Contributing

Contributions are welcome. go-ruby-benchmark/benchmark is built to a small set of non-negotiable rules — they are what keep it pure-Go, correct, and MRI-compatible. Please read these before opening a pull request.

Hard rules

• Build from source — no vendoring. Everything compiles from source. Being able to compile from source is a guarantee of independence.
• 100% test coverage target, enforced in CI. New code ships with tests, and coverage is a CI gate. Fill the error branches, not just the happy path.
• All GitHub content in English. Issues, pull requests, commits, comments, and discussions are English-only.
• Differential testing against MRI. Correctness is defined by reference Ruby. The same fixed numbers and a deterministic clock are fed to both ruby's Benchmark and this library and the formatted output is compared byte-for-byte — not approximated from memory.
• Pure Go, cgo disabled. The whole point is a single static binary with no C toolchain. Code must build with CGO_ENABLED=0. If a feature seems to need C, it needs a pure-Go path instead.
• The clock is injected. The library must not read the system clock itself; measurement comes through the Clock seam so every formatted line stays reproducible. A live clock belongs in the consumer.

Workflow

1. Pick or open an issue describing the change.
2. Work test-first: add the differential / unit tests, then make them pass.

3. Run the full suite with coverage and confirm the gate is green:

   COVERPKG=$(go list ./... | paste -sd, -)
   go test -race -coverpkg="$COVERPKG" -coverprofile=cover.out ./...
   go tool cover -func=cover.out | tail -1   # 100.0%
   

4. Open a PR in English, referencing the issue.

Where things live

The library is in github.com/go-ruby-benchmark/benchmark. This documentation site is in github.com/go-ruby-benchmark/docs. Start from the Usage & API page and the Roadmap to find the right place for your change.