mode-features-and-debugging)) - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: Jakub Žádník <kubouch@gmail.com>
2024-01-29 20:42:27 +01:00
} else if content == b"not" {
not_start_spans.push(spans[idx].start);
idx += 1;
while idx < spans.len() {
let next_value = working_set.get_span_contents(spans[idx]);
if next_value == b"not" {
not_start_spans.push(spans[idx].start);
idx += 1;
} else {
break;
}
}
if idx == spans.len() {
working_set.error(ParseError::Expected(
"expression",
Span::new(spans[idx - 1].end, spans[idx - 1].end),
));
return garbage(working_set, spans[idx - 1]);
Fix precedence of 'not' operator (#11672) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> # Description A bit hackish but this fixes the precedence of the `not` operator. Before: `not false and false` => true Now: `not false and false` => false Fixes #11633 # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass (on Windows make sure to [enable developer mode](https://learn.microsoft.com/en-us/windows/apps/get-started/developer-mode-features-and-debugging)) - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: Jakub Žádník <kubouch@gmail.com>
2024-01-29 20:42:27 +01:00
}
}
let mut rhs = parse_value(working_set, spans[idx], &SyntaxShape::Any);
for not_start_span in not_start_spans.iter().rev() {
// rhs = Expression {
// expr: Expr::UnaryNot(Box::new(rhs)),
// span: Span::new(*not_start_span, spans[idx].end),
// ty: Type::Bool,
// custom_completion: None,
// };
rhs = Expression::new(
working_set,
Expr::UnaryNot(Box::new(rhs)),
Span::new(*not_start_span, spans[idx].end),
Type::Bool,
);
support env and mut assignment with `if` block and `match` guard (#9650) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> # Description <!-- Thank you for improving Nushell. Please, check our [contributing guide](../CONTRIBUTING.md) and talk to the core team before making major changes. Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience. --> Fixes: https://github.com/nushell/nushell/issues/9595 So we can do the following in nushell: ```nushell mut a = 3 $a = if 4 == 3 { 10 } else {20} ``` or ```nushell $env.BUILD_EXT = match 3 { 1 => { 'yes!' }, _ => { 'no!' } } ``` # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: WindSoilder <windsoilder@DESKTOP-R8GRJ1D.localdomain>
2023-07-13 10:55:41 +02:00
}
Fix precedence of 'not' operator (#11672) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> # Description A bit hackish but this fixes the precedence of the `not` operator. Before: `not false and false` => true Now: `not false and false` => false Fixes #11633 # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass (on Windows make sure to [enable developer mode](https://learn.microsoft.com/en-us/windows/apps/get-started/developer-mode-features-and-debugging)) - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: Jakub Žádník <kubouch@gmail.com>
2024-01-29 20:42:27 +01:00
not_start_spans.clear();
2021-07-02 08:44:37 +02:00
2022-03-25 04:23:08 +01:00
while op_prec <= last_prec && expr_stack.len() > 1 {
2021-11-06 08:31:28 +01:00
// Collapse the right associated operations first
// so that we can get back to a stack with a lower precedence
let mut rhs = expr_stack
.pop()
.expect("internal error: expression stack empty");
let mut op = expr_stack
.pop()
.expect("internal error: expression stack empty");
2022-03-25 04:23:08 +01:00
last_prec = op.precedence();
if last_prec < op_prec {
expr_stack.push(op);
expr_stack.push(rhs);
break;
}
2021-11-06 08:31:28 +01:00
let mut lhs = expr_stack
.pop()
.expect("internal error: expression stack empty");
if let Some(row_var_id) = lhs_row_var_id {
expand_to_cell_path(working_set, &mut lhs, row_var_id);
2021-11-06 08:31:28 +01:00
}
2021-09-09 23:47:20 +02:00
2021-11-06 08:31:28 +01:00
let (result_ty, err) = math_result_type(working_set, &mut lhs, &mut op, &mut rhs);
if let Some(err) = err {
working_set.error(err);
}
2021-07-02 08:44:37 +02:00
let op_span = Span::append(lhs.span, rhs.span);
expr_stack.push(Expression::new(
working_set,
Expr::BinaryOp(Box::new(lhs), Box::new(op), Box::new(rhs)),
op_span,
result_ty,
));
2021-07-02 08:44:37 +02:00
}
2021-09-02 10:25:22 +02:00
expr_stack.push(op);
expr_stack.push(rhs);
2021-07-02 08:44:37 +02:00
2021-09-02 10:25:22 +02:00
last_prec = op_prec;
2021-07-23 23:19:30 +02:00
2021-09-02 10:25:22 +02:00
idx += 1;
}
2021-07-02 08:44:37 +02:00
2021-09-02 10:25:22 +02:00
while expr_stack.len() != 1 {
let mut rhs = expr_stack
.pop()
.expect("internal error: expression stack empty");
let mut op = expr_stack
2021-07-02 08:44:37 +02:00
.pop()
.expect("internal error: expression stack empty");
2021-09-02 10:25:22 +02:00
let mut lhs = expr_stack
.pop()
.expect("internal error: expression stack empty");
2021-09-09 23:47:20 +02:00
if let Some(row_var_id) = lhs_row_var_id {
expand_to_cell_path(working_set, &mut lhs, row_var_id);
2021-09-09 23:47:20 +02:00
}
2021-09-02 10:25:22 +02:00
let (result_ty, err) = math_result_type(working_set, &mut lhs, &mut op, &mut rhs);
if let Some(err) = err {
working_set.error(err)
}
2021-07-02 08:44:37 +02:00
let binary_op_span = Span::append(lhs.span, rhs.span);
expr_stack.push(Expression::new(
working_set,
Expr::BinaryOp(Box::new(lhs), Box::new(op), Box::new(rhs)),
binary_op_span,
result_ty,
));
2021-07-01 02:01:04 +02:00
}
expr_stack
2021-09-02 10:25:22 +02:00
.pop()
.expect("internal error: expression stack empty")
2021-09-02 10:25:22 +02:00
}
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
pub fn parse_expression(working_set: &mut StateWorkingSet, spans: &[Span]) -> Expression {
trace!("parsing: expression");
2021-11-04 03:32:35 +01:00
let mut pos = 0;
let mut shorthand = vec![];
2021-09-02 10:25:22 +02:00
2021-11-04 03:32:35 +01:00
while pos < spans.len() {
// Check if there is any environment shorthand
let name = working_set.get_span_contents(spans[pos]);
let split = name.splitn(2, |x| *x == b'=');
2021-11-04 03:32:35 +01:00
let split: Vec<_> = split.collect();
range operator accepts bot..=top as well as bot..top (#8382) # Description A compromise fix for #8162. Nushell range operator now accepts `..=` to mean the range includes the top value, so you can use your Rust habits. But the unadorned `..` range operator also includes the value, so you can also use your Nushell habits. _(Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience.)_ ```nushell 〉1..5 ╭───┬───╮ │ 0 │ 1 │ │ 1 │ 2 │ │ 2 │ 3 │ │ 3 │ 4 │ │ 4 │ 5 │ ╰───┴───╯ -------------------------------------------- /home/bobhy/src/rust/nushell -------------------------------------------- 〉1..=5 ╭───┬───╮ │ 0 │ 1 │ │ 1 │ 2 │ │ 2 │ 3 │ │ 3 │ 4 │ │ 4 │ 5 │ ╰───┴───╯ -------------------------------------------- /home/bobhy/src/rust/nushell -------------------------------------------- 〉1..<5 ╭───┬───╮ │ 0 │ 1 │ │ 1 │ 2 │ │ 2 │ 3 │ │ 3 │ 4 │ ╰───┴───╯ ``` # User-Facing Changes Existing scripts with range operator will continue to operate as heretofore. _(List of all changes that impact the user experience here. This helps us keep track of breaking changes.)_ # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [x] `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [x] `cargo test --workspace` to check that all tests pass # After Submitting Will update the book to include new syntax.
2023-04-07 13:40:05 +02:00
if !name.starts_with(b"^")
&& split.len() == 2
&& !split[0].is_empty()
&& !split[0].ends_with(b"..")
// was range op ..=
{
2021-11-04 03:32:35 +01:00
let point = split[0].len() + 1;
let starting_error_count = working_set.parse_errors.len();
2021-11-04 03:32:35 +01:00
let lhs = parse_string_strict(
working_set,
Protocol: debug_assert!() Span to reflect a valid slice (#6806) Also enforce this by #[non_exhaustive] span such that going forward we cannot, in debug builds (1), construct invalid spans. The motivation for this stems from #6431 where I've seen crashes due to invalid slice indexing. My hope is this will mitigate such senarios 1. https://github.com/nushell/nushell/pull/6431#issuecomment-1278147241 # Description (description of your pull request here) # Tests Make sure you've done the following: - [ ] Add tests that cover your changes, either in the command examples, the crate/tests folder, or in the /tests folder. - [ ] Try to think about corner cases and various ways how your changes could break. Cover them with tests. - [ ] If adding tests is not possible, please document in the PR body a minimal example with steps on how to reproduce so one can verify your change works. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [ ] `cargo clippy --workspace --features=extra -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [ ] `cargo test --workspace --features=extra` to check that all the tests pass # Documentation - [ ] If your PR touches a user-facing nushell feature then make sure that there is an entry in the documentation (https://github.com/nushell/nushell.github.io) for the feature, and update it if necessary.
2022-12-03 10:44:12 +01:00
Span::new(spans[pos].start, spans[pos].start + point - 1),
2021-11-04 03:32:35 +01:00
);
let rhs = if spans[pos].start + point < spans[pos].end {
Protocol: debug_assert!() Span to reflect a valid slice (#6806) Also enforce this by #[non_exhaustive] span such that going forward we cannot, in debug builds (1), construct invalid spans. The motivation for this stems from #6431 where I've seen crashes due to invalid slice indexing. My hope is this will mitigate such senarios 1. https://github.com/nushell/nushell/pull/6431#issuecomment-1278147241 # Description (description of your pull request here) # Tests Make sure you've done the following: - [ ] Add tests that cover your changes, either in the command examples, the crate/tests folder, or in the /tests folder. - [ ] Try to think about corner cases and various ways how your changes could break. Cover them with tests. - [ ] If adding tests is not possible, please document in the PR body a minimal example with steps on how to reproduce so one can verify your change works. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [ ] `cargo clippy --workspace --features=extra -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [ ] `cargo test --workspace --features=extra` to check that all the tests pass # Documentation - [ ] If your PR touches a user-facing nushell feature then make sure that there is an entry in the documentation (https://github.com/nushell/nushell.github.io) for the feature, and update it if necessary.
2022-12-03 10:44:12 +01:00
let rhs_span = Span::new(spans[pos].start + point, spans[pos].end);
if working_set.get_span_contents(rhs_span).starts_with(b"$") {
parse_dollar_expr(working_set, rhs_span)
} else {
parse_string_strict(working_set, rhs_span)
}
2021-11-04 03:32:35 +01:00
} else {
Expression::new(
working_set,
Expr::String(String::new()),
Span::unknown(),
Type::Nothing,
)
2021-11-04 03:32:35 +01:00
};
if starting_error_count == working_set.parse_errors.len() {
shorthand.push((lhs, rhs));
2021-11-04 03:32:35 +01:00
pos += 1;
} else {
working_set.parse_errors.truncate(starting_error_count);
2021-11-04 03:32:35 +01:00
break;
}
} else {
break;
}
}
if pos == spans.len() {
working_set.error(ParseError::UnknownCommand(spans[0]));
return garbage(working_set, Span::concat(spans));
2021-11-04 03:32:35 +01:00
}
let output = if is_math_expression_like(working_set, spans[pos]) {
parse_math_expression(working_set, &spans[pos..], None)
2021-11-04 03:32:35 +01:00
} else {
let bytes = working_set.get_span_contents(spans[pos]).to_vec();
// For now, check for special parses of certain keywords
match bytes.as_slice() {
b"def" | b"extern" | b"for" | b"module" | b"use" | b"source" | b"alias" | b"export"
| b"hide" => {
working_set.error(ParseError::BuiltinCommandInPipeline(
String::from_utf8(bytes)
.expect("builtin commands bytes should be able to convert to string"),
spans[0],
));
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
parse_call(working_set, &spans[pos..], spans[0])
}
b"let" | b"const" | b"mut" => {
working_set.error(ParseError::AssignInPipeline(
String::from_utf8(bytes)
.expect("builtin commands bytes should be able to convert to string"),
String::from_utf8_lossy(match spans.len() {
1..=3 => b"value",
_ => working_set.get_span_contents(spans[3]),
})
.to_string(),
String::from_utf8_lossy(match spans.len() {
1 => b"variable",
_ => working_set.get_span_contents(spans[1]),
})
.to_string(),
spans[0],
));
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
parse_call(working_set, &spans[pos..], spans[0])
}
Overlays (#5375) * WIP: Start laying overlays * Rename Overlay->Module; Start adding overlay * Revamp adding overlay * Add overlay add tests; Disable debug print * Fix overlay add; Add overlay remove * Add overlay remove tests * Add missing overlay remove file * Add overlay list command * (WIP?) Enable overlays for env vars * Move OverlayFrames to ScopeFrames * (WIP) Move everything to overlays only ScopeFrame contains nothing but overlays now * Fix predecls * Fix wrong overlay id translation and aliases * Fix broken env lookup logic * Remove TODOs * Add overlay add + remove for environment * Add a few overlay tests; Fix overlay add name * Some cleanup; Fix overlay add/remove names * Clippy * Fmt * Remove walls of comments * List overlays from stack; Add debugging flag Currently, the engine state ordering is somehow broken. * Fix (?) overlay list test * Fix tests on Windows * Fix activated overlay ordering * Check for active overlays equality in overlay list This removes the -p flag: Either both parser and engine will have the same overlays, or the command will fail. * Add merging on overlay remove * Change help message and comment * Add some remove-merge/discard tests * (WIP) Track removed overlays properly * Clippy; Fmt * Fix getting last overlay; Fix predecls in overlays * Remove merging; Fix re-add overwriting stuff Also some error message tweaks. * Fix overlay error in the engine * Update variable_completions.rs * Adds flags and optional arguments to view-source (#5446) * added flags and optional arguments to view-source * removed redundant code * removed redundant code * fmt * fix bug in shell_integration (#5450) * fix bug in shell_integration * add some comments * enable cd to work with directory abbreviations (#5452) * enable cd to work with abbreviations * add abbreviation example * fix tests * make it configurable * make cd recornize symblic link (#5454) * implement seq char command to generate single character sequence (#5453) * add tmp code * add seq char command * Add split number flag in `split row` (#5434) Signed-off-by: Yuheng Su <gipsyh.icu@gmail.com> * Add two more overlay tests * Add ModuleId to OverlayFrame * Fix env conversion accidentally activating overlay It activated overlay from permanent state prematurely which would cause `overlay add` to misbehave. * Remove unused parameter; Add overlay list test * Remove added traces * Add overlay commands examples * Modify TODO * Fix $nu.scope iteration * Disallow removing default overlay * Refactor some parser errors * Remove last overlay if no argument * Diversify overlay examples * Make it possible to update overlay's module In case the origin module updates, the overlay add loads the new module, makes it overlay's origin and applies the changes. Before, it was impossible to update the overlay if the module changed. Co-authored-by: JT <547158+jntrnr@users.noreply.github.com> Co-authored-by: pwygab <88221256+merelymyself@users.noreply.github.com> Co-authored-by: Darren Schroeder <343840+fdncred@users.noreply.github.com> Co-authored-by: WindSoilder <WindSoilder@outlook.com> Co-authored-by: Yuheng Su <gipsyh.icu@gmail.com>
2022-05-07 21:39:22 +02:00
b"overlay" => {
if spans.len() > 1 && working_set.get_span_contents(spans[1]) == b"list" {
// whitelist 'overlay list'
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
parse_call(working_set, &spans[pos..], spans[0])
Overlays (#5375) * WIP: Start laying overlays * Rename Overlay->Module; Start adding overlay * Revamp adding overlay * Add overlay add tests; Disable debug print * Fix overlay add; Add overlay remove * Add overlay remove tests * Add missing overlay remove file * Add overlay list command * (WIP?) Enable overlays for env vars * Move OverlayFrames to ScopeFrames * (WIP) Move everything to overlays only ScopeFrame contains nothing but overlays now * Fix predecls * Fix wrong overlay id translation and aliases * Fix broken env lookup logic * Remove TODOs * Add overlay add + remove for environment * Add a few overlay tests; Fix overlay add name * Some cleanup; Fix overlay add/remove names * Clippy * Fmt * Remove walls of comments * List overlays from stack; Add debugging flag Currently, the engine state ordering is somehow broken. * Fix (?) overlay list test * Fix tests on Windows * Fix activated overlay ordering * Check for active overlays equality in overlay list This removes the -p flag: Either both parser and engine will have the same overlays, or the command will fail. * Add merging on overlay remove * Change help message and comment * Add some remove-merge/discard tests * (WIP) Track removed overlays properly * Clippy; Fmt * Fix getting last overlay; Fix predecls in overlays * Remove merging; Fix re-add overwriting stuff Also some error message tweaks. * Fix overlay error in the engine * Update variable_completions.rs * Adds flags and optional arguments to view-source (#5446) * added flags and optional arguments to view-source * removed redundant code * removed redundant code * fmt * fix bug in shell_integration (#5450) * fix bug in shell_integration * add some comments * enable cd to work with directory abbreviations (#5452) * enable cd to work with abbreviations * add abbreviation example * fix tests * make it configurable * make cd recornize symblic link (#5454) * implement seq char command to generate single character sequence (#5453) * add tmp code * add seq char command * Add split number flag in `split row` (#5434) Signed-off-by: Yuheng Su <gipsyh.icu@gmail.com> * Add two more overlay tests * Add ModuleId to OverlayFrame * Fix env conversion accidentally activating overlay It activated overlay from permanent state prematurely which would cause `overlay add` to misbehave. * Remove unused parameter; Add overlay list test * Remove added traces * Add overlay commands examples * Modify TODO * Fix $nu.scope iteration * Disallow removing default overlay * Refactor some parser errors * Remove last overlay if no argument * Diversify overlay examples * Make it possible to update overlay's module In case the origin module updates, the overlay add loads the new module, makes it overlay's origin and applies the changes. Before, it was impossible to update the overlay if the module changed. Co-authored-by: JT <547158+jntrnr@users.noreply.github.com> Co-authored-by: pwygab <88221256+merelymyself@users.noreply.github.com> Co-authored-by: Darren Schroeder <343840+fdncred@users.noreply.github.com> Co-authored-by: WindSoilder <WindSoilder@outlook.com> Co-authored-by: Yuheng Su <gipsyh.icu@gmail.com>
2022-05-07 21:39:22 +02:00
} else {
working_set.error(ParseError::BuiltinCommandInPipeline(
"overlay".into(),
spans[0],
));
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
parse_call(working_set, &spans[pos..], spans[0])
Overlays (#5375) * WIP: Start laying overlays * Rename Overlay->Module; Start adding overlay * Revamp adding overlay * Add overlay add tests; Disable debug print * Fix overlay add; Add overlay remove * Add overlay remove tests * Add missing overlay remove file * Add overlay list command * (WIP?) Enable overlays for env vars * Move OverlayFrames to ScopeFrames * (WIP) Move everything to overlays only ScopeFrame contains nothing but overlays now * Fix predecls * Fix wrong overlay id translation and aliases * Fix broken env lookup logic * Remove TODOs * Add overlay add + remove for environment * Add a few overlay tests; Fix overlay add name * Some cleanup; Fix overlay add/remove names * Clippy * Fmt * Remove walls of comments * List overlays from stack; Add debugging flag Currently, the engine state ordering is somehow broken. * Fix (?) overlay list test * Fix tests on Windows * Fix activated overlay ordering * Check for active overlays equality in overlay list This removes the -p flag: Either both parser and engine will have the same overlays, or the command will fail. * Add merging on overlay remove * Change help message and comment * Add some remove-merge/discard tests * (WIP) Track removed overlays properly * Clippy; Fmt * Fix getting last overlay; Fix predecls in overlays * Remove merging; Fix re-add overwriting stuff Also some error message tweaks. * Fix overlay error in the engine * Update variable_completions.rs * Adds flags and optional arguments to view-source (#5446) * added flags and optional arguments to view-source * removed redundant code * removed redundant code * fmt * fix bug in shell_integration (#5450) * fix bug in shell_integration * add some comments * enable cd to work with directory abbreviations (#5452) * enable cd to work with abbreviations * add abbreviation example * fix tests * make it configurable * make cd recornize symblic link (#5454) * implement seq char command to generate single character sequence (#5453) * add tmp code * add seq char command * Add split number flag in `split row` (#5434) Signed-off-by: Yuheng Su <gipsyh.icu@gmail.com> * Add two more overlay tests * Add ModuleId to OverlayFrame * Fix env conversion accidentally activating overlay It activated overlay from permanent state prematurely which would cause `overlay add` to misbehave. * Remove unused parameter; Add overlay list test * Remove added traces * Add overlay commands examples * Modify TODO * Fix $nu.scope iteration * Disallow removing default overlay * Refactor some parser errors * Remove last overlay if no argument * Diversify overlay examples * Make it possible to update overlay's module In case the origin module updates, the overlay add loads the new module, makes it overlay's origin and applies the changes. Before, it was impossible to update the overlay if the module changed. Co-authored-by: JT <547158+jntrnr@users.noreply.github.com> Co-authored-by: pwygab <88221256+merelymyself@users.noreply.github.com> Co-authored-by: Darren Schroeder <343840+fdncred@users.noreply.github.com> Co-authored-by: WindSoilder <WindSoilder@outlook.com> Co-authored-by: Yuheng Su <gipsyh.icu@gmail.com>
2022-05-07 21:39:22 +02:00
}
}
b"where" => parse_where_expr(working_set, &spans[pos..]),
#[cfg(feature = "plugin")]
b"register" => {
working_set.error(ParseError::BuiltinCommandInPipeline(
Deprecate `register` and add `plugin use` (#12607) # Description Adds a new keyword, `plugin use`. Unlike `register`, this merely loads the signatures from the plugin cache file. The file is configurable with the `--plugin-config` option either to `nu` or to `plugin use` itself, just like the other `plugin` family of commands. At the REPL, one might do this to replace `register`: ```nushell > plugin add ~/.cargo/bin/nu_plugin_foo > plugin use foo ``` This will not work in a script, because `plugin use` is a keyword and `plugin add` does not evaluate at parse time (intentionally). This means we no longer run random binaries during parse. The `--plugins` option has been added to allow running `nu` with certain plugins in one step. This is used especially for the `nu_with_plugins!` test macro, but I'd imagine is generally useful. The only weird quirk is that it has to be a list, and we don't really do this for any of our other CLI args at the moment. `register` now prints a deprecation parse warning. This should fix #11923, as we now have a complete alternative to `register`. # User-Facing Changes - Add `plugin use` command - Deprecate `register` - Add `--plugins` option to `nu` to replace a common use of `register` # Tests + Formatting I think I've tested it thoroughly enough and every existing test passes. Testing nu CLI options and alternate config files is a little hairy and I wish there were some more generic helpers for this, so this will go on my TODO list for refactoring. - :green_circle: `toolkit fmt` - :green_circle: `toolkit clippy` - :green_circle: `toolkit test` - :green_circle: `toolkit test stdlib` # After Submitting - [ ] Update plugins sections of book - [ ] Release notes
