More stability checker options (#5299)

## Summary

This contains three changes:
* repos in `check_ecosystem.py` are stored as `org:name` instead of
`org/name` to create a flat directory layout
* `check_ecosystem.py` performs a maximum of 50 parallel jobs at the
same time to avoid consuming to much RAM
* `check-formatter-stability` gets a new option `--multi-project` so
it's possible to do `cargo run --bin ruff_dev --
check-formatter-stability --multi-project target/checkouts`
With these three changes it becomes easy to check the formatter
stability over a larger number of repositories. This is part of the
integration of integrating formatter regressions checks into the
ecosystem checks.

## Test Plan

```shell
python scripts/check_ecosystem.py --checkouts target/checkouts --projects github_search.jsonl -v $(which true) $(which true)
cargo run --bin ruff_dev -- check-formatter-stability --multi-project target/checkouts
```
This commit is contained in:
konstin 2023-06-22 17:48:11 +02:00 committed by GitHub
parent f9f0cf7524
commit 03694ef649
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
2 changed files with 68 additions and 17 deletions

View file

@ -42,6 +42,9 @@ pub(crate) struct Args {
/// Print only the first error and exit, `-x` is same as pytest
#[arg(long, short = 'x')]
pub(crate) exit_first_error: bool,
/// Checks each project inside a directory
#[arg(long)]
pub(crate) multi_project: bool,
}
/// Generate ourself a `try_parse_from` impl for `CheckArgs`. This is a strange way to use clap but
@ -54,6 +57,35 @@ struct WrapperArgs {
}
pub(crate) fn main(args: &Args) -> anyhow::Result<ExitCode> {
let all_success = if args.multi_project {
let mut all_success = true;
for base_dir in &args.files {
for dir in base_dir.read_dir()? {
let dir = dir?;
println!("Starting {}", dir.path().display());
let success = check_repo(&Args {
files: vec![dir.path().clone()],
..*args
});
println!("Finished {}: {:?}", dir.path().display(), success);
if !matches!(success, Ok(true)) {
all_success = false;
}
}
}
all_success
} else {
check_repo(args)?
};
if all_success {
Ok(ExitCode::SUCCESS)
} else {
Ok(ExitCode::FAILURE)
}
}
/// Returns whether the check was successful
pub(crate) fn check_repo(args: &Args) -> anyhow::Result<bool> {
let start = Instant::now();
// Find files to check (or in this case, format twice). Adapted from ruff_cli
@ -77,13 +109,20 @@ pub(crate) fn main(args: &Args) -> anyhow::Result<ExitCode> {
let (paths, _resolver) = python_files_in_path(&cli.files, &pyproject_config, &overrides)?;
assert!(!paths.is_empty(), "no python files in {:?}", cli.files);
let mut formatted_counter = 0;
let errors = paths
.into_iter()
.map(|dir_entry| {
// Doesn't make sense to recover here in this test script
let file = dir_entry
.expect("Iterating the files in the repository failed")
.into_path();
dir_entry.expect("Iterating the files in the repository failed")
})
.filter(|dir_entry| {
// For some reason it does not filter in the beginning
dir_entry.file_name() != "pyproject.toml"
})
.map(|dir_entry| {
let file = dir_entry.path().to_path_buf();
formatted_counter += 1;
// Handle panics (mostly in `debug_assert!`)
let result = match catch_unwind(|| check_file(&file)) {
Ok(result) => result,
@ -166,20 +205,20 @@ Formatted twice:
}
if args.exit_first_error {
return Ok(ExitCode::FAILURE);
return Ok(false);
}
}
let duration = start.elapsed();
println!(
"Formatting {} files twice took {:.2}s",
cli.files.len(),
formatted_counter,
duration.as_secs_f32()
);
if any_errors {
Ok(ExitCode::FAILURE)
Ok(false)
} else {
Ok(ExitCode::SUCCESS)
Ok(true)
}
}