nushell/crates/nu-cli/src/completions/command_completions.rs
Yash Thakur 671640b0a9
Avoid recomputing fuzzy match scores (#13700)
<!--
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.
-->

This PR makes it so that when using fuzzy matching, the score isn't
recomputed when sorting. Instead, filtering and sorting suggestions is
handled by a new `NuMatcher` struct. This struct accepts suggestions
and, if they match the user's typed text, stores those suggestions
(along with their scores and values). At the end, it returns a sorted
list of suggestions.

This probably won't have a noticeable impact on performance, but it
might be helpful if we start using Nucleo in the future.

Minor change: Makes `find_commands_by_predicate` in `StateWorkingSet`
and `EngineState` take `FnMut` rather than `Fn` for the predicate.

# User-Facing Changes
<!-- List of all changes that impact the user experience here. This
helps us keep track of breaking changes. -->

When using case-insensitive matching, if you have two matches `FOO` and
`abc`, `abc` will be shown before `FOO` rather than the other way
around. I think this way makes more sense than the current behavior.
When I brought this up on Discord, WindSoilder did say it would make
sense to show uppercase matches first if the user typed, say, `F`.
However, that would be a lot more complicated to implement.

# 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 toolkit.nu; toolkit test stdlib"` 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 a test for the changes in
https://github.com/nushell/nushell/pull/13302.