2024-04-23 13:37:50 +02:00
"register".into(),
spans[0],
));
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
parse_call(working_set, &spans[pos..], spans[0])
}
Deprecate `register` and add `plugin use` (#12607) # Description Adds a new keyword, `plugin use`. Unlike `register`, this merely loads the signatures from the plugin cache file. The file is configurable with the `--plugin-config` option either to `nu` or to `plugin use` itself, just like the other `plugin` family of commands. At the REPL, one might do this to replace `register`: ```nushell > plugin add ~/.cargo/bin/nu_plugin_foo > plugin use foo ``` This will not work in a script, because `plugin use` is a keyword and `plugin add` does not evaluate at parse time (intentionally). This means we no longer run random binaries during parse. The `--plugins` option has been added to allow running `nu` with certain plugins in one step. This is used especially for the `nu_with_plugins!` test macro, but I'd imagine is generally useful. The only weird quirk is that it has to be a list, and we don't really do this for any of our other CLI args at the moment. `register` now prints a deprecation parse warning. This should fix #11923, as we now have a complete alternative to `register`. # User-Facing Changes - Add `plugin use` command - Deprecate `register` - Add `--plugins` option to `nu` to replace a common use of `register` # Tests + Formatting I think I've tested it thoroughly enough and every existing test passes. Testing nu CLI options and alternate config files is a little hairy and I wish there were some more generic helpers for this, so this will go on my TODO list for refactoring. - :green_circle: `toolkit fmt` - :green_circle: `toolkit clippy` - :green_circle: `toolkit test` - :green_circle: `toolkit test stdlib` # After Submitting - [ ] Update plugins sections of book - [ ] Release notes
2024-04-23 13:37:50 +02:00
#[cfg(feature = "plugin")]
b"plugin" => {
if spans.len() > 1 && working_set.get_span_contents(spans[1]) == b"use" {
// only 'plugin use' is banned
working_set.error(ParseError::BuiltinCommandInPipeline(
"plugin use".into(),
spans[0],
));
}
parse_call(working_set, &spans[pos..], spans[0])
}
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
_ => parse_call(working_set, &spans[pos..], spans[0]),
}
2021-11-04 03:32:35 +01:00
};
if !shorthand.is_empty() {
let with_env = working_set.find_decl(b"with-env");
2021-11-04 03:32:35 +01:00
if let Some(decl_id) = with_env {
let mut block = Block::default();
let ty = output.ty.clone();
block.pipelines = vec![Pipeline::from_vec(vec![output])];
2021-11-04 03:32:35 +01:00
Make EngineState clone cheaper with Arc on all of the heavy objects (#12229) # Description This makes many of the larger objects in `EngineState` into `Arc`, and uses `Arc::make_mut` to do clone-on-write if the reference is not unique. This is generally very cheap, giving us the best of both worlds - allowing us to mutate without cloning if we have an exclusive reference, and cloning if we don't. This started as more of a curiosity for me after remembering that `Arc::make_mut` exists and can make using `Arc` for mostly immutable data that sometimes needs to be changed very convenient, and also after hearing someone complain about memory usage on Discord - this is a somewhat significant win for that. The exact objects that were wrapped in `Arc`: - `files`, `file_contents` - the strings and byte buffers - `decls` - the whole `Vec`, but mostly to avoid lots of individual `malloc()` calls on Clone rather than for memory usage - `blocks` - the blocks themselves, rather than the outer Vec - `modules` - the modules themselves, rather than the outer Vec - `env_vars`, `previous_env_vars` - the entire maps - `config` The changes required were relatively minimal, but this is a breaking API change. In particular, blocks are added as Arcs, to allow the parser cache functionality to work. With my normal nu config, running on Linux, this saves me about 15 MiB of process memory usage when running interactively (65 MiB → 50 MiB). This also makes quick command executions cheaper, particularly since every REPL loop now involves a clone of the engine state so that we can recover from a panic. It also reduces memory usage where engine state needs to be cloned and sent to another thread or kept within an iterator. # User-Facing Changes Shouldn't be any, since it's all internal stuff, but it does change some public interfaces so it's a breaking change
2024-03-19 19:07:00 +01:00
let block_id = working_set.add_block(Arc::new(block));
2021-11-04 03:32:35 +01:00
let mut env_vars = vec![];
for sh in shorthand {
env_vars.push(RecordItem::Pair(sh.0, sh.1));
2021-11-04 03:32:35 +01:00
}
let arguments = vec![
Argument::Positional(Expression::new(
working_set,
Expr::Record(env_vars),
Span::concat(&spans[..pos]),
Type::Any,
)),
Argument::Positional(Expression::new(
working_set,
Expr::Closure(block_id),
Span::concat(&spans[pos..]),
Type::Closure,
)),
2021-11-04 03:32:35 +01:00
];
2022-02-21 18:58:04 +01:00
let expr = Expr::Call(Box::new(Call {
Protocol: debug_assert!() Span to reflect a valid slice (#6806) Also enforce this by #[non_exhaustive] span such that going forward we cannot, in debug builds (1), construct invalid spans. The motivation for this stems from #6431 where I've seen crashes due to invalid slice indexing. My hope is this will mitigate such senarios 1. https://github.com/nushell/nushell/pull/6431#issuecomment-1278147241 # Description (description of your pull request here) # Tests Make sure you've done the following: - [ ] Add tests that cover your changes, either in the command examples, the crate/tests folder, or in the /tests folder. - [ ] Try to think about corner cases and various ways how your changes could break. Cover them with tests. - [ ] If adding tests is not possible, please document in the PR body a minimal example with steps on how to reproduce so one can verify your change works. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [ ] `cargo clippy --workspace --features=extra -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [ ] `cargo test --workspace --features=extra` to check that all the tests pass # Documentation - [ ] If your PR touches a user-facing nushell feature then make sure that there is an entry in the documentation (https://github.com/nushell/nushell.github.io) for the feature, and update it if necessary.
2022-12-03 10:44:12 +01:00
head: Span::unknown(),
2022-02-21 18:58:04 +01:00
decl_id,
arguments,
parser_info: HashMap::new(),
2022-02-21 18:58:04 +01:00
}));
Expression::new(working_set, expr, Span::concat(spans), ty)
2021-11-04 03:32:35 +01:00
} else {
output
2021-11-04 03:32:35 +01:00
}
} else {
output
2021-07-01 02:01:04 +02:00
}
2021-09-02 10:25:22 +02:00
}
2021-07-01 02:01:04 +02:00
pub fn parse_variable(working_set: &mut StateWorkingSet, span: Span) -> Option<VarId> {
2021-09-02 10:25:22 +02:00
let bytes = working_set.get_span_contents(span);
2021-07-01 02:01:04 +02:00
2021-09-02 10:25:22 +02:00
if is_variable(bytes) {
Remove broken compile-time overload system (#9677) # Description This PR removes the compile-time overload system. Unfortunately, this system never worked correctly because in a gradual type system where types can be `Any`, you don't have enough information to correctly resolve function calls with overloads. These resolutions must be done at runtime, if they're supported. That said, there's a bit of work that needs to go into resolving input/output types (here overloads do not execute separate commands, but the same command and each overload explains how each output type corresponds to input types). This PR also removes the type scope, which would give incorrect answers in cases where multiple subexpressions were used in a pipeline. # User-Facing Changes Finishes removing compile-time overloads. These were only used in a few places in the code base, but it's possible it may impact user code. I'll mark this as breaking change so we can review. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-13 21:05:03 +02:00
working_set.find_variable(bytes)
2021-09-02 10:25:22 +02:00
} else {
working_set.error(ParseError::Expected("valid variable name", span));
None
2021-07-01 02:01:04 +02:00
}
2021-09-02 10:25:22 +02:00
}
2021-07-01 02:01:04 +02:00
pub fn parse_builtin_commands(
working_set: &mut StateWorkingSet,
lite_command: &LiteCommand,
) -> Pipeline {
Improve type hovers (#9515) # Description This PR does a few things to help improve type hovers and, in the process, fixes a few outstanding issues in the type system. Here's a list of the changes: * `for` now will try to infer the type of the iteration variable based on the expression it's given. This fixes things like `for x in [1, 2, 3] { }` where `x` now properly gets the int type. * Removed old input/output type fields from the signature, focuses on the vec of signatures. Updated a bunch of dataframe commands that hadn't moved over. This helps tie things together a bit better * Fixed inference of types from subexpressions to use the last expression in the block * Fixed handling of explicit types in `let` and `mut` calls, so we now respect that as the authoritative type I also tried to add `def` input/output type inference, but unfortunately we only know the predecl types universally, which means we won't have enough information to properly know what the types of the custom commands are. # User-Facing Changes Script typechecking will get tighter in some cases Hovers should be more accurate in some cases that previously resorted to any. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: Darren Schroeder <343840+fdncred@users.noreply.github.com>
2023-06-28 19:19:48 +02:00
trace!("parsing: builtin commands");
if !is_math_expression_like(working_set, lite_command.parts[0])
2023-03-10 22:20:31 +01:00
&& !is_unaliasable_parser_keyword(working_set, &lite_command.parts)
{
Improve type hovers (#9515) # Description This PR does a few things to help improve type hovers and, in the process, fixes a few outstanding issues in the type system. Here's a list of the changes: * `for` now will try to infer the type of the iteration variable based on the expression it's given. This fixes things like `for x in [1, 2, 3] { }` where `x` now properly gets the int type. * Removed old input/output type fields from the signature, focuses on the vec of signatures. Updated a bunch of dataframe commands that hadn't moved over. This helps tie things together a bit better * Fixed inference of types from subexpressions to use the last expression in the block * Fixed handling of explicit types in `let` and `mut` calls, so we now respect that as the authoritative type I also tried to add `def` input/output type inference, but unfortunately we only know the predecl types universally, which means we won't have enough information to properly know what the types of the custom commands are. # User-Facing Changes Script typechecking will get tighter in some cases Hovers should be more accurate in some cases that previously resorted to any. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: Darren Schroeder <343840+fdncred@users.noreply.github.com>
2023-06-28 19:19:48 +02:00
trace!("parsing: not math expression or unaliasable parser keyword");
2023-03-10 22:20:31 +01:00
let name = working_set.get_span_contents(lite_command.parts[0]);
Remove broken compile-time overload system (#9677) # Description This PR removes the compile-time overload system. Unfortunately, this system never worked correctly because in a gradual type system where types can be `Any`, you don't have enough information to correctly resolve function calls with overloads. These resolutions must be done at runtime, if they're supported. That said, there's a bit of work that needs to go into resolving input/output types (here overloads do not execute separate commands, but the same command and each overload explains how each output type corresponds to input types). This PR also removes the type scope, which would give incorrect answers in cases where multiple subexpressions were used in a pipeline. # User-Facing Changes Finishes removing compile-time overloads. These were only used in a few places in the code base, but it's possible it may impact user code. I'll mark this as breaking change so we can review. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-13 21:05:03 +02:00
if let Some(decl_id) = working_set.find_decl(name) {
2023-03-10 22:20:31 +01:00
let cmd = working_set.get_decl(decl_id);
if cmd.is_alias() {
// Parse keywords that can be aliased. Note that we check for "unaliasable" keywords
// because alias can have any name, therefore, we can't check for "aliasable" keywords.
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
let call_expr = parse_call(working_set, &lite_command.parts, lite_command.parts[0]);
2023-03-10 22:20:31 +01:00
if let Expression {
expr: Expr::Call(call),
..
} = call_expr
{
// Apply parse keyword side effects
let cmd = working_set.get_decl(call.decl_id);
match cmd.name() {
"overlay hide" => return parse_overlay_hide(working_set, call),
"overlay new" => return parse_overlay_new(working_set, call),
"overlay use" => return parse_overlay_use(working_set, call),
_ => { /* this alias is not a parser keyword */ }
2023-03-10 22:20:31 +01:00
}
}
}
}
}
Improve type hovers (#9515) # Description This PR does a few things to help improve type hovers and, in the process, fixes a few outstanding issues in the type system. Here's a list of the changes: * `for` now will try to infer the type of the iteration variable based on the expression it's given. This fixes things like `for x in [1, 2, 3] { }` where `x` now properly gets the int type. * Removed old input/output type fields from the signature, focuses on the vec of signatures. Updated a bunch of dataframe commands that hadn't moved over. This helps tie things together a bit better * Fixed inference of types from subexpressions to use the last expression in the block * Fixed handling of explicit types in `let` and `mut` calls, so we now respect that as the authoritative type I also tried to add `def` input/output type inference, but unfortunately we only know the predecl types universally, which means we won't have enough information to properly know what the types of the custom commands are. # User-Facing Changes Script typechecking will get tighter in some cases Hovers should be more accurate in some cases that previously resorted to any. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: Darren Schroeder <343840+fdncred@users.noreply.github.com>
2023-06-28 19:19:48 +02:00
trace!("parsing: checking for keywords");
let name = working_set.get_span_contents(lite_command.parts[0]);
2021-09-13 21:59:11 +02:00
match name {
b"def" => parse_def(working_set, lite_command, None).0,
b"extern" => parse_extern(working_set, lite_command, None),
Let with pipeline (#9589) # Description This changes the default behaviour of `let` to be able to take a pipeline as its initial value. For example: ``` > let x = "hello world" | str length ``` This is a change from the existing behaviour, where the right hand side is assumed to be an expression. Pipelines are more general, and can be more powerful. My google foo is failing me, but this also fixes this issue: ``` let x = foo ``` Currently, this reads `foo` as a bareword that gets converted to a string rather than running the `foo` command. In practice, this is really annoying and is a really hard to spot bug in a script. # User-Facing Changes BREAKING CHANGE BREAKING CHANGE `let` gains the power to be assigned via a pipeline. However, this changes the behaviour of `let x = foo` from assigning the string "foo" to `$x` to being "run the command `foo` and give the result to `$x`" # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-03 07:45:10 +02:00
b"let" => parse_let(working_set, &lite_command.parts),
b"const" => parse_const(working_set, &lite_command.parts),
b"mut" => parse_mut(working_set, &lite_command.parts),
b"for" => {
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
let expr = parse_for(working_set, lite_command);
Pipeline::from_vec(vec![expr])
}
b"alias" => parse_alias(working_set, lite_command, None),
2023-05-06 22:55:10 +02:00
b"module" => parse_module(working_set, lite_command, None).0,
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
b"use" => parse_use(working_set, lite_command).0,
b"overlay" => {
if let Some(redirection) = lite_command.redirection.as_ref() {
working_set.error(redirecting_builtin_error("overlay", redirection));
return garbage_pipeline(working_set, &lite_command.parts);
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
}
parse_keyword(working_set, lite_command)
}
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
b"source" | b"source-env" => parse_source(working_set, lite_command),
b"export" => parse_export_in_block(working_set, lite_command),
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
b"hide" => parse_hide(working_set, lite_command),
b"where" => parse_where(working_set, lite_command),
2021-11-02 21:56:00 +01:00
#[cfg(feature = "plugin")]
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
b"register" => parse_register(working_set, lite_command),
Deprecate `register` and add `plugin use` (#12607) # Description Adds a new keyword, `plugin use`. Unlike `register`, this merely loads the signatures from the plugin cache file. The file is configurable with the `--plugin-config` option either to `nu` or to `plugin use` itself, just like the other `plugin` family of commands. At the REPL, one might do this to replace `register`: ```nushell > plugin add ~/.cargo/bin/nu_plugin_foo > plugin use foo ``` This will not work in a script, because `plugin use` is a keyword and `plugin add` does not evaluate at parse time (intentionally). This means we no longer run random binaries during parse. The `--plugins` option has been added to allow running `nu` with certain plugins in one step. This is used especially for the `nu_with_plugins!` test macro, but I'd imagine is generally useful. The only weird quirk is that it has to be a list, and we don't really do this for any of our other CLI args at the moment. `register` now prints a deprecation parse warning. This should fix #11923, as we now have a complete alternative to `register`. # User-Facing Changes - Add `plugin use` command - Deprecate `register` - Add `--plugins` option to `nu` to replace a common use of `register` # Tests + Formatting I think I've tested it thoroughly enough and every existing test passes. Testing nu CLI options and alternate config files is a little hairy and I wish there were some more generic helpers for this, so this will go on my TODO list for refactoring. - :green_circle: `toolkit fmt` - :green_circle: `toolkit clippy` - :green_circle: `toolkit test` - :green_circle: `toolkit test stdlib` # After Submitting - [ ] Update plugins sections of book - [ ] Release notes
2024-04-23 13:37:50 +02:00
// Only "plugin use" is a keyword
#[cfg(feature = "plugin")]
b"plugin"
if lite_command
.parts
.get(1)
.is_some_and(|span| working_set.get_span_contents(*span) == b"use") =>
{
if let Some(redirection) = lite_command.redirection.as_ref() {
working_set.error(redirecting_builtin_error("plugin use", redirection));
return garbage_pipeline(working_set, &lite_command.parts);
Deprecate `register` and add `plugin use` (#12607) # Description Adds a new keyword, `plugin use`. Unlike `register`, this merely loads the signatures from the plugin cache file. The file is configurable with the `--plugin-config` option either to `nu` or to `plugin use` itself, just like the other `plugin` family of commands. At the REPL, one might do this to replace `register`: ```nushell > plugin add ~/.cargo/bin/nu_plugin_foo > plugin use foo ``` This will not work in a script, because `plugin use` is a keyword and `plugin add` does not evaluate at parse time (intentionally). This means we no longer run random binaries during parse. The `--plugins` option has been added to allow running `nu` with certain plugins in one step. This is used especially for the `nu_with_plugins!` test macro, but I'd imagine is generally useful. The only weird quirk is that it has to be a list, and we don't really do this for any of our other CLI args at the moment. `register` now prints a deprecation parse warning. This should fix #11923, as we now have a complete alternative to `register`. # User-Facing Changes - Add `plugin use` command - Deprecate `register` - Add `--plugins` option to `nu` to replace a common use of `register` # Tests + Formatting I think I've tested it thoroughly enough and every existing test passes. Testing nu CLI options and alternate config files is a little hairy and I wish there were some more generic helpers for this, so this will go on my TODO list for refactoring. - :green_circle: `toolkit fmt` - :green_circle: `toolkit clippy` - :green_circle: `toolkit test` - :green_circle: `toolkit test stdlib` # After Submitting - [ ] Update plugins sections of book - [ ] Release notes
2024-04-23 13:37:50 +02:00
}
parse_keyword(working_set, lite_command)
}
2021-09-13 21:59:11 +02:00
_ => {
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
let element = parse_pipeline_element(working_set, lite_command);
2023-03-10 22:20:31 +01:00
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
Pipeline {
elements: vec![element],
}
}
}
2021-09-02 10:25:22 +02:00
}
2021-06-30 03:42:56 +02:00
pub fn parse_record(working_set: &mut StateWorkingSet, span: Span) -> Expression {
2021-11-11 00:14:00 +01:00
let bytes = working_set.get_span_contents(span);
let mut start = span.start;
let mut end = span.end;
if bytes.starts_with(b"{") {
start += 1;
} else {
working_set.error(ParseError::Expected("{", Span::new(start, start + 1)));
return garbage(working_set, span);
2021-11-11 00:14:00 +01:00
}
if bytes.ends_with(b"}") {
end -= 1;
} else {
working_set.error(ParseError::Unclosed("}".into(), Span::new(end, end)));
2021-11-11 00:14:00 +01:00
}
Protocol: debug_assert!() Span to reflect a valid slice (#6806) Also enforce this by #[non_exhaustive] span such that going forward we cannot, in debug builds (1), construct invalid spans. The motivation for this stems from #6431 where I've seen crashes due to invalid slice indexing. My hope is this will mitigate such senarios 1. https://github.com/nushell/nushell/pull/6431#issuecomment-1278147241 # Description (description of your pull request here) # Tests Make sure you've done the following: - [ ] Add tests that cover your changes, either in the command examples, the crate/tests folder, or in the /tests folder. - [ ] Try to think about corner cases and various ways how your changes could break. Cover them with tests. - [ ] If adding tests is not possible, please document in the PR body a minimal example with steps on how to reproduce so one can verify your change works. Make sure you've run and fixed any issues with these commands: - [x] `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - [ ] `cargo clippy --workspace --features=extra -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - [ ] `cargo test --workspace --features=extra` to check that all the tests pass # Documentation - [ ] If your PR touches a user-facing nushell feature then make sure that there is an entry in the documentation (https://github.com/nushell/nushell.github.io) for the feature, and update it if necessary.
