Add Automatic Gitignore Support

[technicalpickles]

Summary

Adds automatic .gitignore support to pks, respecting gitignored files during analysis. Enabled by default with option to disable for packwerk-identical behavior.

Motivation

Previously, pks analyzed all files regardless of .gitignore status, including build artifacts (node_modules/), temporary files (scratch/), and logs. This meant:

• Unnecessary analysis of ignored files
• Manual exclusion patterns needed
• Behavior inconsistent with git-aware tools

What Changed

For users with .gitignore files:

• Gitignored files are now automatically skipped
• Performance improvement from pruning ignored directories early
• No manual exclusion configuration needed

Configuration:

# packwerk.yml - disable to match packwerk behavior
respect_gitignore: false  # default: true

Supports:

• Local .gitignore files
• Global gitignore (~/.gitignore_global or core.excludesFile)
• .git/info/exclude
• All standard patterns (wildcards, negation, etc.)

Implementation

• Uses ignore crate (same as ripgrep) for battle-tested performance
• Prunes ignored directories during traversal (no stat calls)
• Gracefully handles missing/malformed gitignore files
• Works correctly in non-git directories

Testing

• 10 integration tests covering core scenarios
• 7 unit tests for helper functions
• Test fixtures validate common patterns and edge cases
• All existing tests pass (no regressions)

Troubleshooting

Files unexpectedly ignored?

# Check what's matching
git check-ignore -v path/to/file.rb

# Temporarily disable
echo "respect_gitignore: false" >> packwerk.yml

See ADVANCED_USAGE.md for detailed documentation.

Backward Compatibility

✅ Existing projects work without changes
✅ Can be disabled if needed
✅ Non-git directories continue working
⚠️ Behavior change: gitignored files now skipped (generally desirable)

---

Ready for review! This brings pks behavior in line with git-aware tooling while providing an escape hatch for compatibility.

[Copilot]

Pull Request Overview

This PR adds automatic .gitignore support to pks, making it respect gitignored files during analysis by default. Users can disable this behavior to maintain packwerk-identical functionality.

• Automatically excludes gitignored files from violation checking and file listing
• Adds respect_gitignore configuration option (defaults to true)
• Supports local, global, and git-specific gitignore files with full pattern support

Reviewed Changes

Copilot reviewed 22 out of 23 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
tests/gitignore_test.rs	Comprehensive integration tests covering gitignore functionality scenarios
tests/fixtures/app_with_gitignore_disabled/*	Test fixture demonstrating disabled gitignore behavior
tests/fixtures/app_with_gitignore/*	Test fixture with gitignore patterns for testing
src/packs/walk_directory.rs	Core gitignore implementation with matcher building and directory walking integration
src/packs/raw_configuration.rs	Configuration schema extension for respect_gitignore option
src/packs.rs	Module visibility change to expose walk_directory functions
README.md	Documentation updates describing gitignore support
Cargo.toml	Added ignore and tempfile dependencies
ADVANCED_USAGE.md	Detailed gitignore configuration and troubleshooting guide

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[perryqh]

Performance looks good!

[Copilot]

Pull Request Overview

Copilot reviewed 40 out of 41 changed files in this pull request and generated 9 comments.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[dduugg]

Review

Overall this is well-scoped and well-tested. A few observations:

Concerns

1. Breaking behavior change by default — The default of true means existing users silently get different results without any opt-in. Projects that previously analyzed files in gitignored paths (e.g. vendored code, generated files that someone actually wants checked) will get different output without knowing why. This is the right long-term default, but worth calling out clearly in release notes, and potentially adding a one-time warning on first run.

2. Subprocess on every run — get_global_gitignore() spawns git config --global core.excludesFile via std::process::Command on every walk_directory call. Gracefully returns None when git isn't on PATH, which is correct. Worth noting that the ignore crate already knows how to locate the global gitignore internally — see concern #4.

3. Double Arc wrapping — The code wraps Option<Arc<Gitignore>> in another Arc, resulting in Arc<Option<Arc<Gitignore>>>. The inner Arc is unnecessary; Arc<Option<Gitignore>> suffices.

4. Missed opportunity: ignore::WalkBuilder — The ignore crate has a high-level WalkBuilder that handles gitignore-aware traversal natively (it's what ripgrep uses). The PR manually integrates the lower-level Gitignore type with jwalk, which works fine. If the project ever reconsiders its parallelism strategy, WalkBuilder would handle all of this automatically.

5. Pinned dependencies — globset = "=0.4.16" and ignore = "=0.4.23" are pinned to avoid a Rust edition 2024 requirement in newer patches. Reasonable workaround, but these won't receive patch updates via cargo update. Worth a follow-up to upgrade to Rust edition 2024 and unpin.

Minor Notes

• walk_directory visibility change (pub(crate) → pub): Done to allow gitignore_test.rs to import build_gitignore_matcher. Fine, but slightly expands the public API surface.
• per_file_cache.rs fix: The create_dir_all fix is correct and legitimate — just worth noting it's a bundled unrelated bug fix.
• Test fixtures: Well-structured. The app_with_gitignore_disabled fixture correctly validates opt-out behavior.

None of these are blockers. The feature is solid, the ignore crate is battle-tested, and the test coverage is thorough. Main ask is to document the behavioral change prominently in release notes.

[technicalpickles]

Thanks for the thorough review @dduugg! All five concerns addressed:

1. Breaking change documentation — Added CHANGELOG.md documenting the behavioral change, who's affected, and the opt-out (respect_gitignore: false in packwerk.yml). Skipped the one-time runtime warning — the changelog entry covers it without adding runtime noise.

2+4. Subprocess and WalkBuilder — get_global_gitignore() no longer spawns a subprocess. It now calls ignore::gitignore::gitconfig_excludes_path() directly, which reads ~/.gitconfig (and $XDG_CONFIG_HOME/git/config) without invoking git.

Regarding WalkBuilder: evaluated it, but it can't replace jwalk here. pks uses a custom ProcessReadDirState with jwalk's ClientState trait to propagate current_package_yml from parent to child directories in a single O(1) pass per entry. WalkBuilder's visitor pattern has no equivalent hook for carrying per-directory state across levels. Added a code comment in walk_directory.rs explaining the tradeoff so the next person doesn't have to re-derive it.

3. Double Arc wrapping — Fixed. build_gitignore_matcher() now returns Option<Gitignore> (no inner Arc), and the call site wraps it once in Arc::new() where needed.

5. Pinned dependencies — Unpinned. rust-toolchain.toml is already at 1.92.0 which supports edition 2024, so ignore and globset are now on semver ranges ("0.4") and will pick up patch updates normally.

---

Smoke-tested against zenpayroll (46k files): zero behavioral delta with respect_gitignore: true vs false on the real codebase. The packwerk.yml exclude block already covers the same directories gitignore would filter. Feature verified correct with a synthetic coverage/fake_service.rb test file.