ryl 0.3.4

Fast YAML linter (Rust) packaged for PyPI

ryl

ryl - the Rust Yaml Linter is intended to ultimately be a drop in replacement for yamllint. It is usable today, but parity and edge-case behavior are still maturing.

Compatibility note:

• ryl aims to match yamllint behavior and includes many parity tests.
• ryl uses the saphyr parser stack, while yamllint uses the PyYAML parser stack.
• saphyr and PyYAML do not always agree on which files are valid YAML.

Usage

ryl accepts one or more paths: files and/or directories.

Basic:

ryl <PATH_OR_FILE> [PATH_OR_FILE...]

Behavior:

• Files: parsed as YAML even if the extension is not .yml/.yaml.
• Directories: recursively lints .yml and .yaml files.
  • Respects .gitignore, global git ignores, and git excludes.
  • Does not follow symlinks.

Exit codes:

• 0 when all parsed files are valid (or no files found).
• 1 when any invalid YAML is found.
• 2 for CLI usage errors (for example, no paths provided).

Examples:

# Single file
ryl myfile.yml

# Multiple inputs (mix files and directories)
ryl config/ another.yml

# Multiple directories
ryl dir1 dir2

# Explicit non-YAML extension (parsed as YAML)
ryl notes.txt

Help and version:

• ryl -h or ryl --help shows auto-generated help.
• ryl -V or ryl --version prints the version.

The CLI is built with clap, which auto-generates --help and --version.

Performance benchmarking

This repo includes a standalone benchmark script that compares PyPI ryl and yamllint using synthetic YAML corpora and hyperfine.

Prerequisites:

• uv
• hyperfine

Run a quick sample:

uv run scripts/benchmark_perf_vs_yamllint.py --file-counts 25,100 --file-sizes-kib 1,8 --runs 5 --warmup 1

Run a fuller matrix (explicit lists):

uv run scripts/benchmark_perf_vs_yamllint.py --file-counts 25,100,400,1000 --file-sizes-kib 1,8,32,128 --runs 10 --warmup 2

Run a fuller matrix (ranges with increments):

uv run scripts/benchmark_perf_vs_yamllint.py --file-count-start 100 --file-count-end 1000 --file-count-step 100 --file-size-start-kib 4 --file-size-end-kib 64 --file-size-step-kib 4 --runs 10 --warmup 2

The script uses Typer; use --help for all options.

Artifacts are written under manual_outputs/benchmarks/<UTC_TIMESTAMP>/:

• benchmark.png and benchmark.svg: side-by-side facet plot with shared Y axis.
• summary.csv: aggregated timing table.
• meta.json: tool versions and run parameters.
• hyperfine-json/: raw results from hyperfine.

Example benchmark figure (5x5 matrix, 5 runs per point):

Configuration

• Flags:
  • -c, --config-file <FILE>: path to a YAML config file.
  • -d, --config-data <YAML>: inline YAML config (highest precedence).
  • --list-files: print files that would be linted after applying ignores and exit.
  • -f, --format, -s, --strict, --no-warnings: reserved for compatibility.
• Discovery precedence: inline --config-data > --config-file > env YAMLLINT_CONFIG_FILE (global) > nearest project config up the tree (.yamllint, .yamllint.yml, .yamllint.yaml) > user-global ($XDG_CONFIG_HOME/yamllint/config or ~/.config/yamllint/config) > built-in defaults.
• Per-file behavior: unless a global config is set via --config-data, --config-file, or YAMLLINT_CONFIG_FILE, each file discovers its nearest project config. Ignores apply to directory scans and explicit files (parity).
• Presets and extends: supports yamllint’s built-in default, relaxed, and empty via extends. Rule maps are deep-merged; scalars/sequences overwrite.