2022-12-03 10:44:12 +01:00
let inner_span = Span::new(start, end);
2022-01-03 06:21:26 +01:00
let source = working_set.get_span_contents(inner_span);
2021-11-11 00:14:00 +01:00
2021-11-21 19:13:09 +01:00
let (tokens, err) = lex(source, start, &[b'\n', b'\r', b','], &[b':'], true);
if let Some(err) = err {
working_set.error(err);
}
2021-11-11 00:14:00 +01:00
let mut output = vec![];
let mut idx = 0;
let mut field_types = Some(vec![]);
2021-11-11 00:14:00 +01:00
while idx < tokens.len() {
Spread operator in record literals (#11144) Goes towards implementing #10598, which asks for a spread operator in lists, in records, and when calling commands (continuation of #11006, which only implements it in lists) # Description This PR is for adding a spread operator that can be used when building records. Additional functionality can be added later. Changes: - Previously, the `Expr::Record` variant held `(Expression, Expression)` pairs. It now holds instances of an enum `RecordItem` (the name isn't amazing) that allows either a key-value mapping or a spread operator. - `...` will be treated as the spread operator when it appears before `$`, `{`, or `(` inside records (no whitespace allowed in between) (not implemented yet) - The error message for duplicate columns now includes the column name itself, because if two spread records are involved in such an error, you can't tell which field was duplicated from the spans alone `...` will still be treated as a normal string outside records, and even in records, it is not treated as a spread operator when not followed immediately by a `$`, `{`, or `(`. # User-Facing Changes Users will be able to use `...` when building records. ``` > let rec = { x: 1, ...{ a: 2 } } > $rec ╭───┬───╮ │ x │ 1 │ │ a │ 2 │ ╰───┴───╯ > { foo: bar, ...$rec, baz: blah } ╭─────┬──────╮ │ foo │ bar │ │ x │ 1 │ │ a │ 2 │ │ baz │ blah │ ╰─────┴──────╯ ``` If you want to update a field of a record, you'll have to use `merge` instead: ``` > { ...$rec, x: 5 } Error: nu::shell::column_defined_twice × Record field or table column used twice: x ╭─[entry #2:1:1] 1 │ { ...$rec, x: 5 } · ──┬─ ┬ · │ ╰── field redefined here · ╰── field first defined here ╰──── > $rec | merge { x: 5 } ╭───┬───╮ │ x │ 5 │ │ a │ 2 │ ╰───┴───╯ ``` # Tests + Formatting # After Submitting
2023-11-29 18:31:31 +01:00
let curr_span = tokens[idx].span;
let curr_tok = working_set.get_span_contents(curr_span);
if curr_tok.starts_with(b"...")
&& curr_tok.len() > 3
&& (curr_tok[3] == b'$' || curr_tok[3] == b'{' || curr_tok[3] == b'(')
{
// Parse spread operator
let inner = parse_value(
working_set,
Span::new(curr_span.start + 3, curr_span.end),
&SyntaxShape::Record(vec![]),
);
idx += 1;
2021-11-11 00:14:00 +01:00
Spread operator in record literals (#11144) Goes towards implementing #10598, which asks for a spread operator in lists, in records, and when calling commands (continuation of #11006, which only implements it in lists) # Description This PR is for adding a spread operator that can be used when building records. Additional functionality can be added later. Changes: - Previously, the `Expr::Record` variant held `(Expression, Expression)` pairs. It now holds instances of an enum `RecordItem` (the name isn't amazing) that allows either a key-value mapping or a spread operator. - `...` will be treated as the spread operator when it appears before `$`, `{`, or `(` inside records (no whitespace allowed in between) (not implemented yet) - The error message for duplicate columns now includes the column name itself, because if two spread records are involved in such an error, you can't tell which field was duplicated from the spans alone `...` will still be treated as a normal string outside records, and even in records, it is not treated as a spread operator when not followed immediately by a `$`, `{`, or `(`. # User-Facing Changes Users will be able to use `...` when building records. ``` > let rec = { x: 1, ...{ a: 2 } } > $rec ╭───┬───╮ │ x │ 1 │ │ a │ 2 │ ╰───┴───╯ > { foo: bar, ...$rec, baz: blah } ╭─────┬──────╮ │ foo │ bar │ │ x │ 1 │ │ a │ 2 │ │ baz │ blah │ ╰─────┴──────╯ ``` If you want to update a field of a record, you'll have to use `merge` instead: ``` > { ...$rec, x: 5 } Error: nu::shell::column_defined_twice × Record field or table column used twice: x ╭─[entry #2:1:1] 1 │ { ...$rec, x: 5 } · ──┬─ ┬ · │ ╰── field redefined here · ╰── field first defined here ╰──── > $rec | merge { x: 5 } ╭───┬───╮ │ x │ 5 │ │ a │ 2 │ ╰───┴───╯ ``` # Tests + Formatting # After Submitting
2023-11-29 18:31:31 +01:00
match &inner.ty {
Type::Record(inner_fields) => {
if let Some(fields) = &mut field_types {
for (field, ty) in inner_fields.as_ref() {
Spread operator in record literals (#11144) Goes towards implementing #10598, which asks for a spread operator in lists, in records, and when calling commands (continuation of #11006, which only implements it in lists) # Description This PR is for adding a spread operator that can be used when building records. Additional functionality can be added later. Changes: - Previously, the `Expr::Record` variant held `(Expression, Expression)` pairs. It now holds instances of an enum `RecordItem` (the name isn't amazing) that allows either a key-value mapping or a spread operator. - `...` will be treated as the spread operator when it appears before `$`, `{`, or `(` inside records (no whitespace allowed in between) (not implemented yet) - The error message for duplicate columns now includes the column name itself, because if two spread records are involved in such an error, you can't tell which field was duplicated from the spans alone `...` will still be treated as a normal string outside records, and even in records, it is not treated as a spread operator when not followed immediately by a `$`, `{`, or `(`. # User-Facing Changes Users will be able to use `...` when building records. ``` > let rec = { x: 1, ...{ a: 2 } } > $rec ╭───┬───╮ │ x │ 1 │ │ a │ 2 │ ╰───┴───╯ > { foo: bar, ...$rec, baz: blah } ╭─────┬──────╮ │ foo │ bar │ │ x │ 1 │ │ a │ 2 │ │ baz │ blah │ ╰─────┴──────╯ ``` If you want to update a field of a record, you'll have to use `merge` instead: ``` > { ...$rec, x: 5 } Error: nu::shell::column_defined_twice × Record field or table column used twice: x ╭─[entry #2:1:1] 1 │ { ...$rec, x: 5 } · ──┬─ ┬ · │ ╰── field redefined here · ╰── field first defined here ╰──── > $rec | merge { x: 5 } ╭───┬───╮ │ x │ 5 │ │ a │ 2 │ ╰───┴───╯ ``` # Tests + Formatting # After Submitting
2023-11-29 18:31:31 +01:00
fields.push((field.clone(), ty.clone()));
}
}
}
_ => {
// We can't properly see all the field types
// so fall back to the Any type later
field_types = None;
}
}
Spread operator in record literals (#11144) Goes towards implementing #10598, which asks for a spread operator in lists, in records, and when calling commands (continuation of #11006, which only implements it in lists) # Description This PR is for adding a spread operator that can be used when building records. Additional functionality can be added later. Changes: - Previously, the `Expr::Record` variant held `(Expression, Expression)` pairs. It now holds instances of an enum `RecordItem` (the name isn't amazing) that allows either a key-value mapping or a spread operator. - `...` will be treated as the spread operator when it appears before `$`, `{`, or `(` inside records (no whitespace allowed in between) (not implemented yet) - The error message for duplicate columns now includes the column name itself, because if two spread records are involved in such an error, you can't tell which field was duplicated from the spans alone `...` will still be treated as a normal string outside records, and even in records, it is not treated as a spread operator when not followed immediately by a `$`, `{`, or `(`. # User-Facing Changes Users will be able to use `...` when building records. ``` > let rec = { x: 1, ...{ a: 2 } } > $rec ╭───┬───╮ │ x │ 1 │ │ a │ 2 │ ╰───┴───╯ > { foo: bar, ...$rec, baz: blah } ╭─────┬──────╮ │ foo │ bar │ │ x │ 1 │ │ a │ 2 │ │ baz │ blah │ ╰─────┴──────╯ ``` If you want to update a field of a record, you'll have to use `merge` instead: ``` > { ...$rec, x: 5 } Error: nu::shell::column_defined_twice × Record field or table column used twice: x ╭─[entry #2:1:1] 1 │ { ...$rec, x: 5 } · ──┬─ ┬ · │ ╰── field redefined here · ╰── field first defined here ╰──── > $rec | merge { x: 5 } ╭───┬───╮ │ x │ 5 │ │ a │ 2 │ ╰───┴───╯ ``` # Tests + Formatting # After Submitting
2023-11-29 18:31:31 +01:00
output.push(RecordItem::Spread(
Span::new(curr_span.start, curr_span.start + 3),
inner,
));
} else {
Spread operator in record literals (#11144) Goes towards implementing #10598, which asks for a spread operator in lists, in records, and when calling commands (continuation of #11006, which only implements it in lists) # Description This PR is for adding a spread operator that can be used when building records. Additional functionality can be added later. Changes: - Previously, the `Expr::Record` variant held `(Expression, Expression)` pairs. It now holds instances of an enum `RecordItem` (the name isn't amazing) that allows either a key-value mapping or a spread operator. - `...` will be treated as the spread operator when it appears before `$`, `{`, or `(` inside records (no whitespace allowed in between) (not implemented yet) - The error message for duplicate columns now includes the column name itself, because if two spread records are involved in such an error, you can't tell which field was duplicated from the spans alone `...` will still be treated as a normal string outside records, and even in records, it is not treated as a spread operator when not followed immediately by a `$`, `{`, or `(`. # User-Facing Changes Users will be able to use `...` when building records. ``` > let rec = { x: 1, ...{ a: 2 } } > $rec ╭───┬───╮ │ x │ 1 │ │ a │ 2 │ ╰───┴───╯ > { foo: bar, ...$rec, baz: blah } ╭─────┬──────╮ │ foo │ bar │ │ x │ 1 │ │ a │ 2 │ │ baz │ blah │ ╰─────┴──────╯ ``` If you want to update a field of a record, you'll have to use `merge` instead: ``` > { ...$rec, x: 5 } Error: nu::shell::column_defined_twice × Record field or table column used twice: x ╭─[entry #2:1:1] 1 │ { ...$rec, x: 5 } · ──┬─ ┬ · │ ╰── field redefined here · ╰── field first defined here ╰──── > $rec | merge { x: 5 } ╭───┬───╮ │ x │ 5 │ │ a │ 2 │ ╰───┴───╯ ``` # Tests + Formatting # After Submitting
2023-11-29 18:31:31 +01:00
// Normal key-value pair
let field = parse_value(working_set, curr_span, &SyntaxShape::Any);
idx += 1;
if idx == tokens.len() {
More specific errors for missing values in records (#11423) # Description Currently, when writing a record, if you don't give the value for a field, the syntax error highlights the entire record instead of pinpointing the issue. Here's some examples: ```nushell > { a: 2, 3 } # Missing colon (and value) Error: nu::parser::parse_mismatch × Parse mismatch during operation. ╭─[entry #2:1:1] 1 │ { a: 2, 3 } · ─────┬───── · ╰── expected record ╰──── > { a: 2, 3: } # Missing value Error: nu::parser::parse_mismatch × Parse mismatch during operation. ╭─[entry #3:1:1] 1 │ { a: 2, 3: } · ──────┬───── · ╰── expected record ╰──── > { a: 2, 3 4 } # Missing colon Error: nu::parser::parse_mismatch × Parse mismatch during operation. ╭─[entry #4:1:1] 1 │ { a: 2, 3 4 } · ──────┬────── · ╰── expected record ╰──── ``` In all of them, the entire record is highlighted red because an `Expr::Garbage` is returned covering that whole span: ![image](https://github.com/nushell/nushell/assets/45539777/36660b50-23be-4353-b180-3f84eff3c220) This PR is for highlighting only the part inside the record that could not be parsed. If the record literal is big, an error message pointing to the start of where the parser thinks things went wrong should help people fix their code. # User-Facing Changes Below are screenshots of the new errors: If there's a stray record key right before the record ends, it highlights only that key and tells the user it expected a colon after it: ![image](https://github.com/nushell/nushell/assets/45539777/94503256-8ea2-47dd-b69a-4b520c66f7b6) If the record ends before the value for the last field was given, it highlights the key and colon of that field and tells the user it expected a value after the colon: ![image](https://github.com/nushell/nushell/assets/45539777/2f3837ec-3b35-4b81-8c57-706f8056ac04) If there are two consecutive expressions without a colon between them, it highlights everything from the second expression to the end of the record and tells the user it expected a colon. I was tempted to add a help message suggesting adding a colon in between, but that may not always be the right thing to do. ![image](https://github.com/nushell/nushell/assets/45539777/1abaaaa8-1896-4909-bbb7-9a38cece5250) # Tests + Formatting # After Submitting
2023-12-27 10:15:12 +01:00
working_set.error(ParseError::Expected(
"':'",
Span::new(curr_span.end, curr_span.end),
));
output.push(RecordItem::Pair(
garbage(working_set, curr_span),
garbage(working_set, Span::new(curr_span.end, curr_span.end)),
More specific errors for missing values in records (#11423) # Description Currently, when writing a record, if you don't give the value for a field, the syntax error highlights the entire record instead of pinpointing the issue. Here's some examples: ```nushell > { a: 2, 3 } # Missing colon (and value) Error: nu::parser::parse_mismatch × Parse mismatch during operation. ╭─[entry #2:1:1] 1 │ { a: 2, 3 } · ─────┬───── · ╰── expected record ╰──── > { a: 2, 3: } # Missing value Error: nu::parser::parse_mismatch × Parse mismatch during operation. ╭─[entry #3:1:1] 1 │ { a: 2, 3: } · ──────┬───── · ╰── expected record ╰──── > { a: 2, 3 4 } # Missing colon Error: nu::parser::parse_mismatch × Parse mismatch during operation. ╭─[entry #4:1:1] 1 │ { a: 2, 3 4 } · ──────┬────── · ╰── expected record ╰──── ``` In all of them, the entire record is highlighted red because an `Expr::Garbage` is returned covering that whole span: ![image](https://github.com/nushell/nushell/assets/45539777/36660b50-23be-4353-b180-3f84eff3c220) This PR is for highlighting only the part inside the record that could not be parsed. If the record literal is big, an error message pointing to the start of where the parser thinks things went wrong should help people fix their code. # User-Facing Changes Below are screenshots of the new errors: If there's a stray record key right before the record ends, it highlights only that key and tells the user it expected a colon after it: ![image](https://github.com/nushell/nushell/assets/45539777/94503256-8ea2-47dd-b69a-4b520c66f7b6) If the record ends before the value for the last field was given, it highlights the key and colon of that field and tells the user it expected a value after the colon: ![image](https://github.com/nushell/nushell/assets/45539777/2f3837ec-3b35-4b81-8c57-706f8056ac04) If there are two consecutive expressions without a colon between them, it highlights everything from the second expression to the end of the record and tells the user it expected a colon. I was tempted to add a help message suggesting adding a colon in between, but that may not always be the right thing to do. ![image](https://github.com/nushell/nushell/assets/45539777/1abaaaa8-1896-4909-bbb7-9a38cece5250) # Tests + Formatting # After Submitting
2023-12-27 10:15:12 +01:00
));
break;
Spread operator in record literals (#11144) Goes towards implementing #10598, which asks for a spread operator in lists, in records, and when calling commands (continuation of #11006, which only implements it in lists) # Description This PR is for adding a spread operator that can be used when building records. Additional functionality can be added later. Changes: - Previously, the `Expr::Record` variant held `(Expression, Expression)` pairs. It now holds instances of an enum `RecordItem` (the name isn't amazing) that allows either a key-value mapping or a spread operator. - `...` will be treated as the spread operator when it appears before `$`, `{`, or `(` inside records (no whitespace allowed in between) (not implemented yet) - The error message for duplicate columns now includes the column name itself, because if two spread records are involved in such an error, you can't tell which field was duplicated from the spans alone `...` will still be treated as a normal string outside records, and even in records, it is not treated as a spread operator when not followed immediately by a `$`, `{`, or `(`. # User-Facing Changes Users will be able to use `...` when building records. ``` > let rec = { x: 1, ...{ a: 2 } } > $rec ╭───┬───╮ │ x │ 1 │ │ a │ 2 │ ╰───┴───╯ > { foo: bar, ...$rec, baz: blah } ╭─────┬──────╮ │ foo │ bar │ │ x │ 1 │ │ a │ 2 │ │ baz │ blah │ ╰─────┴──────╯ ``` If you want to update a field of a record, you'll have to use `merge` instead: ``` > { ...$rec, x: 5 } Error: nu::shell::column_defined_twice × Record field or table column used twice: x ╭─[entry #2:1:1] 1 │ { ...$rec, x: 5 } · ──┬─ ┬ · │ ╰── field redefined here · ╰── field first defined here ╰──── > $rec | merge { x: 5 } ╭───┬───╮ │ x │ 5 │ │ a │ 2 │ ╰───┴───╯ ``` # Tests + Formatting # After Submitting
2023-11-29 18:31:31 +01:00
}
More specific errors for missing values in records (#11423) # Description Currently, when writing a record, if you don't give the value for a field, the syntax error highlights the entire record instead of pinpointing the issue. Here's some examples: ```nushell > { a: 2, 3 } # Missing colon (and value) Error: nu::parser::parse_mismatch × Parse mismatch during operation. ╭─[entry #2:1:1] 1 │ { a: 2, 3 } · ─────┬───── · ╰── expected record ╰──── > { a: 2, 3: } # Missing value Error: nu::parser::parse_mismatch × Parse mismatch during operation. ╭─[entry #3:1:1] 1 │ { a: 2, 3: } · ──────┬───── · ╰── expected record ╰──── > { a: 2, 3 4 } # Missing colon Error: nu::parser::parse_mismatch × Parse mismatch during operation. ╭─[entry #4:1:1] 1 │ { a: 2, 3 4 } · ──────┬────── · ╰── expected record ╰──── ``` In all of them, the entire record is highlighted red because an `Expr::Garbage` is returned covering that whole span: ![image](https://github.com/nushell/nushell/assets/45539777/36660b50-23be-4353-b180-3f84eff3c220) This PR is for highlighting only the part inside the record that could not be parsed. If the record literal is big, an error message pointing to the start of where the parser thinks things went wrong should help people fix their code. # User-Facing Changes Below are screenshots of the new errors: If there's a stray record key right before the record ends, it highlights only that key and tells the user it expected a colon after it: ![image](https://github.com/nushell/nushell/assets/45539777/94503256-8ea2-47dd-b69a-4b520c66f7b6) If the record ends before the value for the last field was given, it highlights the key and colon of that field and tells the user it expected a value after the colon: ![image](https://github.com/nushell/nushell/assets/45539777/2f3837ec-3b35-4b81-8c57-706f8056ac04) If there are two consecutive expressions without a colon between them, it highlights everything from the second expression to the end of the record and tells the user it expected a colon. I was tempted to add a help message suggesting adding a colon in between, but that may not always be the right thing to do. ![image](https://github.com/nushell/nushell/assets/45539777/1abaaaa8-1896-4909-bbb7-9a38cece5250) # Tests + Formatting # After Submitting
2023-12-27 10:15:12 +01:00
let colon_span = tokens[idx].span;
let colon = working_set.get_span_contents(colon_span);
