Prior to this PR, `ls` would return `nothing` in an empty directory.
After this PR, it returns an empty `List`. This makes the behaviour of
`ls` more consistent and easier to reason about (IMO).
This was prompted by a user noticing that `ls | where size == 0KB and
type == file` breaks when run in an empty directory:
```
× Input type not supported.
╭─[entry #12:1:1]
1 │ ls | where size == 0KB and type == file
· ─┬ ──┬──
· │ ╰── only list, binary, raw data or range input data is supported
· ╰── input type: nothing
╰────
```
If people agree with this change, let's wait until after the 0.77
release so we have a bit more time to test it.
# Description
Our `ShellError` at the moment has a `std::mem::size_of<ShellError>` of
136 bytes (on AMD64). As a result `Value` directly storing the struct
also required 136 bytes (thanks to alignment requirements).
This change stores the `Value::Error` `ShellError` on the heap.
Pro:
- Value now needs just 80 bytes
- Should be 1 cacheline less (still at least 2 cachelines)
Con:
- More small heap allocations when dealing with `Value::Error`
- More heap fragmentation
- Potential for additional required memcopies
# Further code changes
Includes a small refactor of `try` due to a type mismatch in its large
match.
# User-Facing Changes
None for regular users.
Plugin authors may have to update their matches on `Value` if they use
`nu-protocol`
Needs benchmarking to see if there is a benefit in real world workloads.
**Update** small improvements in runtime for workloads with high volume
of values. Significant reduction in maximum resident set size, when many
values are held in memory.
# Tests + Formatting
# Description
Adds a `profile` command that profiles each pipeline element of a block
and can also recursively step into child blocks.
# Limitations
* It is implemented using pipeline metadata which currently get lost in
some circumstances (e.g.,
https://github.com/nushell/nushell/issues/4501). This means that the
profiler will lose data coming from subexpressions. This issue will
hopefully be solved in the future.
* It also does not step into individual loop iteration which I'm not
sure why but maybe that's a good thing.
# User-Facing Changes
Shouldn't change any existing behavior.
# 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.
---------
Co-authored-by: Darren Schroeder <343840+fdncred@users.noreply.github.com>
This PR is an incremental improvement to `ls` when it encounters
'illegal' file paths on Windows. Related:
https://github.com/nushell/nushell/issues/7869
## Context
We have trouble with filenames that Windows doesn't like, for example
[files with a `.` at the end of their
name](https://learn.microsoft.com/en-us/windows/win32/fileio/naming-a-file#naming-conventions).
To make a long story short, the Rust stdlib and several Win32 APIs will
choke if asked to do something with an illegal filepath. This is a
problem because files with illegal names can be created via other means
(like `touch foo.` in MINGW bash).
Previously `ls` would fail completely in a directory with a bad file,
which isn't great. After this PR, bad files get included in `ls` results
but without any metadata columns. This is not quite where we want to be
— eventually we want to be able to display file metadata for _all_ files
(even naughty ones) — but it's an improvement on the status quo.
### Before
### After
## Future work
Try the workarounds @ChrisDenton suggested:
https://github.com/nushell/nushell/issues/7869#issuecomment-1405977221
Some info on verbatim paths:
https://users.rust-lang.org/t/understanding-windows-paths/58583
## Testing
I tried to write a test for this, but it looks like our testing sandbox
can't create files with illegal filenames.😔 Here's the code in case it
proves useful someday:
```rust
/// Windows doesn't like certain file names, like file names ending with a period:
/// https://learn.microsoft.com/en-us/windows/win32/fileio/naming-a-file#naming-conventions
/// However, those files can still be created with tools like MINGW bash.
/// We may not be able to get full metadata for those files, but we should test that we can at least include them in ls results
#[test]
#[cfg(windows)]
fn can_list_illegal_files() {
Playground::setup("ls_test_all_columns", |dirs, sandbox| {
sandbox.with_files(vec![
EmptyFile("foo"),
EmptyFile("bar."),
EmptyFile("baz"),
]);
let actual = nu!(
cwd: dirs.test(),
"ls | length"
);
assert_eq!(actual.out, "3");
let actual = nu!(
cwd: dirs.test(),
"ls"
);
assert_eq!(actual.out, "1");
let actual = nu!(
cwd: dirs.test(),
"ls | where {|f| $f.name | str ends-with 'bar.'} | length"
);
assert_eq!(actual.out, "1");
})
}
```
# Description
_(Thank you for improving Nushell. Please, check our [contributing
guide](../CONTRIBUTING.md) and talk to the core team before making major
changes.)_
I opened this PR to unify the run command method. It's mainly to improve
consistency across the tree.
# User-Facing Changes
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
# 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.
# Description
This PR adds the `mime-type` to the `type` column if you add the
`--mime-type(-m)` flag to `ls`.
<img width="853" alt="Screenshot 2022-12-27 at 11 43 20 AM"
src="https://user-images.githubusercontent.com/343840/209705499-27fe40fe-0356-4d9d-97f2-4b2dc52e0963.png">
<img width="781" alt="Screenshot 2022-12-27 at 11 45 53 AM"
src="https://user-images.githubusercontent.com/343840/209705509-4d677389-fd68-401e-a7af-3fc6052743b6.png">
# User-Facing Changes
If you specify the `-m` flag, you get the "guessed at" mime type. The
guess is based on the file name and uses this crate
https://docs.rs/mime_guess/latest/mime_guess/ for the guessing.
