Currently `last n` memory use is O(input), while it should be O(n). This
patch replaces code collecting all of last's input into a Vec<_> with
collecting into a bounded VecDeque<_>. UI/UX remain are unchanged.
Alters `all`, `any`, `each while`, `each`, `insert`, `par-each`, `reduce`, `update`, `upsert` and `where`,
so that their blocks take an optional parameter containing the index.
* Make json require string and pass around metadata
The json deserializer was accepting any inputs by coercing non-strings
into strings. As an example, if the input was `[1, 2]` the coercion
would turn into `[12]` and deserialize as a list containing number
twelve instead of a list of two numbers, one and two. This could lead
to silent data corruption.
Aside from that pipeline metadata wasn't passed aroud.
This commit fixes the type issue by adding a strict conversion
function that errors if the input type is not a string or external
stream. It then uses this function instead of the original
`collect_string()`. In addition, this function returns the pipeline
metadata so it can be passed along.
* Make other formats require string
The problem with json coercing non-string types to string was present in
all other text formats. This reuses the `collect_string_strict` function
to fix them.
* `IntoPipelineData` cleanup
The method `into_pipeline_data_with_metadata` can now be conveniently
used.
* add signature information when help on one command
* tell user that one command support operated on cell paths
Also, make type output to be more friendly, like `record<>` should just be `record`
And the same to `table<>`, which should be `table`
* simplify code
* don't show signatures for parser keyword
* update comment
* output arg syntax shape as type, so it's the same as describe command
* fix string when no positional args
* update signature body
* update
* add help signature test
* fix arg output format for composed data type like list or record
* fix clippy
* add comment
* Grouped config commands better
* Tweaked test slightly
* Fix merge conflict(?)
* Remove recently-added test case
* Revert rm.always_trash default
* Untweak rm help messages
* Formatting
* Remove example
* Add deprecation warning
* Remove deprecation timeline
Not sure we want to commit to a specific timeline just yet
Co-authored-by: Reilly Wood <26268125+rgwood@users.noreply.github.com>
* removes unused features.
* Adds back multithreading feature to sysinfo.
* Adds back alloc for percent-encoding
* Adds updated lock file.
* Missed one sysinfo.
* `indexmap` just defaults
* Revert `miette``default-features=false`
Co-authored-by: Stefan Holderbach <sholderbach@users.noreply.github.com>
Co-authored-by: JT <547158+jntrnr@users.noreply.github.com>
This adds support for (limited) mutable variables. Mutable variables are created with mut much the same way immutable variables are made with let.
Mutable variables allow mutation via the assignment operator (=).
❯ mut x = 100
❯ $x = 200
❯ print $x
200
Mutable variables are limited in that they're only tended to be used in the local code block. Trying to capture a local variable will result in an error:
❯ mut x = 123; {|| $x }
Error: nu::parser::expected_keyword (link)
× Capture of mutable variable.
The intent of this limitation is to reduce some of the issues with mutable variables in general: namely they make code that's harder to reason about. By reducing the scope that a mutable variable can be used it, we can help create local reasoning about them.
Mutation can occur with fields as well, as in this case:
❯ mut y = {abc: 123}
❯ $y.abc = 456
❯ $y
On a historical note: mutable variables are something that we resisted for quite a long time, leaning as much as we could on the functional style of pipelines and dataflow. That said, we've watched folks struggle to work with reduce as an approximation for patterns that would be trivial to express with local mutation. With that in mind, we're leaning towards the happy path.
- Custom commands are true for builtin and custom
- Add classification as external command
- Specify wildcard in keyword: keyword is true for builtin and keyword
* Fix ignore-errors for select
* fix Value::List match
* fix invalid rows
* add tests
* fix ListStream match
* add one more test for ListStream
* add more tests
* tweak words
* Add failing test that list of ints and floats is List<Number>
* Start defining subtype relation
* Make it possible to declare input and output types for commands
- Enforce them in tests
* Declare input and output types of commands
* Add formatted signatures to `help commands` table
* Revert SyntaxShape::Table -> Type::Table change
* Revert unnecessary derive(Hash) on SyntaxShape
Co-authored-by: JT <547158+jntrnr@users.noreply.github.com>
Allows use of slightly optimized variants that check if they have to use
the heavier vte parser. Tries to avoid unnnecessary allocations. Initial
performance characteristics proven out in #4378.
Also reduces boilerplate with right-ward drift.
* fix: fixcd
try to fix
Log: try to fix the bug with can enter a permisson error fold
* change wording
* fat
* fmt
Co-authored-by: Darren Schroeder <343840+fdncred@users.noreply.github.com>
* make format filesize more flexible
* make code simpler
* finish refactor on bytes commands
* finish refactor on str commands
* fimplify code
* rename from column_paths to cell_paths
* fix: ✨ "saner" default for mv
fixes#6747
As highlighted in the issue, the default behavior of nu currently
is to overwrite the destination file without notice.
This is not a "standard" expectation users that want this behavior
can create a dedicated alias.
* fix: 📝 edit the comment
* fix: ✨ updated the tests
* fix: 🚧 use --force for case test
This reduces the number of dependencies to build for `cargo build` or
`cargo run` by around 19.
Will not speed up CI as we need to `cargo test` but should help for
`cargo install` or users just building from source.
Time saved on my machine ~0.8 secs so likely unnoticable in noise.
Larger future goal is reducing longer dependency chains to allow more
parallel compilation.
Fixes#6895
Warning: `Iterator::map_while` does not return a `FusedIterator`!
Depending on the consuming adaptor or code (e.g. for loop) the iteration
may be stopped but this is not guaranteed.
Adds a test example but Examples handle iterators like
`FusedIterator` and thus don't catch the regression
Cleanup the message on another test
Disable backtrace on miette
- gimli crate requires several seconds
Disable diagnostics on wax
- depends on an outdated miette version
Builds fine, no observable loss in diagnostics quality of life
Removes 10 crates that have to be compiled.
* Remove unnecessary `#[allow]` annots
Reduce the number of lint exceptions that are not necessary with the
current state of the code (or more recent toolchain)
* Remove dead code from `FileStructure` in nu-command
* Replace `allow(unused)` with relevant feature switch
* Deal with `needless_collect` with annotations
* Change hack for needless_collect in `from json`
This change obviates the need for `allow(needless_collect)`
Removes a pessimistic allocation for empty strings, but increases
allocation size to `Value`
Probably not really worth it.
* Revert "Deal with `needless_collect` with annotations"
This reverts commit 05aca98445.
The previous state seems to better from a performance perspective as a
`Vec<String>` is lighter weight than `Vec<Value>`
* Wrap `open` parse errors from `from` commands
Minimal fix for #6843
This propagates the underlying errors from the called `from` commands
and adds a top-level error with the full path and the understood file
extension and resulting called command.
* Repoint inner span for `from ...` to `open`
* Add actionable message: refer to help or use --raw
* Add support to render right prompt on last line of the prompt
* reset reedline to main branch
* update reedline to fix right prompt to be rendered correctly
* reset reedline to main branch again