Spread operator in record literals (#11144) Goes towards implementing #10598, which asks for a spread operator in lists, in records, and when calling commands (continuation of #11006, which only implements it in lists) # Description This PR is for adding a spread operator that can be used when building records. Additional functionality can be added later. Changes: - Previously, the `Expr::Record` variant held `(Expression, Expression)` pairs. It now holds instances of an enum `RecordItem` (the name isn't amazing) that allows either a key-value mapping or a spread operator. - `...` will be treated as the spread operator when it appears before `$`, `{`, or `(` inside records (no whitespace allowed in between) (not implemented yet) - The error message for duplicate columns now includes the column name itself, because if two spread records are involved in such an error, you can't tell which field was duplicated from the spans alone `...` will still be treated as a normal string outside records, and even in records, it is not treated as a spread operator when not followed immediately by a `$`, `{`, or `(`. # User-Facing Changes Users will be able to use `...` when building records. ``` > let rec = { x: 1, ...{ a: 2 } } > $rec ╭───┬───╮ │ x │ 1 │ │ a │ 2 │ ╰───┴───╯ > { foo: bar, ...$rec, baz: blah } ╭─────┬──────╮ │ foo │ bar │ │ x │ 1 │ │ a │ 2 │ │ baz │ blah │ ╰─────┴──────╯ ``` If you want to update a field of a record, you'll have to use `merge` instead: ``` > { ...$rec, x: 5 } Error: nu::shell::column_defined_twice × Record field or table column used twice: x ╭─[entry #2:1:1] 1 │ { ...$rec, x: 5 } · ──┬─ ┬ · │ ╰── field redefined here · ╰── field first defined here ╰──── > $rec | merge { x: 5 } ╭───┬───╮ │ x │ 5 │ │ a │ 2 │ ╰───┴───╯ ``` # Tests + Formatting # After Submitting
2023-11-29 18:31:31 +01:00
idx += 1;
More specific errors for missing values in records (#11423) # Description Currently, when writing a record, if you don't give the value for a field, the syntax error highlights the entire record instead of pinpointing the issue. Here's some examples: ```nushell > { a: 2, 3 } # Missing colon (and value) Error: nu::parser::parse_mismatch × Parse mismatch during operation. ╭─[entry #2:1:1] 1 │ { a: 2, 3 } · ─────┬───── · ╰── expected record ╰──── > { a: 2, 3: } # Missing value Error: nu::parser::parse_mismatch × Parse mismatch during operation. ╭─[entry #3:1:1] 1 │ { a: 2, 3: } · ──────┬───── · ╰── expected record ╰──── > { a: 2, 3 4 } # Missing colon Error: nu::parser::parse_mismatch × Parse mismatch during operation. ╭─[entry #4:1:1] 1 │ { a: 2, 3 4 } · ──────┬────── · ╰── expected record ╰──── ``` In all of them, the entire record is highlighted red because an `Expr::Garbage` is returned covering that whole span: ![image](https://github.com/nushell/nushell/assets/45539777/36660b50-23be-4353-b180-3f84eff3c220) This PR is for highlighting only the part inside the record that could not be parsed. If the record literal is big, an error message pointing to the start of where the parser thinks things went wrong should help people fix their code. # User-Facing Changes Below are screenshots of the new errors: If there's a stray record key right before the record ends, it highlights only that key and tells the user it expected a colon after it: ![image](https://github.com/nushell/nushell/assets/45539777/94503256-8ea2-47dd-b69a-4b520c66f7b6) If the record ends before the value for the last field was given, it highlights the key and colon of that field and tells the user it expected a value after the colon: ![image](https://github.com/nushell/nushell/assets/45539777/2f3837ec-3b35-4b81-8c57-706f8056ac04) If there are two consecutive expressions without a colon between them, it highlights everything from the second expression to the end of the record and tells the user it expected a colon. I was tempted to add a help message suggesting adding a colon in between, but that may not always be the right thing to do. ![image](https://github.com/nushell/nushell/assets/45539777/1abaaaa8-1896-4909-bbb7-9a38cece5250) # Tests + Formatting # After Submitting
2023-12-27 10:15:12 +01:00
if colon != b":" {
working_set.error(ParseError::Expected(
"':'",
Span::new(colon_span.start, colon_span.start),
));
output.push(RecordItem::Pair(
field,
garbage(
working_set,
Span::new(colon_span.start, tokens[tokens.len() - 1].span.end),
),
More specific errors for missing values in records (#11423) # Description Currently, when writing a record, if you don't give the value for a field, the syntax error highlights the entire record instead of pinpointing the issue. Here's some examples: ```nushell > { a: 2, 3 } # Missing colon (and value) Error: nu::parser::parse_mismatch × Parse mismatch during operation. ╭─[entry #2:1:1] 1 │ { a: 2, 3 } · ─────┬───── · ╰── expected record ╰──── > { a: 2, 3: } # Missing value Error: nu::parser::parse_mismatch × Parse mismatch during operation. ╭─[entry #3:1:1] 1 │ { a: 2, 3: } · ──────┬───── · ╰── expected record ╰──── > { a: 2, 3 4 } # Missing colon Error: nu::parser::parse_mismatch × Parse mismatch during operation. ╭─[entry #4:1:1] 1 │ { a: 2, 3 4 } · ──────┬────── · ╰── expected record ╰──── ``` In all of them, the entire record is highlighted red because an `Expr::Garbage` is returned covering that whole span: ![image](https://github.com/nushell/nushell/assets/45539777/36660b50-23be-4353-b180-3f84eff3c220) This PR is for highlighting only the part inside the record that could not be parsed. If the record literal is big, an error message pointing to the start of where the parser thinks things went wrong should help people fix their code. # User-Facing Changes Below are screenshots of the new errors: If there's a stray record key right before the record ends, it highlights only that key and tells the user it expected a colon after it: ![image](https://github.com/nushell/nushell/assets/45539777/94503256-8ea2-47dd-b69a-4b520c66f7b6) If the record ends before the value for the last field was given, it highlights the key and colon of that field and tells the user it expected a value after the colon: ![image](https://github.com/nushell/nushell/assets/45539777/2f3837ec-3b35-4b81-8c57-706f8056ac04) If there are two consecutive expressions without a colon between them, it highlights everything from the second expression to the end of the record and tells the user it expected a colon. I was tempted to add a help message suggesting adding a colon in between, but that may not always be the right thing to do. ![image](https://github.com/nushell/nushell/assets/45539777/1abaaaa8-1896-4909-bbb7-9a38cece5250) # Tests + Formatting # After Submitting
2023-12-27 10:15:12 +01:00
));
break;
}
if idx == tokens.len() {
working_set.error(ParseError::Expected(
"value for record field",
Span::new(colon_span.end, colon_span.end),
));
output.push(RecordItem::Pair(
garbage(working_set, Span::new(curr_span.start, colon_span.end)),
garbage(
working_set,
Span::new(colon_span.end, tokens[tokens.len() - 1].span.end),
),
More specific errors for missing values in records (#11423) # Description Currently, when writing a record, if you don't give the value for a field, the syntax error highlights the entire record instead of pinpointing the issue. Here's some examples: ```nushell > { a: 2, 3 } # Missing colon (and value) Error: nu::parser::parse_mismatch × Parse mismatch during operation. ╭─[entry #2:1:1] 1 │ { a: 2, 3 } · ─────┬───── · ╰── expected record ╰──── > { a: 2, 3: } # Missing value Error: nu::parser::parse_mismatch × Parse mismatch during operation. ╭─[entry #3:1:1] 1 │ { a: 2, 3: } · ──────┬───── · ╰── expected record ╰──── > { a: 2, 3 4 } # Missing colon Error: nu::parser::parse_mismatch × Parse mismatch during operation. ╭─[entry #4:1:1] 1 │ { a: 2, 3 4 } · ──────┬────── · ╰── expected record ╰──── ``` In all of them, the entire record is highlighted red because an `Expr::Garbage` is returned covering that whole span: ![image](https://github.com/nushell/nushell/assets/45539777/36660b50-23be-4353-b180-3f84eff3c220) This PR is for highlighting only the part inside the record that could not be parsed. If the record literal is big, an error message pointing to the start of where the parser thinks things went wrong should help people fix their code. # User-Facing Changes Below are screenshots of the new errors: If there's a stray record key right before the record ends, it highlights only that key and tells the user it expected a colon after it: ![image](https://github.com/nushell/nushell/assets/45539777/94503256-8ea2-47dd-b69a-4b520c66f7b6) If the record ends before the value for the last field was given, it highlights the key and colon of that field and tells the user it expected a value after the colon: ![image](https://github.com/nushell/nushell/assets/45539777/2f3837ec-3b35-4b81-8c57-706f8056ac04) If there are two consecutive expressions without a colon between them, it highlights everything from the second expression to the end of the record and tells the user it expected a colon. I was tempted to add a help message suggesting adding a colon in between, but that may not always be the right thing to do. ![image](https://github.com/nushell/nushell/assets/45539777/1abaaaa8-1896-4909-bbb7-9a38cece5250) # Tests + Formatting # After Submitting
2023-12-27 10:15:12 +01:00
));
break;
Spread operator in record literals (#11144) Goes towards implementing #10598, which asks for a spread operator in lists, in records, and when calling commands (continuation of #11006, which only implements it in lists) # Description This PR is for adding a spread operator that can be used when building records. Additional functionality can be added later. Changes: - Previously, the `Expr::Record` variant held `(Expression, Expression)` pairs. It now holds instances of an enum `RecordItem` (the name isn't amazing) that allows either a key-value mapping or a spread operator. - `...` will be treated as the spread operator when it appears before `$`, `{`, or `(` inside records (no whitespace allowed in between) (not implemented yet) - The error message for duplicate columns now includes the column name itself, because if two spread records are involved in such an error, you can't tell which field was duplicated from the spans alone `...` will still be treated as a normal string outside records, and even in records, it is not treated as a spread operator when not followed immediately by a `$`, `{`, or `(`. # User-Facing Changes Users will be able to use `...` when building records. ``` > let rec = { x: 1, ...{ a: 2 } } > $rec ╭───┬───╮ │ x │ 1 │ │ a │ 2 │ ╰───┴───╯ > { foo: bar, ...$rec, baz: blah } ╭─────┬──────╮ │ foo │ bar │ │ x │ 1 │ │ a │ 2 │ │ baz │ blah │ ╰─────┴──────╯ ``` If you want to update a field of a record, you'll have to use `merge` instead: ``` > { ...$rec, x: 5 } Error: nu::shell::column_defined_twice × Record field or table column used twice: x ╭─[entry #2:1:1] 1 │ { ...$rec, x: 5 } · ──┬─ ┬ · │ ╰── field redefined here · ╰── field first defined here ╰──── > $rec | merge { x: 5 } ╭───┬───╮ │ x │ 5 │ │ a │ 2 │ ╰───┴───╯ ``` # Tests + Formatting # After Submitting
2023-11-29 18:31:31 +01:00
}
let value = parse_value(working_set, tokens[idx].span, &SyntaxShape::Any);
idx += 1;
if let Some(field) = field.as_string() {
if let Some(fields) = &mut field_types {
fields.push((field, value.ty.clone()));
}
} else {
// We can't properly see all the field types
// so fall back to the Any type later
field_types = None;
}
output.push(RecordItem::Pair(field, value));
}
2021-11-11 00:14:00 +01:00
}
Expression::new(
working_set,
Expr::Record(output),
span,
if let Some(fields) = field_types {
Type::Record(fields.into())
} else {
Type::Any
},
)
2021-11-11 00:14:00 +01:00
}
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
fn parse_redirection_target(
working_set: &mut StateWorkingSet,
target: &LiteRedirectionTarget,
) -> RedirectionTarget {
match target {
LiteRedirectionTarget::File {
connector,
file,
append,
} => RedirectionTarget::File {
expr: parse_value(working_set, *file, &SyntaxShape::Any),
append: *append,
span: *connector,
},
LiteRedirectionTarget::Pipe { connector } => RedirectionTarget::Pipe { span: *connector },
}
}
pub(crate) fn parse_redirection(
working_set: &mut StateWorkingSet,
target: &LiteRedirection,
) -> PipelineRedirection {
match target {
LiteRedirection::Single { source, target } => PipelineRedirection::Single {
source: *source,
target: parse_redirection_target(working_set, target),
},
LiteRedirection::Separate { out, err } => PipelineRedirection::Separate {
out: parse_redirection_target(working_set, out),
err: parse_redirection_target(working_set, err),
},
}
}
fn parse_pipeline_element(
working_set: &mut StateWorkingSet,
command: &LiteCommand,
) -> PipelineElement {
trace!("parsing: pipeline element");
let expr = parse_expression(working_set, &command.parts);
let redirection = command
.redirection
.as_ref()
.map(|r| parse_redirection(working_set, r));
PipelineElement {
pipe: command.pipe,
expr,
redirection,
}
}
pub(crate) fn redirecting_builtin_error(
name: &'static str,
redirection: &LiteRedirection,
) -> ParseError {
match redirection {
LiteRedirection::Single { target, .. } => {
ParseError::RedirectingBuiltinCommand(name, target.connector(), None)
}
LiteRedirection::Separate { out, err } => ParseError::RedirectingBuiltinCommand(
name,
out.connector().min(err.connector()),
Some(out.connector().max(err.connector())),
),
}
}
Let with pipeline (#9589) # Description This changes the default behaviour of `let` to be able to take a pipeline as its initial value. For example: ``` > let x = "hello world" | str length ``` This is a change from the existing behaviour, where the right hand side is assumed to be an expression. Pipelines are more general, and can be more powerful. My google foo is failing me, but this also fixes this issue: ``` let x = foo ``` Currently, this reads `foo` as a bareword that gets converted to a string rather than running the `foo` command. In practice, this is really annoying and is a really hard to spot bug in a script. # User-Facing Changes BREAKING CHANGE BREAKING CHANGE `let` gains the power to be assigned via a pipeline. However, this changes the behaviour of `let x = foo` from assigning the string "foo" to `$x` to being "run the command `foo` and give the result to `$x`" # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-03 07:45:10 +02:00
pub fn parse_pipeline(
2021-09-02 10:25:22 +02:00
working_set: &mut StateWorkingSet,
Let with pipeline (#9589) # Description This changes the default behaviour of `let` to be able to take a pipeline as its initial value. For example: ``` > let x = "hello world" | str length ``` This is a change from the existing behaviour, where the right hand side is assumed to be an expression. Pipelines are more general, and can be more powerful. My google foo is failing me, but this also fixes this issue: ``` let x = foo ``` Currently, this reads `foo` as a bareword that gets converted to a string rather than running the `foo` command. In practice, this is really annoying and is a really hard to spot bug in a script. # User-Facing Changes BREAKING CHANGE BREAKING CHANGE `let` gains the power to be assigned via a pipeline. However, this changes the behaviour of `let x = foo` from assigning the string "foo" to `$x` to being "run the command `foo` and give the result to `$x`" # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-03 07:45:10 +02:00
pipeline: &LitePipeline,
is_subexpression: bool,
Let with pipeline (#9589) # Description This changes the default behaviour of `let` to be able to take a pipeline as its initial value. For example: ``` > let x = "hello world" | str length ``` This is a change from the existing behaviour, where the right hand side is assumed to be an expression. Pipelines are more general, and can be more powerful. My google foo is failing me, but this also fixes this issue: ``` let x = foo ``` Currently, this reads `foo` as a bareword that gets converted to a string rather than running the `foo` command. In practice, this is really annoying and is a really hard to spot bug in a script. # User-Facing Changes BREAKING CHANGE BREAKING CHANGE `let` gains the power to be assigned via a pipeline. However, this changes the behaviour of `let x = foo` from assigning the string "foo" to `$x` to being "run the command `foo` and give the result to `$x`" # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-03 07:45:10 +02:00
pipeline_index: usize,
) -> Pipeline {
if pipeline.commands.len() > 1 {
// Special case: allow `let` and `mut` to consume the whole pipeline, eg) `let abc = "foo" | str length`
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
if let Some(&first) = pipeline.commands[0].parts.first() {
let first = working_set.get_span_contents(first);
if first == b"let" || first == b"mut" {
let name = if first == b"let" { "let" } else { "mut" };
let mut new_command = LiteCommand {
comments: vec![],
parts: pipeline.commands[0].parts.clone(),
pipe: None,
redirection: None,
};
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
if let Some(redirection) = pipeline.commands[0].redirection.as_ref() {
working_set.error(redirecting_builtin_error(name, redirection));
}
2022-01-01 22:42:50 +01:00
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
for element in &pipeline.commands[1..] {
if let Some(redirection) = pipeline.commands[0].redirection.as_ref() {
working_set.error(redirecting_builtin_error(name, redirection));
} else {
new_command.parts.push(element.pipe.expect("pipe span"));
new_command.comments.extend_from_slice(&element.comments);
new_command.parts.extend_from_slice(&element.parts);
Let with pipeline (#9589) # Description This changes the default behaviour of `let` to be able to take a pipeline as its initial value. For example: ``` > let x = "hello world" | str length ``` This is a change from the existing behaviour, where the right hand side is assumed to be an expression. Pipelines are more general, and can be more powerful. My google foo is failing me, but this also fixes this issue: ``` let x = foo ``` Currently, this reads `foo` as a bareword that gets converted to a string rather than running the `foo` command. In practice, this is really annoying and is a really hard to spot bug in a script. # User-Facing Changes BREAKING CHANGE BREAKING CHANGE `let` gains the power to be assigned via a pipeline. However, this changes the behaviour of `let x = foo` from assigning the string "foo" to `$x` to being "run the command `foo` and give the result to `$x`" # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-03 07:45:10 +02:00
}
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
}
2021-06-30 03:42:56 +02:00
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
// if the 'let' is complete enough, use it, if not, fall through for now
if new_command.parts.len() > 3 {
let rhs_span = Span::concat(&new_command.parts[3..]);
2021-07-01 02:01:04 +02:00
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
new_command.parts.truncate(3);
new_command.parts.push(rhs_span);
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
let mut pipeline = parse_builtin_commands(working_set, &new_command);
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
if pipeline_index == 0 {
let let_decl_id = working_set.find_decl(b"let");
let mut_decl_id = working_set.find_decl(b"mut");
for element in pipeline.elements.iter_mut() {
if let Expr::Call(call) = &element.expr.expr {
if Some(call.decl_id) == let_decl_id
|| Some(call.decl_id) == mut_decl_id
{
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
// Do an expansion
if let Some(Expression {
expr: Expr::Block(block_id),
..