Part of issue #7612 and and #7524
There's some debate on if the `mime-type` should be added to the `type`
column or if there should be a separate `mime` column. I tend to lean on
the side of `type` since it's technically a type and it's only in that
column if you ask it to be there. Also, I'd prefer to reuse a column
rather than having a list of sprawling columns. Also, as @KodiCraft
suggested, there is precedence as with `ls -d` where the summed size is
in the size column.
I could go either way and if someone wants to create a `mime` column,
we'd probably accept it.
# 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.
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.
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 ls panics when a file or directory not exists
Fixes#6146
Signed-off-by: nibon7 <nibon7@163.com>
* add test
Signed-off-by: nibon7 <nibon7@163.com>
* Avoid extending the directory without globs in `nu_engine::glob_from`
* avoid joining a `*` to the directory without globs
* remove checks on directory permission and whether it is empty
The previous implemention of `nu_engine::glob_from` will extend the
given directory even if it containes no glob pattern. This commit
overcomes lack of consistency with the function `nu_glob::glob`.
* Add flag -D to ls, to list the directory itself instead of its contents
* add --directory (-d) flag to ls
* correct the difference between the given path and the cwd
* set default path to `.` instead of `./*` when --directory (-d) flag is true
* add comments
* add an example
* add tests
* fmt
* Fix `ls` for Windows system files
* Fix non-Windows builds
* Make Clippy happy on non-Windows platforms
* Fix new test on GitHub runners
* Move ls Windows code into its own module
* Switch to short-names when the path is a relative_path (a dir) and exit with an error if the path does not exist
* Remove debugging print line
* Show relative filenames... It does not work yet for ls ../
* Try something else to fix relative paths... it works, but the ../ code part is not very pretty
* Add canonicalize check and remove code clones
* Fix the canonicalize_with issue pointed out by kubouch. Not sure the prefix_str is what kubouch suggested
* Fix the canonicalize_with issue pointed out by kubouch. Not sure the prefix_str is what kubouch suggested
* Add single-dot expansion to nu-path
* Move value path expansion from parser to eval
Fixes#745
* Remove single dot expansion from parser
It is not necessary since it will get expanded anyway in the eval.
* Fix ls to display globs with relative paths
* Use pathdiff crate to get relative paths for ls
Co-authored-by: Stefan Stanciulescu <contact@stefanstanciulescu.com>
* Switch to short-names when the path is a relative_path (a dir) and exit with an error if the path does not exist
* Remove debugging print line
* Show relative filenames... It does not work yet for ls ../
* Try something else to fix relative paths... it works, but the ../ code part is not very pretty
* Add canonicalize check and remove code clones
* Fix the canonicalize_with issue pointed out by kubouch. Not sure the prefix_str is what kubouch suggested
* Fix the canonicalize_with issue pointed out by kubouch. Not sure the prefix_str is what kubouch suggested
* Use only $nu.env.PWD for getting current directory
Because setting and reading to/from std::env changes the global state
shich is problematic if we call `cd` from multiple threads (e.g., in a
`par-each` block).
With this change, when engine-q starts, it will either inherit existing
PWD env var, or create a new one from `std::env::current_dir()`.
Otherwise, everything that needs the current directory will get it from
`$nu.env.PWD`. Each spawned external command will get its current
directory per-process which should be thread-safe.
One thing left to do is to patch nu-path for this as well since it uses
`std::env::current_dir()` in its expansions.
* Rename nu-path functions
*_with is not *_relative which should be more descriptive and frees
"with" for use in a followup commit.
* Clone stack every each iter; Fix some commands
Cloning the stack each iteration of `each` makes sure we're not reusing
PWD between iterations.
Some fixes in commands to make them use the new PWD.
* Post-rebase cleanup, fmt, clippy
* Change back _relative to _with in nu-path funcs
Didn't use the idea I had for the new "_with".
* Remove leftover current_dir from rebase
* Add cwd sync at merge_delta()
This makes sure the parser and completer always have up-to-date cwd.
* Always pass absolute path to glob in ls
* Do not allow PWD a relative path; Allow recovery
Makes it possible to recover PWD by proceeding with the REPL cycle.
* Clone stack in each also for byte/string stream
* (WIP) Start moving env variables to engine state
* (WIP) Move env vars to engine state (ugly)
Quick and dirty code.
* (WIP) Remove unused mut and args; Fmt
* (WIP) Fix dataframe tests
* (WIP) Fix missing args after rebase
* (WIP) Clone only env vars, not the whole stack
* (WIP) Add env var clone to `for` loop as well
* Minor edits
* Refactor merge_delta() to include stack merging.
Less error-prone than doing it manually.
* Clone env for each `update` command iteration
* Mark env var hidden only when found in eng. state
* Fix clippt warnings
* Add TODO about env var reading
* Do not clone empty environment in loops
* Remove extra cwd collection
* Split current_dir() into str and path; Fix autocd
* Make completions respect PWD env var