# 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.
-->
2024-11-22 06:29:00 -06:00

339 lines
11 KiB
Rust

use std::collections::HashMap;
use crate::{
completions::{Completer, CompletionOptions},
SuggestionKind,
};
use nu_parser::FlatShape;
use nu_protocol::{
engine::{CachedFile, Stack, StateWorkingSet},
Span,
};
use reedline::Suggestion;
use super::{completion_options::NuMatcher, SemanticSuggestion};
pub struct CommandCompletion {
flattened: Vec<(Span, FlatShape)>,
flat_shape: FlatShape,
force_completion_after_space: bool,
}
impl CommandCompletion {
pub fn new(
flattened: Vec<(Span, FlatShape)>,
flat_shape: FlatShape,
force_completion_after_space: bool,
) -> Self {
Self {
flattened,
flat_shape,
force_completion_after_space,
}
}
fn external_command_completion(
&self,
working_set: &StateWorkingSet,
sugg_span: reedline::Span,
matched_internal: impl Fn(&str) -> bool,
matcher: &mut NuMatcher<String>,
) -> HashMap<String, SemanticSuggestion> {
let mut suggs = HashMap::new();
// os agnostic way to get the PATH env var
let paths = working_set.permanent_state.get_path_env_var();
if let Some(paths) = paths {
if let Ok(paths) = paths.as_list() {
for path in paths {
let path = path.coerce_str().unwrap_or_default();
if let Ok(mut contents) = std::fs::read_dir(path.as_ref()) {
while let Some(Ok(item)) = contents.next() {
if working_set
.permanent_state
.config
.completions
.external
.max_results
<= suggs.len() as i64
{
break;
}
let Ok(name) = item.file_name().into_string() else {
continue;
};
let value = if matched_internal(&name) {
format!("^{}", name)
} else {
name.clone()
};
if suggs.contains_key(&value) {
continue;
}
if matcher.matches(&name) && is_executable::is_executable(item.path()) {
// If there's an internal command with the same name, adds ^cmd to the
// matcher so that both the internal and external command are included
matcher.add(&name, value.clone());
suggs.insert(
value.clone(),
SemanticSuggestion {
suggestion: Suggestion {
value,
span: sugg_span,
append_whitespace: true,
..Default::default()
},
// TODO: is there a way to create a test?
kind: None,
},
);
}
}
}
}
}
}
suggs
}
fn complete_commands(
&self,
working_set: &StateWorkingSet,
span: Span,
offset: usize,
find_externals: bool,
options: &CompletionOptions,
) -> Vec<SemanticSuggestion> {
let partial = working_set.get_span_contents(span);
let mut matcher = NuMatcher::new(String::from_utf8_lossy(partial), options.clone());
let sugg_span = reedline::Span::new(span.start - offset, span.end - offset);
let mut internal_suggs = HashMap::new();
let filtered_commands = working_set.find_commands_by_predicate(
|name| {
let name = String::from_utf8_lossy(name);
matcher.add(&name, name.to_string())
},
true,
);
for (name, description, typ) in filtered_commands {
let name = String::from_utf8_lossy(&name);
internal_suggs.insert(
name.to_string(),
SemanticSuggestion {
suggestion: Suggestion {
value: name.to_string(),
description,
span: sugg_span,
append_whitespace: true,
..Suggestion::default()
},
kind: Some(SuggestionKind::Command(typ)),
},
);
}
let mut external_suggs = if find_externals {
self.external_command_completion(
working_set,
sugg_span,
|name| internal_suggs.contains_key(name),
&mut matcher,
)
} else {
HashMap::new()
};
let mut res = Vec::new();
for cmd_name in matcher.results() {
if let Some(sugg) = internal_suggs
.remove(&cmd_name)
.or_else(|| external_suggs.remove(&cmd_name))
{
res.push(sugg);
}
}
res
}
}
impl Completer for CommandCompletion {
fn fetch(
&mut self,
working_set: &StateWorkingSet,
_stack: &Stack,
_prefix: &[u8],
span: Span,
offset: usize,
pos: usize,
options: &CompletionOptions,
) -> Vec<SemanticSuggestion> {
let last = self
.flattened
.iter()
.rev()
.skip_while(|x| x.0.end > pos)
.take_while(|x| {
matches!(
x.1,
FlatShape::InternalCall(_)
| FlatShape::External
| FlatShape::ExternalArg
| FlatShape::Literal
| FlatShape::String
)
})
.last();
// The last item here would be the earliest shape that could possible by part of this subcommand
let subcommands = if let Some(last) = last {
self.complete_commands(
working_set,
Span::new(last.0.start, pos),
offset,
false,
options,
)
} else {
vec![]
};
if !subcommands.is_empty() {
return subcommands;
}
let config = working_set.get_config();
if matches!(self.flat_shape, nu_parser::FlatShape::External)
|| matches!(self.flat_shape, nu_parser::FlatShape::InternalCall(_))
|| ((span.end - span.start) == 0)
|| is_passthrough_command(working_set.delta.get_file_contents())
{
// we're in a gap or at a command
if working_set.get_span_contents(span).is_empty() && !self.force_completion_after_space
{
return vec![];
}
self.complete_commands(
working_set,
span,
offset,
config.completions.external.enable,
options,
)
} else {
vec![]
}
}
}
pub fn find_non_whitespace_index(contents: &[u8], start: usize) -> usize {
match contents.get(start..) {
Some(contents) => {
contents
.iter()
.take_while(|x| x.is_ascii_whitespace())
.count()
+ start
}
None => start,
}
}
pub fn is_passthrough_command(working_set_file_contents: &[CachedFile]) -> bool {
for cached_file in working_set_file_contents {
let contents = &cached_file.content;
let last_pipe_pos_rev = contents.iter().rev().position(|x| x == &b'|');
let last_pipe_pos = last_pipe_pos_rev.map(|x| contents.len() - x).unwrap_or(0);
let cur_pos = find_non_whitespace_index(contents, last_pipe_pos);
let result = match contents.get(cur_pos..) {
Some(contents) => contents.starts_with(b"sudo ") || contents.starts_with(b"doas "),
None => false,
};
if result {
return true;
}
}
false
}
#[cfg(test)]
mod command_completions_tests {
use super::*;
use nu_protocol::engine::EngineState;
use std::sync::Arc;
#[test]
fn test_find_non_whitespace_index() {
let commands = [
(" hello", 4),
("sudo ", 0),
(" sudo ", 2),
(" sudo ", 2),
(" hello ", 1),
(" hello ", 3),
(" hello | sudo ", 4),
(" sudo|sudo", 5),
("sudo | sudo ", 0),
(" hello sud", 1),
];
for (idx, ele) in commands.iter().enumerate() {
let index = find_non_whitespace_index(ele.0.as_bytes(), 0);
assert_eq!(index, ele.1, "Failed on index {}", idx);
}
}
#[test]
fn test_is_last_command_passthrough() {
let commands = [
(" hello", false),
(" sudo ", true),
("sudo ", true),
(" hello", false),
(" sudo", false),
(" sudo ", true),
(" sudo ", true),
(" sudo ", true),
(" hello ", false),
(" hello | sudo ", true),
(" sudo|sudo", false),
("sudo | sudo ", true),
(" hello sud", false),
(" sudo | sud ", false),
(" sudo|sudo ", true),
(" sudo | sudo ls | sudo ", true),
];
for (idx, ele) in commands.iter().enumerate() {
let input = ele.0.as_bytes();
let mut engine_state = EngineState::new();
engine_state.add_file("test.nu".into(), Arc::new([]));
let delta = {
let mut working_set = StateWorkingSet::new(&engine_state);
let _ = working_set.add_file("child.nu".into(), input);
working_set.render()
};
let result = engine_state.merge_delta(delta);
assert!(
result.is_ok(),
"Merge delta has failed: {}",
result.err().unwrap()
);
let is_passthrough_command = is_passthrough_command(engine_state.get_file_contents());
assert_eq!(
is_passthrough_command, ele.1,
"index for '{}': {}",
ele.0, idx
);
}
}
}