}) = call.positional_iter().nth(1)
Let with pipeline (#9589) # Description This changes the default behaviour of `let` to be able to take a pipeline as its initial value. For example: ``` > let x = "hello world" | str length ``` This is a change from the existing behaviour, where the right hand side is assumed to be an expression. Pipelines are more general, and can be more powerful. My google foo is failing me, but this also fixes this issue: ``` let x = foo ``` Currently, this reads `foo` as a bareword that gets converted to a string rather than running the `foo` command. In practice, this is really annoying and is a really hard to spot bug in a script. # User-Facing Changes BREAKING CHANGE BREAKING CHANGE `let` gains the power to be assigned via a pipeline. However, this changes the behaviour of `let x = foo` from assigning the string "foo" to `$x` to being "run the command `foo` and give the result to `$x`" # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-03 07:45:10 +02:00
{
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
let block = working_set.get_block(*block_id);
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
if let Some(element) = block
.pipelines
.first()
.and_then(|p| p.elements.first())
.cloned()
{
if element.has_in_variable(working_set) {
let element = wrap_element_with_collect(
working_set,
&element,
);
let block = working_set.get_block_mut(*block_id);
block.pipelines[0].elements[0] = element;
Let with pipeline (#9589) # Description This changes the default behaviour of `let` to be able to take a pipeline as its initial value. For example: ``` > let x = "hello world" | str length ``` This is a change from the existing behaviour, where the right hand side is assumed to be an expression. Pipelines are more general, and can be more powerful. My google foo is failing me, but this also fixes this issue: ``` let x = foo ``` Currently, this reads `foo` as a bareword that gets converted to a string rather than running the `foo` command. In practice, this is really annoying and is a really hard to spot bug in a script. # User-Facing Changes BREAKING CHANGE BREAKING CHANGE `let` gains the power to be assigned via a pipeline. However, this changes the behaviour of `let x = foo` from assigning the string "foo" to `$x` to being "run the command `foo` and give the result to `$x`" # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-03 07:45:10 +02:00
}
}
}
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
continue;
} else if element.has_in_variable(working_set) && !is_subexpression
{
*element = wrap_element_with_collect(working_set, element);
Let with pipeline (#9589) # Description This changes the default behaviour of `let` to be able to take a pipeline as its initial value. For example: ``` > let x = "hello world" | str length ``` This is a change from the existing behaviour, where the right hand side is assumed to be an expression. Pipelines are more general, and can be more powerful. My google foo is failing me, but this also fixes this issue: ``` let x = foo ``` Currently, this reads `foo` as a bareword that gets converted to a string rather than running the `foo` command. In practice, this is really annoying and is a really hard to spot bug in a script. # User-Facing Changes BREAKING CHANGE BREAKING CHANGE `let` gains the power to be assigned via a pipeline. However, this changes the behaviour of `let x = foo` from assigning the string "foo" to `$x` to being "run the command `foo` and give the result to `$x`" # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-03 07:45:10 +02:00
}
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
} else if element.has_in_variable(working_set) && !is_subexpression {
*element = wrap_element_with_collect(working_set, element);
Let with pipeline (#9589) # Description This changes the default behaviour of `let` to be able to take a pipeline as its initial value. For example: ``` > let x = "hello world" | str length ``` This is a change from the existing behaviour, where the right hand side is assumed to be an expression. Pipelines are more general, and can be more powerful. My google foo is failing me, but this also fixes this issue: ``` let x = foo ``` Currently, this reads `foo` as a bareword that gets converted to a string rather than running the `foo` command. In practice, this is really annoying and is a really hard to spot bug in a script. # User-Facing Changes BREAKING CHANGE BREAKING CHANGE `let` gains the power to be assigned via a pipeline. However, this changes the behaviour of `let x = foo` from assigning the string "foo" to `$x` to being "run the command `foo` and give the result to `$x`" # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-03 07:45:10 +02:00
}
}
}
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
return pipeline;
Let with pipeline (#9589) # Description This changes the default behaviour of `let` to be able to take a pipeline as its initial value. For example: ``` > let x = "hello world" | str length ``` This is a change from the existing behaviour, where the right hand side is assumed to be an expression. Pipelines are more general, and can be more powerful. My google foo is failing me, but this also fixes this issue: ``` let x = foo ``` Currently, this reads `foo` as a bareword that gets converted to a string rather than running the `foo` command. In practice, this is really annoying and is a really hard to spot bug in a script. # User-Facing Changes BREAKING CHANGE BREAKING CHANGE `let` gains the power to be assigned via a pipeline. However, this changes the behaviour of `let x = foo` from assigning the string "foo" to `$x` to being "run the command `foo` and give the result to `$x`" # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-03 07:45:10 +02:00
}
}
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
}
Let with pipeline (#9589) # Description This changes the default behaviour of `let` to be able to take a pipeline as its initial value. For example: ``` > let x = "hello world" | str length ``` This is a change from the existing behaviour, where the right hand side is assumed to be an expression. Pipelines are more general, and can be more powerful. My google foo is failing me, but this also fixes this issue: ``` let x = foo ``` Currently, this reads `foo` as a bareword that gets converted to a string rather than running the `foo` command. In practice, this is really annoying and is a really hard to spot bug in a script. # User-Facing Changes BREAKING CHANGE BREAKING CHANGE `let` gains the power to be assigned via a pipeline. However, this changes the behaviour of `let x = foo` from assigning the string "foo" to `$x` to being "run the command `foo` and give the result to `$x`" # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-03 07:45:10 +02:00
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
let mut elements = pipeline
Let with pipeline (#9589) # Description This changes the default behaviour of `let` to be able to take a pipeline as its initial value. For example: ``` > let x = "hello world" | str length ``` This is a change from the existing behaviour, where the right hand side is assumed to be an expression. Pipelines are more general, and can be more powerful. My google foo is failing me, but this also fixes this issue: ``` let x = foo ``` Currently, this reads `foo` as a bareword that gets converted to a string rather than running the `foo` command. In practice, this is really annoying and is a really hard to spot bug in a script. # User-Facing Changes BREAKING CHANGE BREAKING CHANGE `let` gains the power to be assigned via a pipeline. However, this changes the behaviour of `let x = foo` from assigning the string "foo" to `$x` to being "run the command `foo` and give the result to `$x`" # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-03 07:45:10 +02:00
.commands
.iter()
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
.map(|element| parse_pipeline_element(working_set, element))
.collect::<Vec<_>>();
Let with pipeline (#9589) # Description This changes the default behaviour of `let` to be able to take a pipeline as its initial value. For example: ``` > let x = "hello world" | str length ``` This is a change from the existing behaviour, where the right hand side is assumed to be an expression. Pipelines are more general, and can be more powerful. My google foo is failing me, but this also fixes this issue: ``` let x = foo ``` Currently, this reads `foo` as a bareword that gets converted to a string rather than running the `foo` command. In practice, this is really annoying and is a really hard to spot bug in a script. # User-Facing Changes BREAKING CHANGE BREAKING CHANGE `let` gains the power to be assigned via a pipeline. However, this changes the behaviour of `let x = foo` from assigning the string "foo" to `$x` to being "run the command `foo` and give the result to `$x`" # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-03 07:45:10 +02:00
if is_subexpression {
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
for element in elements.iter_mut().skip(1) {
Let with pipeline (#9589) # Description This changes the default behaviour of `let` to be able to take a pipeline as its initial value. For example: ``` > let x = "hello world" | str length ``` This is a change from the existing behaviour, where the right hand side is assumed to be an expression. Pipelines are more general, and can be more powerful. My google foo is failing me, but this also fixes this issue: ``` let x = foo ``` Currently, this reads `foo` as a bareword that gets converted to a string rather than running the `foo` command. In practice, this is really annoying and is a really hard to spot bug in a script. # User-Facing Changes BREAKING CHANGE BREAKING CHANGE `let` gains the power to be assigned via a pipeline. However, this changes the behaviour of `let x = foo` from assigning the string "foo" to `$x` to being "run the command `foo` and give the result to `$x`" # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-03 07:45:10 +02:00
if element.has_in_variable(working_set) {
*element = wrap_element_with_collect(working_set, element);
}
}
} else {
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
for element in elements.iter_mut() {
Let with pipeline (#9589) # Description This changes the default behaviour of `let` to be able to take a pipeline as its initial value. For example: ``` > let x = "hello world" | str length ``` This is a change from the existing behaviour, where the right hand side is assumed to be an expression. Pipelines are more general, and can be more powerful. My google foo is failing me, but this also fixes this issue: ``` let x = foo ``` Currently, this reads `foo` as a bareword that gets converted to a string rather than running the `foo` command. In practice, this is really annoying and is a really hard to spot bug in a script. # User-Facing Changes BREAKING CHANGE BREAKING CHANGE `let` gains the power to be assigned via a pipeline. However, this changes the behaviour of `let x = foo` from assigning the string "foo" to `$x` to being "run the command `foo` and give the result to `$x`" # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-03 07:45:10 +02:00
if element.has_in_variable(working_set) {
*element = wrap_element_with_collect(working_set, element);
}
}
}
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
Pipeline { elements }
Let with pipeline (#9589) # Description This changes the default behaviour of `let` to be able to take a pipeline as its initial value. For example: ``` > let x = "hello world" | str length ``` This is a change from the existing behaviour, where the right hand side is assumed to be an expression. Pipelines are more general, and can be more powerful. My google foo is failing me, but this also fixes this issue: ``` let x = foo ``` Currently, this reads `foo` as a bareword that gets converted to a string rather than running the `foo` command. In practice, this is really annoying and is a really hard to spot bug in a script. # User-Facing Changes BREAKING CHANGE BREAKING CHANGE `let` gains the power to be assigned via a pipeline. However, this changes the behaviour of `let x = foo` from assigning the string "foo" to `$x` to being "run the command `foo` and give the result to `$x`" # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-03 07:45:10 +02:00
} else {
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
if let Some(&first) = pipeline.commands[0].parts.first() {
let first = working_set.get_span_contents(first);
if first == b"let" || first == b"mut" {
if let Some(redirection) = pipeline.commands[0].redirection.as_ref() {
let name = if first == b"let" { "let" } else { "mut" };
working_set.error(redirecting_builtin_error(name, redirection));
}
}
}
Let with pipeline (#9589) # Description This changes the default behaviour of `let` to be able to take a pipeline as its initial value. For example: ``` > let x = "hello world" | str length ``` This is a change from the existing behaviour, where the right hand side is assumed to be an expression. Pipelines are more general, and can be more powerful. My google foo is failing me, but this also fixes this issue: ``` let x = foo ``` Currently, this reads `foo` as a bareword that gets converted to a string rather than running the `foo` command. In practice, this is really annoying and is a really hard to spot bug in a script. # User-Facing Changes BREAKING CHANGE BREAKING CHANGE `let` gains the power to be assigned via a pipeline. However, this changes the behaviour of `let x = foo` from assigning the string "foo" to `$x` to being "run the command `foo` and give the result to `$x`" # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-03 07:45:10 +02:00
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
let mut pipeline = parse_builtin_commands(working_set, &pipeline.commands[0]);
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
let let_decl_id = working_set.find_decl(b"let");
let mut_decl_id = working_set.find_decl(b"mut");
if pipeline_index == 0 {
for element in pipeline.elements.iter_mut() {
if let Expr::Call(call) = &element.expr.expr {
if Some(call.decl_id) == let_decl_id || Some(call.decl_id) == mut_decl_id {
// Do an expansion
if let Some(Expression {
expr: Expr::Block(block_id),
..
}) = call.positional_iter().nth(1)
{
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
let block = working_set.get_block(*block_id);
Let with pipeline (#9589) # Description This changes the default behaviour of `let` to be able to take a pipeline as its initial value. For example: ``` > let x = "hello world" | str length ``` This is a change from the existing behaviour, where the right hand side is assumed to be an expression. Pipelines are more general, and can be more powerful. My google foo is failing me, but this also fixes this issue: ``` let x = foo ``` Currently, this reads `foo` as a bareword that gets converted to a string rather than running the `foo` command. In practice, this is really annoying and is a really hard to spot bug in a script. # User-Facing Changes BREAKING CHANGE BREAKING CHANGE `let` gains the power to be assigned via a pipeline. However, this changes the behaviour of `let x = foo` from assigning the string "foo" to `$x` to being "run the command `foo` and give the result to `$x`" # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-03 07:45:10 +02:00
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
if let Some(element) = block
.pipelines
.first()
.and_then(|p| p.elements.first())
.cloned()
{
if element.has_in_variable(working_set) {
let element = wrap_element_with_collect(working_set, &element);
let block = working_set.get_block_mut(*block_id);
block.pipelines[0].elements[0] = element;
}
}
}
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
continue;
} else if element.has_in_variable(working_set) && !is_subexpression {
*element = wrap_element_with_collect(working_set, element);
}
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
} else if element.has_in_variable(working_set) && !is_subexpression {
*element = wrap_element_with_collect(working_set, element);
Let with pipeline (#9589) # Description This changes the default behaviour of `let` to be able to take a pipeline as its initial value. For example: ``` > let x = "hello world" | str length ``` This is a change from the existing behaviour, where the right hand side is assumed to be an expression. Pipelines are more general, and can be more powerful. My google foo is failing me, but this also fixes this issue: ``` let x = foo ``` Currently, this reads `foo` as a bareword that gets converted to a string rather than running the `foo` command. In practice, this is really annoying and is a really hard to spot bug in a script. # User-Facing Changes BREAKING CHANGE BREAKING CHANGE `let` gains the power to be assigned via a pipeline. However, this changes the behaviour of `let x = foo` from assigning the string "foo" to `$x` to being "run the command `foo` and give the result to `$x`" # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-03 07:45:10 +02:00
}
}
}
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
pipeline
Let with pipeline (#9589) # Description This changes the default behaviour of `let` to be able to take a pipeline as its initial value. For example: ``` > let x = "hello world" | str length ``` This is a change from the existing behaviour, where the right hand side is assumed to be an expression. Pipelines are more general, and can be more powerful. My google foo is failing me, but this also fixes this issue: ``` let x = foo ``` Currently, this reads `foo` as a bareword that gets converted to a string rather than running the `foo` command. In practice, this is really annoying and is a really hard to spot bug in a script. # User-Facing Changes BREAKING CHANGE BREAKING CHANGE `let` gains the power to be assigned via a pipeline. However, this changes the behaviour of `let x = foo` from assigning the string "foo" to `$x` to being "run the command `foo` and give the result to `$x`" # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-03 07:45:10 +02:00
}
}
pub fn parse_block(
working_set: &mut StateWorkingSet,
tokens: &[Token],
span: Span,
scoped: bool,
is_subexpression: bool,
) -> Block {
let (lite_block, err) = lite_parse(tokens);
if let Some(err) = err {
working_set.error(err);
}
trace!("parsing block: {:?}", lite_block);
if scoped {
working_set.enter_scope();
}
// Pre-declare any definition so that definitions
// that share the same block can see each other
for pipeline in &lite_block.block {
if pipeline.commands.len() == 1 {
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
parse_def_predecl(working_set, &pipeline.commands[0].parts)
}
}
2021-06-30 03:42:56 +02:00
Let with pipeline (#9589) # Description This changes the default behaviour of `let` to be able to take a pipeline as its initial value. For example: ``` > let x = "hello world" | str length ``` This is a change from the existing behaviour, where the right hand side is assumed to be an expression. Pipelines are more general, and can be more powerful. My google foo is failing me, but this also fixes this issue: ``` let x = foo ``` Currently, this reads `foo` as a bareword that gets converted to a string rather than running the `foo` command. In practice, this is really annoying and is a really hard to spot bug in a script. # User-Facing Changes BREAKING CHANGE BREAKING CHANGE `let` gains the power to be assigned via a pipeline. However, this changes the behaviour of `let x = foo` from assigning the string "foo" to `$x` to being "run the command `foo` and give the result to `$x`" # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-03 07:45:10 +02:00
let mut block = Block::new_with_capacity(lite_block.block.len());
for (idx, lite_pipeline) in lite_block.block.iter().enumerate() {
let pipeline = parse_pipeline(working_set, lite_pipeline, is_subexpression, idx);
block.pipelines.push(pipeline);
}
2021-09-02 10:25:22 +02:00
if scoped {
working_set.exit_scope();
2021-06-30 03:42:56 +02:00
}
Reuse the cached parse results of parsed files (#8949) # Description This does a lookup in the cache of parsed files to see if a span can be found for a file that was previously loaded with the same contents, then uses that span to find the parsed block for that file. The end result should, in theory, be identical but doesn't require any reparsing or creating new blocks/new definitions that aren't needed. This drops the sg.nu benchmark from: ``` ╭───┬───────────────────╮ │ 0 │ 280ms 606µs 208ns │ │ 1 │ 282ms 654µs 416ns │ │ 2 │ 252ms 640µs 541ns │ │ 3 │ 250ms 940µs 41ns │ │ 4 │ 241ms 216µs 375ns │ │ 5 │ 257ms 310µs 583ns │ │ 6 │ 196ms 739µs 416ns │ ╰───┴───────────────────╯ ``` to: ``` ╭───┬───────────────────╮ │ 0 │ 118ms 698µs 125ns │ │ 1 │ 121ms 327µs │ │ 2 │ 121ms 873µs 500ns │ │ 3 │ 124ms 94µs 708ns │ │ 4 │ 113ms 733µs 291ns │ │ 5 │ 108ms 663µs 125ns │ │ 6 │ 63ms 482µs 625ns │ ╰───┴───────────────────╯ ``` I was hoping to also see some startup time improvements, but I didn't notice much there. # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-04-21 21:00:33 +02:00
block.span = Some(span);
Custom command input/output types (#9690) # Description This adds input/output types to custom commands. These are input/output pairs that related an input type to an output type. For example (a single int-to-int input/output pair): ``` def foo []: int -> int { ... } ``` You can also have multiple input/output pairs: ``` def bar []: [int -> string, string -> list<string>] { ... } ``` These types are checked during definition time in the parser. If the block does not match the type, the user will get a parser error. This `:` to begin the input/output signatures should immediately follow the argument signature as shown above. The PR also improves type parsing by re-using the shape parser. The shape parser is now the canonical way to parse types/shapes in user code. This PR also splits `extern` into `extern`/`extern-wrapped` because of the parser limitation that a multi-span argument (which Signature now is) can't precede an optional argument. `extern-wrapped` now takes the required block that was previously optional. # User-Facing Changes The change to `extern` to split into `extern` and `extern-wrapped` is a breaking change. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-14 23:51:28 +02:00
let errors = type_check::check_block_input_output(working_set, &block);
if !errors.is_empty() {
working_set.parse_errors.extend_from_slice(&errors);
}
Input output checking (#9680) # Description This PR tights input/output type-checking a bit more. There are a lot of commands that don't have correct input/output types, so part of the effort is updating them. This PR now contains updates to commands that had wrong input/output signatures. It doesn't add examples for these new signatures, but that can be follow-up work. # User-Facing Changes BREAKING CHANGE BREAKING CHANGE This work enforces many more checks on pipeline type correctness than previous nushell versions. This strictness may uncover incompatibilities in existing scripts or shortcomings in the type information for internal commands. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-14 05:20:35 +02:00
block
2021-09-02 10:25:22 +02:00
}
2021-06-30 03:42:56 +02:00
pub fn discover_captures_in_closure(
2021-10-25 22:04:23 +02:00
working_set: &StateWorkingSet,
block: &Block,
seen: &mut Vec<VarId>,
seen_blocks: &mut HashMap<BlockId, Vec<(VarId, Span)>>,
output: &mut Vec<(VarId, Span)>,
) -> Result<(), ParseError> {
2021-10-25 22:04:23 +02:00
for flag in &block.signature.named {
if let Some(var_id) = flag.var_id {
seen.push(var_id);
}
}
for positional in &block.signature.required_positional {
if let Some(var_id) = positional.var_id {
seen.push(var_id);
}
}
for positional in &block.signature.optional_positional {
if let Some(var_id) = positional.var_id {
seen.push(var_id);
}
}
for positional in &block.signature.rest_positional {
if let Some(var_id) = positional.var_id {
seen.push(var_id);
}
}
for pipeline in &block.pipelines {
discover_captures_in_pipeline(working_set, pipeline, seen, seen_blocks, output)?;
2021-10-25 22:04:23 +02:00
}
Ok(())
2021-10-25 22:04:23 +02:00
}
fn discover_captures_in_pipeline(
2021-10-25 22:04:23 +02:00
working_set: &StateWorkingSet,
pipeline: &Pipeline,
seen: &mut Vec<VarId>,
seen_blocks: &mut HashMap<BlockId, Vec<(VarId, Span)>>,
output: &mut Vec<(VarId, Span)>,
) -> Result<(), ParseError> {
for element in &pipeline.elements {
discover_captures_in_pipeline_element(working_set, element, seen, seen_blocks, output)?;
2021-10-25 22:04:23 +02:00
}
Ok(())
2021-10-25 22:04:23 +02:00
}
// Closes over captured variables
pub fn discover_captures_in_pipeline_element(
working_set: &StateWorkingSet,
element: &PipelineElement,
seen: &mut Vec<VarId>,
seen_blocks: &mut HashMap<BlockId, Vec<(VarId, Span)>>,
output: &mut Vec<(VarId, Span)>,
) -> Result<(), ParseError> {
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
discover_captures_in_expr(working_set, &element.expr, seen, seen_blocks, output)?;
if let Some(redirection) = element.redirection.as_ref() {
match redirection {
PipelineRedirection::Single { target, .. } => {
if let Some(expr) = target.expr() {
discover_captures_in_expr(working_set, expr, seen, seen_blocks, output)?;
}
}
PipelineRedirection::Separate { out, err } => {
if let Some(expr) = out.expr() {
discover_captures_in_expr(working_set, expr, seen, seen_blocks, output)?;
}
if let Some(expr) = err.expr() {
discover_captures_in_expr(working_set, expr, seen, seen_blocks, output)?;
}
}
Avoid blocking when `o+e>` redirects too much stderr message (#8784) # Description Fixes: #8565 Here is another pr #7240 tried to address the issue, but it works in a wrong way. After this change `o+e>` won't redirect all stdout message then stderr message and it works more like how bash does. # User-Facing Changes For the given python code: ```python # test.py import sys print('aa'*300, flush=True) print('bb'*999999, file=sys.stderr, flush=True) print('cc'*300, flush=True) ``` Running `python test.py out+err> a.txt` shoudn't hang nushell, and `a.txt` keeps output in the same order ## About the change The core idea is that when doing lite-parsing, introduce a new variant `LiteElement::SameTargetRedirection` if we meet `out+err>` redirection token(which is generated by lex function), During converting from lite block to block, LiteElement::SameTargetRedirection will be converted to PipelineElement::SameTargetRedirection. Then in the block eval process, if we get PipelineElement::SameTargetRedirection, we'll invoke `run-external` with `--redirect-combine` flag, then pipe the result into save command ## What happened internally? Take the following command as example: `^ls o+e> log.txt` lex parsing result(`Tokens`) are not changed, but `LiteBlock` and `Block` is changed after this pr. ### LiteBlock before ```rust LiteBlock { block: [ LitePipeline { commands: [ Command(None, LiteCommand { comments: [], parts: [Span { start: 39041, end: 39044 }] }), // actually the span of first Redirection is wrong too.. Redirection(Span { start: 39058, end: 39062 }, StdoutAndStderr, LiteCommand { comments: [], parts: [Span { start: 39050, end: 39057 }] }), ] }] } ``` ### LiteBlock after ```rust LiteBlock { block: [ LitePipeline { commands: [ SameTargetRedirection { cmd: (None, LiteCommand { comments: [], parts: [Span { start: 147945, end: 147948}]}), redirection: (Span { start: 147949, end: 147957 }, LiteCommand { comments: [], parts: [Span { start: 147958, end: 147965 }]}) } ] } ] } ``` ### Block before ```rust Pipeline { elements: [ Expression(None, Expression { expr: ExternalCall(Expression { expr: String("ls"), span: Span { start: 39042, end: 39044 }, ty: String, custom_completion: None }, [], false), span: Span { start: 39041, end: 39044 }, ty: Any, custom_completion: None }), Redirection(Span { start: 39058, end: 39062 }, StdoutAndStderr, Expression { expr: String("out.txt"), span: Span { start: 39050, end: 39057 }, ty: String, custom_completion: None })] } ``` ### Block after ```rust Pipeline { elements: [ SameTargetRedirection { cmd: (None, Expression { expr: ExternalCall(Expression { expr: String("ls"), span: Span { start: 147946, end: 147948 }, ty: String, custom_completion: None}, [], false), span: Span { start: 147945, end: 147948}, ty: Any, custom_completion: None }), redirection: (Span { start: 147949, end: 147957}, Expression {expr: String("log.txt"), span: Span { start: 147958, end: 147965 },ty: String,custom_completion: None} } ] } ``` # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-utils/standard_library/tests.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-05-18 00:47:03 +02:00
}
}
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
Ok(())
}
pub fn discover_captures_in_pattern(pattern: &MatchPattern, seen: &mut Vec<VarId>) {
match &pattern.pattern {
Pattern::Variable(var_id) => seen.push(*var_id),
Pattern::List(items) => {
for item in items {
discover_captures_in_pattern(item, seen)
}
}
Pattern::Record(items) => {
for item in items {
discover_captures_in_pattern(&item.1, seen)
}
}
Pattern::Or(patterns) => {
for pattern in patterns {
discover_captures_in_pattern(pattern, seen)
}
}
Pattern::Rest(var_id) => seen.push(*var_id),
Pattern::Value(_) | Pattern::IgnoreValue | Pattern::IgnoreRest | Pattern::Garbage => {}
}
}
// Closes over captured variables
pub fn discover_captures_in_expr(
2021-10-25 22:04:23 +02:00
working_set: &StateWorkingSet,
expr: &Expression,
seen: &mut Vec<VarId>,
seen_blocks: &mut HashMap<BlockId, Vec<(VarId, Span)>>,
output: &mut Vec<(VarId, Span)>,
) -> Result<(), ParseError> {
2021-10-25 22:04:23 +02:00
match &expr.expr {
Expr::BinaryOp(lhs, _, rhs) => {
discover_captures_in_expr(working_set, lhs, seen, seen_blocks, output)?;
discover_captures_in_expr(working_set, rhs, seen, seen_blocks, output)?;
2021-10-25 22:04:23 +02:00
}
2022-04-06 21:10:25 +02:00
Expr::UnaryNot(expr) => {
discover_captures_in_expr(working_set, expr, seen, seen_blocks, output)?;
2022-04-06 21:10:25 +02:00
}
Expr::Closure(block_id) => {
let block = working_set.get_block(*block_id);
let results = {
let mut seen = vec![];
let mut results = vec![];
discover_captures_in_closure(
working_set,
block,
&mut seen,
seen_blocks,
&mut results,
)?;
for (var_id, span) in results.iter() {
if !seen.contains(var_id) {
if let Some(variable) = working_set.get_variable_if_possible(*var_id) {
if variable.mutable {
return Err(ParseError::CaptureOfMutableVar(*span));
}
}
}
}
results
};
seen_blocks.insert(*block_id, results.clone());
for (var_id, span) in results.into_iter() {
if !seen.contains(&var_id) {
output.push((var_id, span))
}
}
}
2021-10-25 22:04:23 +02:00
Expr::Block(block_id) => {
let block = working_set.get_block(*block_id);
// FIXME: is this correct?
let results = {
let mut seen = vec![];
let mut results = vec![];
discover_captures_in_closure(
working_set,
block,
&mut seen,
seen_blocks,
&mut results,
)?;
results
};
seen_blocks.insert(*block_id, results.clone());
for (var_id, span) in results.into_iter() {
if !seen.contains(&var_id) {
output.push((var_id, span))
}
}
2021-10-25 22:04:23 +02:00
}
2022-03-01 00:31:53 +01:00
Expr::Binary(_) => {}
2021-10-25 22:04:23 +02:00
Expr::Bool(_) => {}
Expr::Call(call) => {
let decl = working_set.get_decl(call.decl_id);
if let Some(block_id) = decl.block_id() {
match seen_blocks.get(&block_id) {
Some(capture_list) => {
Fix capture logic for inner closures (#9754) # Description This fixes the variable capture logic for closures in two cases: * Closures inside of closures did not properly register the closures (or lack thereof) in the outer closure * Closures which called their inner closures before definition did not properly calculate the closures of the outer closure Example of the first case: ``` do { let b = 3; def c [] { $b }; c } ``` Example of the second case (notice `c` is called before it is defined): ``` do { let b = 3; c; def c [] { $b }; c } ``` # User-Facing Changes This should strictly allow closures to work more correctly. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-20 21:10:54 +02:00
// Push captures onto the outer closure that aren't created by that outer closure
for capture in capture_list {
if !seen.contains(&capture.0) {
output.push(*capture);
}
}
}
None => {
let block = working_set.get_block(block_id);
if !block.captures.is_empty() {
Fix capture logic for inner closures (#9754) # Description This fixes the variable capture logic for closures in two cases: * Closures inside of closures did not properly register the closures (or lack thereof) in the outer closure * Closures which called their inner closures before definition did not properly calculate the closures of the outer closure Example of the first case: ``` do { let b = 3; def c [] { $b }; c } ``` Example of the second case (notice `c` is called before it is defined): ``` do { let b = 3; c; def c [] { $b }; c } ``` # User-Facing Changes This should strictly allow closures to work more correctly. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-20 21:10:54 +02:00
for capture in &block.captures {
if !seen.contains(capture) {
output.push((*capture, call.head));
}
}
} else {
Fix capture logic for inner closures (#9754) # Description This fixes the variable capture logic for closures in two cases: * Closures inside of closures did not properly register the closures (or lack thereof) in the outer closure * Closures which called their inner closures before definition did not properly calculate the closures of the outer closure Example of the first case: ``` do { let b = 3; def c [] { $b }; c } ``` Example of the second case (notice `c` is called before it is defined): ``` do { let b = 3; c; def c [] { $b }; c } ``` # User-Facing Changes This should strictly allow closures to work more correctly. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-20 21:10:54 +02:00
let result = {
let mut seen = vec![];
seen_blocks.insert(block_id, output.clone());
let mut result = vec![];
discover_captures_in_closure(
working_set,
block,
&mut seen,
seen_blocks,
&mut result,
)?;
result
};
// Push captures onto the outer closure that aren't created by that outer closure
for capture in &result {
if !seen.contains(&capture.0) {
output.push(*capture);
}
}
seen_blocks.insert(block_id, result);
}
}
}
}
Allow spreading arguments to commands (#11289) <!-- if this PR closes one or more issues, you can automatically link the PR with them by using one of the [*linking keywords*](https://docs.github.com/en/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue#linking-a-pull-request-to-an-issue-using-a-keyword), e.g. - this PR should close #xxxx - fixes #xxxx you can also mention related issues, PRs or discussions! --> Finishes implementing https://github.com/nushell/nushell/issues/10598, which asks for a spread operator in lists, in records, and when calling commands. # Description <!-- Thank you for improving Nushell. Please, check our [contributing guide](../CONTRIBUTING.md) and talk to the core team before making major changes. Description of your pull request goes here. **Provide examples and/or screenshots** if your changes affect the user experience. --> This PR will allow spreading arguments to commands (both internal and external). It will also deprecate spreading arguments automatically when passing to external commands. # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> - Users will be able to use `...` to spread arguments to custom/builtin commands that have rest parameters or allow unknown arguments, or to any external command - If a custom command doesn't have a rest parameter and it doesn't allow unknown arguments either, the spread operator will not be allowed - Passing lists to external commands without `...` will work for now but will cause a deprecation warning saying that it'll stop working in 0.91 (is 2 versions enough time?) Here's a function to help with demonstrating some behavior: ```nushell > def foo [ a, b, c?, d?, ...rest ] { [$a $b $c $d $rest] | to nuon } ``` You can pass a list of arguments to fill in the `rest` parameter using `...`: ```nushell > foo 1 2 3 4 ...[5 6] [1, 2, 3, 4, [5, 6]] ``` If you don't use `...`, the list `[5 6]` will be treated as a single argument: ```nushell > foo 1 2 3 4 [5 6] # Note the double [[]] [1, 2, 3, 4, [[5, 6]]] ``` You can omit optional parameters before the spread arguments: ```nushell > foo 1 2 3 ...[4 5] # d is omitted here [1, 2, 3, null, [4, 5]] ``` If you have multiple lists, you can spread them all: ```nushell > foo 1 2 3 ...[4 5] 6 7 ...[8] ...[] [1, 2, 3, null, [4, 5, 6, 7, 8]] ``` Here's the kind of error you get when you try to spread arguments to a command with no rest parameter: ![image](https://github.com/nushell/nushell/assets/45539777/93faceae-00eb-4e59-ac3f-17f98436e6e4) And this is the warning you get when you pass a list to an external now (without `...`): ![image](https://github.com/nushell/nushell/assets/45539777/d368f590-201e-49fb-8b20-68476ced415e) # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass (on Windows make sure to [enable developer mode](https://learn.microsoft.com/en-us/windows/apps/get-started/developer-mode-features-and-debugging)) - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> Added tests to cover the following cases: - Spreading arguments to a command that doesn't have a rest parameter (unexpected spread argument error) - Spreading arguments to a command that doesn't have a rest parameter *but* there's also a missing positional argument (missing positional error) - Spreading arguments to a command that doesn't have a rest parameter but does allow unknown arguments, such as `exec` (allowed) - Spreading a list literal containing arguments of the wrong type (parse error) - Spreading a non-list value, both to internal and external commands - Having named arguments in the middle of rest arguments - `explain`ing a command call that spreads its arguments # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> # Examples Suppose you have multiple tables: ```nushell let people = [[id name age]; [0 alice 100] [1 bob 200] [2 eve 300]] let evil_twins = [[id name age]; [0 ecila 100] [-1 bob 200] [-2 eve 300]] ``` Maybe you often find yourself needing to merge multiple tables and want a utility to do that. You could write a function like this: ```nushell def merge_all [ ...tables ] { $tables | reduce { |it, acc| $acc | merge $it } } ``` Then you can use it like this: ```nushell > merge_all ...([$people $evil_twins] | each { |$it| $it | select name age }) ╭───┬───────┬─────╮ │ # │ name │ age │ ├───┼───────┼─────┤ │ 0 │ ecila │ 100 │ │ 1 │ bob │ 200 │ │ 2 │ eve │ 300 │ ╰───┴───────┴─────╯ ``` Except they had duplicate columns, so now you first want to suffix every column with a number to tell you which table the column came from. You can make a command for that: ```nushell def select_and_merge [ --cols: list<string>, ...tables ] { let renamed_tables = $tables | enumerate | each { |it| $it.item | select $cols | rename ...($cols | each { |col| $col + ($it.index | into string) }) }; merge_all ...$renamed_tables } ``` And call it like this: ```nushell > select_and_merge --cols [name age] $people $evil_twins ╭───┬───────┬──────┬───────┬──────╮ │ # │ name0 │ age0 │ name1 │ age1 │ ├───┼───────┼──────┼───────┼──────┤ │ 0 │ alice │ 100 │ ecila │ 100 │ │ 1 │ bob │ 200 │ bob │ 200 │ │ 2 │ eve │ 300 │ eve │ 300 │ ╰───┴───────┴──────┴───────┴──────╯ ``` --- Suppose someone's made a command to search for APT packages: ```nushell # The main command def search-pkgs [ --install # Whether to install any packages it finds log_level: int # Pretend it's a good idea to make this a required positional parameter exclude?: list<string> # Packages to exclude repositories?: list<string> # Which repositories to look in (searches in all if not given) ...pkgs # Package names to search for ] { { install: $install, log_level: $log_level, exclude: ($exclude | to nuon), repositories: ($repositories | to nuon), pkgs: ($pkgs | to nuon) } } ``` It has a lot of parameters to configure it, so you might make your own helper commands to wrap around it for specific cases. Here's one example: ```nushell # Only look for packages locally def search-pkgs-local [ --install # Whether to install any packages it finds log_level: int exclude?: list<string> # Packages to exclude ...pkgs # Package names to search for ] { # All required and optional positional parameters are given search-pkgs --install=$install $log_level [] ["<local URI or something>"] ...$pkgs } ``` And you can run it like this: ```nushell > search-pkgs-local --install=false 5 ...["python2.7" "vim"] ╭──────────────┬──────────────────────────────╮ │ install │ false │ │ log_level │ 5 │ │ exclude │ [] │ │ repositories │ ["<local URI or something>"] │ │ pkgs │ ["python2.7", vim] │ ╰──────────────┴──────────────────────────────╯ ``` One thing I realized when writing this was that if we decide to not allow passing optional arguments using the spread operator, then you can (mis?)use the spread operator to skip optional parameters. Here, I didn't want to give `exclude` explicitly, so I used a spread operator to pass the packages to install. Without it, I would've needed to do `search-pkgs-local --install=false 5 [] "python2.7" "vim"` (explicitly pass `[]` (or `null`, in the general case) to `exclude`). There are probably more idiomatic ways to do this, but I just thought it was something interesting. If you're a virologist of the [xkcd](https://xkcd.com/350/) kind, another helper command you might make is this: ```nushell # Install any packages it finds def live-dangerously [ ...pkgs ] { # One optional argument was given (exclude), while another was not (repositories) search-pkgs 0 [] ...$pkgs --install # Flags can go after spread arguments } ``` Running it: ```nushell > live-dangerously "git" "*vi*" # *vi* because I don't feel like typing out vim and neovim ╭──────────────┬─────────────╮ │ install │ true │ │ log_level │ 0 │ │ exclude │ [] │ │ repositories │ null │ │ pkgs │ [git, *vi*] │ ╰──────────────┴─────────────╯ ``` Here's an example that uses the spread operator more than once within the same command call: ```nushell let extras = [ chrome firefox python java git ] def search-pkgs-curated [ ...pkgs ] { (search-pkgs 1 [emacs] ["example.com", "foo.com"] vim # A must for everyone! ...($pkgs | filter { |p| not ($p | str contains "*") }) # Remove packages with globs python # Good tool to have ...$extras --install=false python3) # I forget, did I already put Python in extras? } ``` Running it: ```nushell > search-pkgs-curated "git" "*vi*" ╭──────────────┬───────────────────────────────────────────────────────────────────╮ │ install │ false │ │ log_level │ 1 │ │ exclude │ [emacs] │ │ repositories │ [example.com, foo.com] │ │ pkgs │ [vim, git, python, chrome, firefox, python, java, git, "python3"] │ ╰──────────────┴───────────────────────────────────────────────────────────────────╯ ```
2023-12-28 08:43:20 +01:00
for arg in &call.arguments {
match arg {
Argument::Named(named) => {
if let Some(arg) = &named.2 {
discover_captures_in_expr(working_set, arg, seen, seen_blocks, output)?;
}
}
Argument::Positional(expr)
| Argument::Unknown(expr)
| Argument::Spread(expr) => {
discover_captures_in_expr(working_set, expr, seen, seen_blocks, output)?;
}
2021-10-25 22:04:23 +02:00
}
}
}
Expr::CellPath(_) => {}
Expr::DateTime(_) => {}
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
Expr::ExternalCall(head, args) => {
discover_captures_in_expr(working_set, head, seen, seen_blocks, output)?;
for ExternalArgument::Regular(expr) | ExternalArgument::Spread(expr) in args.as_ref() {
discover_captures_in_expr(working_set, expr, seen, seen_blocks, output)?;
2021-10-25 22:04:23 +02:00
}
}
do not attempt to glob expand if the file path is wrapped in quotes (#11569) # Description Fixes: #11455 ### For arguments which is annotated with `:path/:directory/:glob` To fix the issue, we need to have a way to know if a path is originally quoted during runtime. So the information needed to be added at several levels: * parse time (from user input to expression) We need to add quoted information into `Expr::Filepath`, `Expr::Directory`, `Expr::GlobPattern` * eval time When convert from `Expr::Filepath`, `Expr::Directory`, `Expr::GlobPattern` to `Value::String` during runtime, we won't auto expanded the path if it's quoted ### For `ls` It's really special, because it accepts a `String` as a pattern, and it generates `glob` expression inside the command itself. So the idea behind the change is introducing a special SyntaxShape to ls: `SyntaxShape::LsGlobPattern`. So we can track if the pattern is originally quoted easier, and we don't auto expand the path either. Then when constructing a glob pattern inside ls, we check if input pattern is quoted, if so: we escape the input pattern, so we can run `ls a[123]b`, because it's already escaped. Finally, to accomplish the checking process, we also need to introduce a new value type called `Value::QuotedString` to differ from `Value::String`, it's used to generate an enum called `NuPath`, which is finally used in `ls` function. `ls` learned from `NuPath` to know if user input is quoted. # User-Facing Changes Actually it contains several changes ### For arguments which is annotated with `:path/:directory/:glob` #### Before ```nushell > def foo [p: path] { echo $p }; print (foo "~/a"); print (foo '~/a') /home/windsoilder/a /home/windsoilder/a > def foo [p: directory] { echo $p }; print (foo "~/a"); print (foo '~/a') /home/windsoilder/a /home/windsoilder/a > def foo [p: glob] { echo $p }; print (foo "~/a"); print (foo '~/a') /home/windsoilder/a /home/windsoilder/a ``` #### After ```nushell > def foo [p: path] { echo $p }; print (foo "~/a"); print (foo '~/a') ~/a ~/a > def foo [p: directory] { echo $p }; print (foo "~/a"); print (foo '~/a') ~/a ~/a > def foo [p: glob] { echo $p }; print (foo "~/a"); print (foo '~/a') ~/a ~/a ``` ### For ls command `touch '[uwu]'` #### Before ``` ❯ ls -D "[uwu]" Error: × No matches found for [uwu] ╭─[entry #6:1:1] 1 │ ls -D "[uwu]" · ───┬─── · ╰── Pattern, file or folder not found ╰──── help: no matches found ``` #### After ``` ❯ ls -D "[uwu]" ╭───┬───────┬──────┬──────┬──────────╮ │ # │ name │ type │ size │ modified │ ├───┼───────┼──────┼──────┼──────────┤ │ 0 │ [uwu] │ file │ 0 B │ now │ ╰───┴───────┴──────┴──────┴──────────╯ ``` # Tests + Formatting Done # After Submitting NaN
2024-01-21 16:22:25 +01:00
Expr::Filepath(_, _) => {}
Expr::Directory(_, _) => {}
2021-10-25 22:04:23 +02:00
Expr::Float(_) => {}
Expr::FullCellPath(cell_path) => {
discover_captures_in_expr(working_set, &cell_path.head, seen, seen_blocks, output)?;
2021-10-25 22:04:23 +02:00
}
Expr::ImportPattern(_) => {}
Expr::Overlay(_) => {}
2021-10-25 22:04:23 +02:00
Expr::Garbage => {}
Expr::Nothing => {}
do not attempt to glob expand if the file path is wrapped in quotes (#11569) # Description Fixes: #11455 ### For arguments which is annotated with `:path/:directory/:glob` To fix the issue, we need to have a way to know if a path is originally quoted during runtime. So the information needed to be added at several levels: * parse time (from user input to expression) We need to add quoted information into `Expr::Filepath`, `Expr::Directory`, `Expr::GlobPattern` * eval time When convert from `Expr::Filepath`, `Expr::Directory`, `Expr::GlobPattern` to `Value::String` during runtime, we won't auto expanded the path if it's quoted ### For `ls` It's really special, because it accepts a `String` as a pattern, and it generates `glob` expression inside the command itself. So the idea behind the change is introducing a special SyntaxShape to ls: `SyntaxShape::LsGlobPattern`. So we can track if the pattern is originally quoted easier, and we don't auto expand the path either. Then when constructing a glob pattern inside ls, we check if input pattern is quoted, if so: we escape the input pattern, so we can run `ls a[123]b`, because it's already escaped. Finally, to accomplish the checking process, we also need to introduce a new value type called `Value::QuotedString` to differ from `Value::String`, it's used to generate an enum called `NuPath`, which is finally used in `ls` function. `ls` learned from `NuPath` to know if user input is quoted. # User-Facing Changes Actually it contains several changes ### For arguments which is annotated with `:path/:directory/:glob` #### Before ```nushell > def foo [p: path] { echo $p }; print (foo "~/a"); print (foo '~/a') /home/windsoilder/a /home/windsoilder/a > def foo [p: directory] { echo $p }; print (foo "~/a"); print (foo '~/a') /home/windsoilder/a /home/windsoilder/a > def foo [p: glob] { echo $p }; print (foo "~/a"); print (foo '~/a') /home/windsoilder/a /home/windsoilder/a ``` #### After ```nushell > def foo [p: path] { echo $p }; print (foo "~/a"); print (foo '~/a') ~/a ~/a > def foo [p: directory] { echo $p }; print (foo "~/a"); print (foo '~/a') ~/a ~/a > def foo [p: glob] { echo $p }; print (foo "~/a"); print (foo '~/a') ~/a ~/a ``` ### For ls command `touch '[uwu]'` #### Before ``` ❯ ls -D "[uwu]" Error: × No matches found for [uwu] ╭─[entry #6:1:1] 1 │ ls -D "[uwu]" · ───┬─── · ╰── Pattern, file or folder not found ╰──── help: no matches found ``` #### After ``` ❯ ls -D "[uwu]" ╭───┬───────┬──────┬──────┬──────────╮ │ # │ name │ type │ size │ modified │ ├───┼───────┼──────┼──────┼──────────┤ │ 0 │ [uwu] │ file │ 0 B │ now │ ╰───┴───────┴──────┴──────┴──────────╯ ``` # Tests + Formatting Done # After Submitting NaN
2024-01-21 16:22:25 +01:00
Expr::GlobPattern(_, _) => {}
2021-10-25 22:04:23 +02:00
Expr::Int(_) => {}
Expr::Keyword(kw) => {
discover_captures_in_expr(working_set, &kw.expr, seen, seen_blocks, output)?;
2021-10-25 22:04:23 +02:00
}
Expr::List(list) => {
for item in list {
discover_captures_in_expr(working_set, item.expr(), seen, seen_blocks, output)?;
2021-10-25 22:04:23 +02:00
}
}
Expr::Operator(_) => {}
Expr::Range(range) => {
if let Some(from) = &range.from {
discover_captures_in_expr(working_set, from, seen, seen_blocks, output)?;
2021-10-25 22:04:23 +02:00
}
if let Some(next) = &range.next {
discover_captures_in_expr(working_set, next, seen, seen_blocks, output)?;
2021-10-25 22:04:23 +02:00
}
if let Some(to) = &range.to {
discover_captures_in_expr(working_set, to, seen, seen_blocks, output)?;
2021-10-25 22:04:23 +02:00
}
}
Spread operator in record literals (#11144) Goes towards implementing #10598, which asks for a spread operator in lists, in records, and when calling commands (continuation of #11006, which only implements it in lists) # Description This PR is for adding a spread operator that can be used when building records. Additional functionality can be added later. Changes: - Previously, the `Expr::Record` variant held `(Expression, Expression)` pairs. It now holds instances of an enum `RecordItem` (the name isn't amazing) that allows either a key-value mapping or a spread operator. - `...` will be treated as the spread operator when it appears before `$`, `{`, or `(` inside records (no whitespace allowed in between) (not implemented yet) - The error message for duplicate columns now includes the column name itself, because if two spread records are involved in such an error, you can't tell which field was duplicated from the spans alone `...` will still be treated as a normal string outside records, and even in records, it is not treated as a spread operator when not followed immediately by a `$`, `{`, or `(`. # User-Facing Changes Users will be able to use `...` when building records. ``` > let rec = { x: 1, ...{ a: 2 } } > $rec ╭───┬───╮ │ x │ 1 │ │ a │ 2 │ ╰───┴───╯ > { foo: bar, ...$rec, baz: blah } ╭─────┬──────╮ │ foo │ bar │ │ x │ 1 │ │ a │ 2 │ │ baz │ blah │ ╰─────┴──────╯ ``` If you want to update a field of a record, you'll have to use `merge` instead: ``` > { ...$rec, x: 5 } Error: nu::shell::column_defined_twice × Record field or table column used twice: x ╭─[entry #2:1:1] 1 │ { ...$rec, x: 5 } · ──┬─ ┬ · │ ╰── field redefined here · ╰── field first defined here ╰──── > $rec | merge { x: 5 } ╭───┬───╮ │ x │ 5 │ │ a │ 2 │ ╰───┴───╯ ``` # Tests + Formatting # After Submitting
2023-11-29 18:31:31 +01:00
Expr::Record(items) => {
for item in items {
match item {
RecordItem::Pair(field_name, field_value) => {
discover_captures_in_expr(
working_set,
field_name,
seen,
seen_blocks,
output,
)?;
discover_captures_in_expr(
working_set,
field_value,
seen,
seen_blocks,
output,
)?;
}
RecordItem::Spread(_, record) => {
discover_captures_in_expr(working_set, record, seen, seen_blocks, output)?;
}
}
2021-11-11 00:14:00 +01:00
}
}
Expr::Signature(sig) => {
// Something with a declaration, similar to a var decl, will introduce more VarIds into the stack at eval
for pos in &sig.required_positional {
if let Some(var_id) = pos.var_id {
seen.push(var_id);
}
}
for pos in &sig.optional_positional {
if let Some(var_id) = pos.var_id {
seen.push(var_id);
}
}
if let Some(rest) = &sig.rest_positional {
if let Some(var_id) = rest.var_id {
seen.push(var_id);
}
}
for named in &sig.named {
if let Some(var_id) = named.var_id {
seen.push(var_id);
}
}
}
2021-10-25 22:04:23 +02:00
Expr::String(_) => {}
add raw-string literal support (#9956) # Description This PR adds raw string support by using `r#` at the beginning of single quoted strings and `#` at the end. Notice that escapes do not process, even within single quotes, parentheses don't mean anything, $variables don't mean anything. It's just a string. ```nushell ❯ echo r#'one\ntwo (blah) ($var)'# one\ntwo (blah) ($var) ``` Notice how they work without `echo` or `print` and how they work without carriage returns. ```nushell ❯ r#'adsfa'# adsfa ❯ r##"asdfa'@qpejq'## asdfa'@qpejq ❯ r#'asdfasdfasf ∙ foqwejfqo@'23rfjqf'# ``` They also have a special configurable color in the repl. (use single quotes though) ![image](https://github.com/nushell/nushell/assets/343840/8780e21d-de4c-45b3-9880-2425f5fe10ef) They should work like rust raw literals and allow `r##`, `r###`, `r####`, etc, to help with having one or many `#`'s in the middle of your raw-string. They should work with `let` as well. ```nushell r#'some\nraw\nstring'# | str upcase ``` closes https://github.com/nushell/nushell/issues/5091 # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: WindSoilder <WindSoilder@outlook.com> Co-authored-by: Ian Manske <ian.manske@pm.me>
2024-05-02 15:36:37 +02:00
Expr::RawString(_) => {}
Expr::StringInterpolation(exprs) => {
for expr in exprs {
discover_captures_in_expr(working_set, expr, seen, seen_blocks, output)?;
}
}
Add pattern matching (#8590) # Description This adds `match` and basic pattern matching. An example: ``` match $x { 1..10 => { print "Value is between 1 and 10" } { foo: $bar } => { print $"Value has a 'foo' field with value ($bar)" } [$a, $b] => { print $"Value is a list with two items: ($a) and ($b)" } _ => { print "Value is none of the above" } } ``` Like the recent changes to `if` to allow it to be used as an expression, `match` can also be used as an expression. This allows you to assign the result to a variable, eg) `let xyz = match ...` I've also included a short-hand pattern for matching records, as I think it might help when doing a lot of record patterns: `{$foo}` which is equivalent to `{foo: $foo}`. There are still missing components, so consider this the first step in full pattern matching support. Currently missing: * Patterns for strings * Or-patterns (like the `|` in Rust) * Patterns for tables (unclear how we want to match a table, so it'll need some design) * Patterns for binary values * And much more # User-Facing Changes [see above] # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-03-24 02:52:01 +01:00
Expr::MatchBlock(match_block) => {
for match_ in match_block {
discover_captures_in_pattern(&match_.0, seen);
discover_captures_in_expr(working_set, &match_.1, seen, seen_blocks, output)?;
Add pattern matching (#8590) # Description This adds `match` and basic pattern matching. An example: ``` match $x { 1..10 => { print "Value is between 1 and 10" } { foo: $bar } => { print $"Value has a 'foo' field with value ($bar)" } [$a, $b] => { print $"Value is a list with two items: ($a) and ($b)" } _ => { print "Value is none of the above" } } ``` Like the recent changes to `if` to allow it to be used as an expression, `match` can also be used as an expression. This allows you to assign the result to a variable, eg) `let xyz = match ...` I've also included a short-hand pattern for matching records, as I think it might help when doing a lot of record patterns: `{$foo}` which is equivalent to `{foo: $foo}`. There are still missing components, so consider this the first step in full pattern matching support. Currently missing: * Patterns for strings * Or-patterns (like the `|` in Rust) * Patterns for tables (unclear how we want to match a table, so it'll need some design) * Patterns for binary values * And much more # User-Facing Changes [see above] # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-03-24 02:52:01 +01:00
}
}
2021-11-26 04:49:03 +01:00
Expr::RowCondition(block_id) | Expr::Subexpression(block_id) => {
let block = working_set.get_block(*block_id);
let results = {
let mut results = vec![];
let mut seen = vec![];
discover_captures_in_closure(
working_set,
block,
&mut seen,
seen_blocks,
&mut results,
)?;
results
};
seen_blocks.insert(*block_id, results.clone());
for (var_id, span) in results.into_iter() {
if !seen.contains(&var_id) {
output.push((var_id, span))
}
}
2021-10-25 22:04:23 +02:00
}
Expr::Table(table) => {
for header in table.columns.as_ref() {
discover_captures_in_expr(working_set, header, seen, seen_blocks, output)?;
2021-10-25 22:04:23 +02:00
}
for row in table.rows.as_ref() {
for cell in row.as_ref() {
discover_captures_in_expr(working_set, cell, seen, seen_blocks, output)?;
2021-10-25 22:04:23 +02:00
}
}
}
Expr::ValueWithUnit(value) => {
discover_captures_in_expr(working_set, &value.expr, seen, seen_blocks, output)?;
2021-10-25 22:04:23 +02:00
}
Expr::Var(var_id) => {
if (*var_id > ENV_VARIABLE_ID || *var_id == IN_VARIABLE_ID) && !seen.contains(var_id) {
output.push((*var_id, expr.span));
2021-10-25 22:04:23 +02:00
}
}
Expr::VarDecl(var_id) => {
seen.push(*var_id);
}
}
Ok(())
2021-10-25 22:04:23 +02:00
}
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
fn wrap_redirection_with_collect(
working_set: &mut StateWorkingSet,
target: &RedirectionTarget,
) -> RedirectionTarget {
match target {
RedirectionTarget::File { expr, append, span } => RedirectionTarget::File {
expr: wrap_expr_with_collect(working_set, expr),
span: *span,
append: *append,
},
RedirectionTarget::Pipe { span } => RedirectionTarget::Pipe { span: *span },
}
}
fn wrap_element_with_collect(
working_set: &mut StateWorkingSet,
element: &PipelineElement,
) -> PipelineElement {
IO and redirection overhaul (#11934) # Description The PR overhauls how IO redirection is handled, allowing more explicit and fine-grain control over `stdout` and `stderr` output as well as more efficient IO and piping. To summarize the changes in this PR: - Added a new `IoStream` type to indicate the intended destination for a pipeline element's `stdout` and `stderr`. - The `stdout` and `stderr` `IoStream`s are stored in the `Stack` and to avoid adding 6 additional arguments to every eval function and `Command::run`. The `stdout` and `stderr` streams can be temporarily overwritten through functions on `Stack` and these functions will return a guard that restores the original `stdout` and `stderr` when dropped. - In the AST, redirections are now directly part of a `PipelineElement` as a `Option<Redirection>` field instead of having multiple different `PipelineElement` enum variants for each kind of redirection. This required changes to the parser, mainly in `lite_parser.rs`. - `Command`s can also set a `IoStream` override/redirection which will apply to the previous command in the pipeline. This is used, for example, in `ignore` to allow the previous external command to have its stdout redirected to `Stdio::null()` at spawn time. In contrast, the current implementation has to create an os pipe and manually consume the output on nushell's side. File and pipe redirections (`o>`, `e>`, `e>|`, etc.) have precedence over overrides from commands. This PR improves piping and IO speed, partially addressing #10763. Using the `throughput` command from that issue, this PR gives the following speedup on my setup for the commands below: | Command | Before (MB/s) | After (MB/s) | Bash (MB/s) | | --------------------------- | -------------:| ------------:| -----------:| | `throughput o> /dev/null` | 1169 | 52938 | 54305 | | `throughput \| ignore` | 840 | 55438 | N/A | | `throughput \| null` | Error | 53617 | N/A | | `throughput \| rg 'x'` | 1165 | 3049 | 3736 | | `(throughput) \| rg 'x'` | 810 | 3085 | 3815 | (Numbers above are the median samples for throughput) This PR also paves the way to refactor our `ExternalStream` handling in the various commands. For example, this PR already fixes the following code: ```nushell ^sh -c 'echo -n "hello "; sleep 0; echo "world"' | find "hello world" ``` This returns an empty list on 0.90.1 and returns a highlighted "hello world" on this PR. Since the `stdout` and `stderr` `IoStream`s are available to commands when they are run, then this unlocks the potential for more convenient behavior. E.g., the `find` command can disable its ansi highlighting if it detects that the output `IoStream` is not the terminal. Knowing the output streams will also allow background job output to be redirected more easily and efficiently. # User-Facing Changes - External commands returned from closures will be collected (in most cases): ```nushell 1..2 | each {|_| nu -c "print a" } ``` This gives `["a", "a"]` on this PR, whereas this used to print "a\na\n" and then return an empty list. ```nushell 1..2 | each {|_| nu -c "print -e a" } ``` This gives `["", ""]` and prints "a\na\n" to stderr, whereas this used to return an empty list and print "a\na\n" to stderr. - Trailing new lines are always trimmed for external commands when piping into internal commands or collecting it as a value. (Failure to decode the output as utf-8 will keep the trailing newline for the last binary value.) In the current nushell version, the following three code snippets differ only in parenthesis placement, but they all also have different outputs: 1. `1..2 | each { ^echo a }` ``` a a ╭────────────╮ │ empty list │ ╰────────────╯ ``` 2. `1..2 | each { (^echo a) }` ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` 3. `1..2 | (each { ^echo a })` ``` ╭───┬───╮ │ 0 │ a │ │ │ │ │ 1 │ a │ │ │ │ ╰───┴───╯ ``` But in this PR, the above snippets will all have the same output: ``` ╭───┬───╮ │ 0 │ a │ │ 1 │ a │ ╰───┴───╯ ``` - All existing flags on `run-external` are now deprecated. - File redirections now apply to all commands inside a code block: ```nushell (nu -c "print -e a"; nu -c "print -e b") e> test.out ``` This gives "a\nb\n" in `test.out` and prints nothing. The same result would happen when printing to stdout and using a `o>` file redirection. - External command output will (almost) never be ignored, and ignoring output must be explicit now: ```nushell (^echo a; ^echo b) ``` This prints "a\nb\n", whereas this used to print only "b\n". This only applies to external commands; values and internal commands not in return position will not print anything (e.g., `(echo a; echo b)` still only prints "b"). - `complete` now always captures stderr (`do` is not necessary). # After Submitting The language guide and other documentation will need to be updated.
2024-03-14 21:51:55 +01:00
PipelineElement {
pipe: element.pipe,
expr: wrap_expr_with_collect(working_set, &element.expr),
redirection: element.redirection.as_ref().map(|r| match r {
PipelineRedirection::Single { source, target } => PipelineRedirection::Single {
source: *source,
target: wrap_redirection_with_collect(working_set, target),
},
PipelineRedirection::Separate { out, err } => PipelineRedirection::Separate {
out: wrap_redirection_with_collect(working_set, out),
err: wrap_redirection_with_collect(working_set, err),
},
}),
}
}
fn wrap_expr_with_collect(working_set: &mut StateWorkingSet, expr: &Expression) -> Expression {
let span = expr.span;
Remove broken compile-time overload system (#9677) # Description This PR removes the compile-time overload system. Unfortunately, this system never worked correctly because in a gradual type system where types can be `Any`, you don't have enough information to correctly resolve function calls with overloads. These resolutions must be done at runtime, if they're supported. That said, there's a bit of work that needs to go into resolving input/output types (here overloads do not execute separate commands, but the same command and each overload explains how each output type corresponds to input types). This PR also removes the type scope, which would give incorrect answers in cases where multiple subexpressions were used in a pipeline. # User-Facing Changes Finishes removing compile-time overloads. These were only used in a few places in the code base, but it's possible it may impact user code. I'll mark this as breaking change so we can review. # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-07-13 21:05:03 +02:00
if let Some(decl_id) = working_set.find_decl(b"collect") {
let mut output = vec![];
let var_id = IN_VARIABLE_ID;
let mut signature = Signature::new("");
signature.required_positional.push(PositionalArg {
var_id: Some(var_id),
2021-12-30 04:26:40 +01:00
name: "$in".into(),
desc: String::new(),
shape: SyntaxShape::Any,
2022-03-07 21:08:56 +01:00
default_value: None,
});
let block = Block {
pipelines: vec![Pipeline::from_vec(vec![expr.clone()])],
signature: Box::new(signature),
..Default::default()
};
Make EngineState clone cheaper with Arc on all of the heavy objects (#12229) # Description This makes many of the larger objects in `EngineState` into `Arc`, and uses `Arc::make_mut` to do clone-on-write if the reference is not unique. This is generally very cheap, giving us the best of both worlds - allowing us to mutate without cloning if we have an exclusive reference, and cloning if we don't. This started as more of a curiosity for me after remembering that `Arc::make_mut` exists and can make using `Arc` for mostly immutable data that sometimes needs to be changed very convenient, and also after hearing someone complain about memory usage on Discord - this is a somewhat significant win for that. The exact objects that were wrapped in `Arc`: - `files`, `file_contents` - the strings and byte buffers - `decls` - the whole `Vec`, but mostly to avoid lots of individual `malloc()` calls on Clone rather than for memory usage - `blocks` - the blocks themselves, rather than the outer Vec - `modules` - the modules themselves, rather than the outer Vec - `env_vars`, `previous_env_vars` - the entire maps - `config` The changes required were relatively minimal, but this is a breaking API change. In particular, blocks are added as Arcs, to allow the parser cache functionality to work. With my normal nu config, running on Linux, this saves me about 15 MiB of process memory usage when running interactively (65 MiB → 50 MiB). This also makes quick command executions cheaper, particularly since every REPL loop now involves a clone of the engine state so that we can recover from a panic. It also reduces memory usage where engine state needs to be cloned and sent to another thread or kept within an iterator. # User-Facing Changes Shouldn't be any, since it's all internal stuff, but it does change some public interfaces so it's a breaking change
2024-03-19 19:07:00 +01:00
let block_id = working_set.add_block(Arc::new(block));
output.push(Argument::Positional(Expression::new(
working_set,
Expr::Closure(block_id),
span,
Type::Any,
)));
output.push(Argument::Named((
Spanned {
item: "keep-env".to_string(),
span: Span::new(0, 0),
},
None,
None,
)));
2021-12-30 04:26:40 +01:00
// The containing, synthetic call to `collect`.
// We don't want to have a real span as it will confuse flattening
// The args are where we'll get the real info
Expression::new(
working_set,
Expr::Call(Box::new(Call {
2021-12-30 04:26:40 +01:00
head: Span::new(0, 0),
arguments: output,
decl_id,
parser_info: HashMap::new(),
})),
span,
Type::Any,
)
} else {
Expression::garbage(working_set, span)
}
}
2021-09-06 22:41:30 +02:00
// Parses a vector of u8 to create an AST Block. If a file name is given, then
// the name is stored in the working set. When parsing a source without a file
// name, the source of bytes is stored as "source"
pub fn parse(
2021-09-02 10:25:22 +02:00
working_set: &mut StateWorkingSet,
2021-09-06 22:41:30 +02:00
fname: Option<&str>,
2021-09-02 10:25:22 +02:00
contents: &[u8],
scoped: bool,
Make EngineState clone cheaper with Arc on all of the heavy objects (#12229) # Description This makes many of the larger objects in `EngineState` into `Arc`, and uses `Arc::make_mut` to do clone-on-write if the reference is not unique. This is generally very cheap, giving us the best of both worlds - allowing us to mutate without cloning if we have an exclusive reference, and cloning if we don't. This started as more of a curiosity for me after remembering that `Arc::make_mut` exists and can make using `Arc` for mostly immutable data that sometimes needs to be changed very convenient, and also after hearing someone complain about memory usage on Discord - this is a somewhat significant win for that. The exact objects that were wrapped in `Arc`: - `files`, `file_contents` - the strings and byte buffers - `decls` - the whole `Vec`, but mostly to avoid lots of individual `malloc()` calls on Clone rather than for memory usage - `blocks` - the blocks themselves, rather than the outer Vec - `modules` - the modules themselves, rather than the outer Vec - `env_vars`, `previous_env_vars` - the entire maps - `config` The changes required were relatively minimal, but this is a breaking API change. In particular, blocks are added as Arcs, to allow the parser cache functionality to work. With my normal nu config, running on Linux, this saves me about 15 MiB of process memory usage when running interactively (65 MiB → 50 MiB). This also makes quick command executions cheaper, particularly since every REPL loop now involves a clone of the engine state so that we can recover from a panic. It also reduces memory usage where engine state needs to be cloned and sent to another thread or kept within an iterator. # User-Facing Changes Shouldn't be any, since it's all internal stuff, but it does change some public interfaces so it's a breaking change
2024-03-19 19:07:00 +01:00
) -> Arc<Block> {
add raw-string literal support (#9956) # Description This PR adds raw string support by using `r#` at the beginning of single quoted strings and `#` at the end. Notice that escapes do not process, even within single quotes, parentheses don't mean anything, $variables don't mean anything. It's just a string. ```nushell ❯ echo r#'one\ntwo (blah) ($var)'# one\ntwo (blah) ($var) ``` Notice how they work without `echo` or `print` and how they work without carriage returns. ```nushell ❯ r#'adsfa'# adsfa ❯ r##"asdfa'@qpejq'## asdfa'@qpejq ❯ r#'asdfasdfasf ∙ foqwejfqo@'23rfjqf'# ``` They also have a special configurable color in the repl. (use single quotes though) ![image](https://github.com/nushell/nushell/assets/343840/8780e21d-de4c-45b3-9880-2425f5fe10ef) They should work like rust raw literals and allow `r##`, `r###`, `r####`, etc, to help with having one or many `#`'s in the middle of your raw-string. They should work with `let` as well. ```nushell r#'some\nraw\nstring'# | str upcase ``` closes https://github.com/nushell/nushell/issues/5091 # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: WindSoilder <WindSoilder@outlook.com> Co-authored-by: Ian Manske <ian.manske@pm.me>
2024-05-02 15:36:37 +02:00
trace!("parse");
2021-09-06 22:41:30 +02:00
let name = match fname {
Adds multi-file support to IDE support (#8857) # Description This adds multi-file support to the in-progress IDE support. The main new features are a `-I` flag that allows you to add a new source search path when starting up the nu binary, and fixes for the current IDE support to support spans in other files. This needs accompanying fixes to the vscode/lsp implementation to pass along the project directory via `-I`. UPDATE: Marking this draft until we have a means to test this. # User-Facing Changes _(List of all changes that impact the user experience here. This helps us keep track of breaking changes.)_ # Tests + Formatting Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` # After Submitting If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date.
2023-04-12 19:36:29 +02:00
Some(fname) => {
// use the canonical name for this filename
nu_path::expand_to_real_path(fname)
.to_string_lossy()
.to_string()
}
2021-09-06 22:41:30 +02:00
None => "source".to_string(),
};
2021-07-01 02:01:04 +02:00
let file_id = working_set.add_file(name, contents);
let new_span = working_set.get_span_for_file(file_id);
2021-07-01 02:01:04 +02:00
Reuse the cached parse results of parsed files (#8949) # Description This does a lookup in the cache of parsed files to see if a span can be found for a file that was previously loaded with the same contents, then uses that span to find the parsed block for that file. The end result should, in theory, be identical but doesn't require any reparsing or creating new blocks/new definitions that aren't needed. This drops the sg.nu benchmark from: ``` ╭───┬───────────────────╮ │ 0 │ 280ms 606µs 208ns │ │ 1 │ 282ms 654µs 416ns │ │ 2 │ 252ms 640µs 541ns │ │ 3 │ 250ms 940µs 41ns │ │ 4 │ 241ms 216µs 375ns │ │ 5 │ 257ms 310µs 583ns │ │ 6 │ 196ms 739µs 416ns │ ╰───┴───────────────────╯ ``` to: ``` ╭───┬───────────────────╮ │ 0 │ 118ms 698µs 125ns │ │ 1 │ 121ms 327µs │ │ 2 │ 121ms 873µs 500ns │ │ 3 │ 124ms 94µs 708ns │ │ 4 │ 113ms 733µs 291ns │ │ 5 │ 108ms 663µs 125ns │ │ 6 │ 63ms 482µs 625ns │ ╰───┴───────────────────╯ ``` I was hoping to also see some startup time improvements, but I didn't notice much there. # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-04-21 21:00:33 +02:00
let previously_parsed_block = working_set.find_block_by_span(new_span);
let mut output = {
if let Some(block) = previously_parsed_block {
add raw-string literal support (#9956) # Description This PR adds raw string support by using `r#` at the beginning of single quoted strings and `#` at the end. Notice that escapes do not process, even within single quotes, parentheses don't mean anything, $variables don't mean anything. It's just a string. ```nushell ❯ echo r#'one\ntwo (blah) ($var)'# one\ntwo (blah) ($var) ``` Notice how they work without `echo` or `print` and how they work without carriage returns. ```nushell ❯ r#'adsfa'# adsfa ❯ r##"asdfa'@qpejq'## asdfa'@qpejq ❯ r#'asdfasdfasf ∙ foqwejfqo@'23rfjqf'# ``` They also have a special configurable color in the repl. (use single quotes though) ![image](https://github.com/nushell/nushell/assets/343840/8780e21d-de4c-45b3-9880-2425f5fe10ef) They should work like rust raw literals and allow `r##`, `r###`, `r####`, etc, to help with having one or many `#`'s in the middle of your raw-string. They should work with `let` as well. ```nushell r#'some\nraw\nstring'# | str upcase ``` closes https://github.com/nushell/nushell/issues/5091 # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: WindSoilder <WindSoilder@outlook.com> Co-authored-by: Ian Manske <ian.manske@pm.me>
2024-05-02 15:36:37 +02:00
// dbg!("previous block");
Reuse the cached parse results of parsed files (#8949) # Description This does a lookup in the cache of parsed files to see if a span can be found for a file that was previously loaded with the same contents, then uses that span to find the parsed block for that file. The end result should, in theory, be identical but doesn't require any reparsing or creating new blocks/new definitions that aren't needed. This drops the sg.nu benchmark from: ``` ╭───┬───────────────────╮ │ 0 │ 280ms 606µs 208ns │ │ 1 │ 282ms 654µs 416ns │ │ 2 │ 252ms 640µs 541ns │ │ 3 │ 250ms 940µs 41ns │ │ 4 │ 241ms 216µs 375ns │ │ 5 │ 257ms 310µs 583ns │ │ 6 │ 196ms 739µs 416ns │ ╰───┴───────────────────╯ ``` to: ``` ╭───┬───────────────────╮ │ 0 │ 118ms 698µs 125ns │ │ 1 │ 121ms 327µs │ │ 2 │ 121ms 873µs 500ns │ │ 3 │ 124ms 94µs 708ns │ │ 4 │ 113ms 733µs 291ns │ │ 5 │ 108ms 663µs 125ns │ │ 6 │ 63ms 482µs 625ns │ ╰───┴───────────────────╯ ``` I was hoping to also see some startup time improvements, but I didn't notice much there. # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-04-21 21:00:33 +02:00
return block;
} else {
add raw-string literal support (#9956) # Description This PR adds raw string support by using `r#` at the beginning of single quoted strings and `#` at the end. Notice that escapes do not process, even within single quotes, parentheses don't mean anything, $variables don't mean anything. It's just a string. ```nushell ❯ echo r#'one\ntwo (blah) ($var)'# one\ntwo (blah) ($var) ``` Notice how they work without `echo` or `print` and how they work without carriage returns. ```nushell ❯ r#'adsfa'# adsfa ❯ r##"asdfa'@qpejq'## asdfa'@qpejq ❯ r#'asdfasdfasf ∙ foqwejfqo@'23rfjqf'# ``` They also have a special configurable color in the repl. (use single quotes though) ![image](https://github.com/nushell/nushell/assets/343840/8780e21d-de4c-45b3-9880-2425f5fe10ef) They should work like rust raw literals and allow `r##`, `r###`, `r####`, etc, to help with having one or many `#`'s in the middle of your raw-string. They should work with `let` as well. ```nushell r#'some\nraw\nstring'# | str upcase ``` closes https://github.com/nushell/nushell/issues/5091 # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: WindSoilder <WindSoilder@outlook.com> Co-authored-by: Ian Manske <ian.manske@pm.me>
2024-05-02 15:36:37 +02:00
// dbg!("starting lex");
Reuse the cached parse results of parsed files (#8949) # Description This does a lookup in the cache of parsed files to see if a span can be found for a file that was previously loaded with the same contents, then uses that span to find the parsed block for that file. The end result should, in theory, be identical but doesn't require any reparsing or creating new blocks/new definitions that aren't needed. This drops the sg.nu benchmark from: ``` ╭───┬───────────────────╮ │ 0 │ 280ms 606µs 208ns │ │ 1 │ 282ms 654µs 416ns │ │ 2 │ 252ms 640µs 541ns │ │ 3 │ 250ms 940µs 41ns │ │ 4 │ 241ms 216µs 375ns │ │ 5 │ 257ms 310µs 583ns │ │ 6 │ 196ms 739µs 416ns │ ╰───┴───────────────────╯ ``` to: ``` ╭───┬───────────────────╮ │ 0 │ 118ms 698µs 125ns │ │ 1 │ 121ms 327µs │ │ 2 │ 121ms 873µs 500ns │ │ 3 │ 124ms 94µs 708ns │ │ 4 │ 113ms 733µs 291ns │ │ 5 │ 108ms 663µs 125ns │ │ 6 │ 63ms 482µs 625ns │ ╰───┴───────────────────╯ ``` I was hoping to also see some startup time improvements, but I didn't notice much there. # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-04-21 21:00:33 +02:00
let (output, err) = lex(contents, new_span.start, &[], &[], false);
add raw-string literal support (#9956) # Description This PR adds raw string support by using `r#` at the beginning of single quoted strings and `#` at the end. Notice that escapes do not process, even within single quotes, parentheses don't mean anything, $variables don't mean anything. It's just a string. ```nushell ❯ echo r#'one\ntwo (blah) ($var)'# one\ntwo (blah) ($var) ``` Notice how they work without `echo` or `print` and how they work without carriage returns. ```nushell ❯ r#'adsfa'# adsfa ❯ r##"asdfa'@qpejq'## asdfa'@qpejq ❯ r#'asdfasdfasf ∙ foqwejfqo@'23rfjqf'# ``` They also have a special configurable color in the repl. (use single quotes though) ![image](https://github.com/nushell/nushell/assets/343840/8780e21d-de4c-45b3-9880-2425f5fe10ef) They should work like rust raw literals and allow `r##`, `r###`, `r####`, etc, to help with having one or many `#`'s in the middle of your raw-string. They should work with `let` as well. ```nushell r#'some\nraw\nstring'# | str upcase ``` closes https://github.com/nushell/nushell/issues/5091 # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect -A clippy::result_large_err` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- -c "use std testing; testing run-tests --path crates/nu-std"` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. --> --------- Co-authored-by: WindSoilder <WindSoilder@outlook.com> Co-authored-by: Ian Manske <ian.manske@pm.me>
2024-05-02 15:36:37 +02:00
// dbg!("finished lex");
// dbg!(&output);
Reuse the cached parse results of parsed files (#8949) # Description This does a lookup in the cache of parsed files to see if a span can be found for a file that was previously loaded with the same contents, then uses that span to find the parsed block for that file. The end result should, in theory, be identical but doesn't require any reparsing or creating new blocks/new definitions that aren't needed. This drops the sg.nu benchmark from: ``` ╭───┬───────────────────╮ │ 0 │ 280ms 606µs 208ns │ │ 1 │ 282ms 654µs 416ns │ │ 2 │ 252ms 640µs 541ns │ │ 3 │ 250ms 940µs 41ns │ │ 4 │ 241ms 216µs 375ns │ │ 5 │ 257ms 310µs 583ns │ │ 6 │ 196ms 739µs 416ns │ ╰───┴───────────────────╯ ``` to: ``` ╭───┬───────────────────╮ │ 0 │ 118ms 698µs 125ns │ │ 1 │ 121ms 327µs │ │ 2 │ 121ms 873µs 500ns │ │ 3 │ 124ms 94µs 708ns │ │ 4 │ 113ms 733µs 291ns │ │ 5 │ 108ms 663µs 125ns │ │ 6 │ 63ms 482µs 625ns │ ╰───┴───────────────────╯ ``` I was hoping to also see some startup time improvements, but I didn't notice much there. # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-04-21 21:00:33 +02:00
if let Some(err) = err {
working_set.error(err)
}
2021-07-01 02:01:04 +02:00
Make EngineState clone cheaper with Arc on all of the heavy objects (#12229) # Description This makes many of the larger objects in `EngineState` into `Arc`, and uses `Arc::make_mut` to do clone-on-write if the reference is not unique. This is generally very cheap, giving us the best of both worlds - allowing us to mutate without cloning if we have an exclusive reference, and cloning if we don't. This started as more of a curiosity for me after remembering that `Arc::make_mut` exists and can make using `Arc` for mostly immutable data that sometimes needs to be changed very convenient, and also after hearing someone complain about memory usage on Discord - this is a somewhat significant win for that. The exact objects that were wrapped in `Arc`: - `files`, `file_contents` - the strings and byte buffers - `decls` - the whole `Vec`, but mostly to avoid lots of individual `malloc()` calls on Clone rather than for memory usage - `blocks` - the blocks themselves, rather than the outer Vec - `modules` - the modules themselves, rather than the outer Vec - `env_vars`, `previous_env_vars` - the entire maps - `config` The changes required were relatively minimal, but this is a breaking API change. In particular, blocks are added as Arcs, to allow the parser cache functionality to work. With my normal nu config, running on Linux, this saves me about 15 MiB of process memory usage when running interactively (65 MiB → 50 MiB). This also makes quick command executions cheaper, particularly since every REPL loop now involves a clone of the engine state so that we can recover from a panic. It also reduces memory usage where engine state needs to be cloned and sent to another thread or kept within an iterator. # User-Facing Changes Shouldn't be any, since it's all internal stuff, but it does change some public interfaces so it's a breaking change
2024-03-19 19:07:00 +01:00
Arc::new(parse_block(working_set, &output, new_span, scoped, false))
Reuse the cached parse results of parsed files (#8949) # Description This does a lookup in the cache of parsed files to see if a span can be found for a file that was previously loaded with the same contents, then uses that span to find the parsed block for that file. The end result should, in theory, be identical but doesn't require any reparsing or creating new blocks/new definitions that aren't needed. This drops the sg.nu benchmark from: ``` ╭───┬───────────────────╮ │ 0 │ 280ms 606µs 208ns │ │ 1 │ 282ms 654µs 416ns │ │ 2 │ 252ms 640µs 541ns │ │ 3 │ 250ms 940µs 41ns │ │ 4 │ 241ms 216µs 375ns │ │ 5 │ 257ms 310µs 583ns │ │ 6 │ 196ms 739µs 416ns │ ╰───┴───────────────────╯ ``` to: ``` ╭───┬───────────────────╮ │ 0 │ 118ms 698µs 125ns │ │ 1 │ 121ms 327µs │ │ 2 │ 121ms 873µs 500ns │ │ 3 │ 124ms 94µs 708ns │ │ 4 │ 113ms 733µs 291ns │ │ 5 │ 108ms 663µs 125ns │ │ 6 │ 63ms 482µs 625ns │ ╰───┴───────────────────╯ ``` I was hoping to also see some startup time improvements, but I didn't notice much there. # User-Facing Changes <!-- List of all changes that impact the user experience here. This helps us keep track of breaking changes. --> # Tests + Formatting <!-- Don't forget to add tests that cover your changes. Make sure you've run and fixed any issues with these commands: - `cargo fmt --all -- --check` to check standard code formatting (`cargo fmt --all` applies these changes) - `cargo clippy --workspace -- -D warnings -D clippy::unwrap_used -A clippy::needless_collect` to check that you're using the standard code style - `cargo test --workspace` to check that all tests pass - `cargo run -- crates/nu-std/tests/run.nu` to run the tests for the standard library > **Note** > from `nushell` you can also use the `toolkit` as follows > ```bash > use toolkit.nu # or use an `env_change` hook to activate it automatically > toolkit check pr > ``` --> # After Submitting <!-- If your PR had any user-facing changes, update [the documentation](https://github.com/nushell/nushell.github.io) after the PR is merged, if necessary. This will help us keep the docs up to date. -->
2023-04-21 21:00:33 +02:00
}
};
2021-09-02 10:25:22 +02:00
let mut seen = vec![];
let mut seen_blocks = HashMap::new();
let mut captures = vec![];
match discover_captures_in_closure(
working_set,
&output,
&mut seen,
&mut seen_blocks,
&mut captures,
) {
Make EngineState clone cheaper with Arc on all of the heavy objects (#12229) # Description This makes many of the larger objects in `EngineState` into `Arc`, and uses `Arc::make_mut` to do clone-on-write if the reference is not unique. This is generally very cheap, giving us the best of both worlds - allowing us to mutate without cloning if we have an exclusive reference, and cloning if we don't. This started as more of a curiosity for me after remembering that `Arc::make_mut` exists and can make using `Arc` for mostly immutable data that sometimes needs to be changed very convenient, and also after hearing someone complain about memory usage on Discord - this is a somewhat significant win for that. The exact objects that were wrapped in `Arc`: - `files`, `file_contents` - the strings and byte buffers - `decls` - the whole `Vec`, but mostly to avoid lots of individual `malloc()` calls on Clone rather than for memory usage - `blocks` - the blocks themselves, rather than the outer Vec - `modules` - the modules themselves, rather than the outer Vec - `env_vars`, `previous_env_vars` - the entire maps - `config` The changes required were relatively minimal, but this is a breaking API change. In particular, blocks are added as Arcs, to allow the parser cache functionality to work. With my normal nu config, running on Linux, this saves me about 15 MiB of process memory usage when running interactively (65 MiB → 50 MiB). This also makes quick command executions cheaper, particularly since every REPL loop now involves a clone of the engine state so that we can recover from a panic. It also reduces memory usage where engine state needs to be cloned and sent to another thread or kept within an iterator. # User-Facing Changes Shouldn't be any, since it's all internal stuff, but it does change some public interfaces so it's a breaking change
2024-03-19 19:07:00 +01:00
Ok(_) => {
Arc::make_mut(&mut output).captures =
captures.into_iter().map(|(var_id, _)| var_id).collect();
}
Err(err) => working_set.error(err),
}
// Also check other blocks that might have been imported
let mut errors = vec![];
for (block_idx, block) in working_set.delta.blocks.iter().enumerate() {
let block_id = block_idx + working_set.permanent_state.num_blocks();
if !seen_blocks.contains_key(&block_id) {
let mut captures = vec![];
match discover_captures_in_closure(
working_set,
block,
&mut seen,
&mut seen_blocks,
&mut captures,
) {
Ok(_) => {
seen_blocks.insert(block_id, captures);
}
Err(err) => {
errors.push(err);
}
}
}
}
for err in errors {
working_set.error(err)
}
for (block_id, captures) in seen_blocks.into_iter() {
// In theory, we should only be updating captures where we have new information
// the only place where this is possible would be blocks that are newly created
// by our working set delta. If we ever tried to modify the permanent state, we'd
// panic (again, in theory, this shouldn't be possible)
let block = working_set.get_block(block_id);
let block_captures_empty = block.captures.is_empty();
// need to check block_id >= working_set.permanent_state.num_blocks()
// to avoid mutate a block that is in the permanent state.
// this can happened if user defines a function with recursive call
// and pipe a variable to the command, e.g:
// def px [] { if true { 42 } else { px } }; # the block px is saved in permanent state.
// let x = 3
// $x | px
// If we don't guard for `block_id`, it will change captures of `px`, which is
// already saved in permanent state
if !captures.is_empty()
&& block_captures_empty
&& block_id >= working_set.permanent_state.num_blocks()
{
let block = working_set.get_block_mut(block_id);
block.captures = captures.into_iter().map(|(var_id, _)| var_id).collect();
}
}
output
2021-06-30 03:42:56 +02:00
}