frostem

Versioning

Example: 1.20260729.0 — facade API v1; last upstream algorithm change on 2026-07-29 UTC.

• API breaks are detected with cargo-semver-checks and bump major.
• Daily automation publishes whenever upstream main HEAD advances.
• minor advances only when upstream algorithms/ differs from the previous pin (stem definitions / new or removed .sbl files). The minor number is that commit’s UTC YYYYMMDD.
• Upstream movement that does not touch algorithms/ (compiler, runtime, docs, other backends, …) keeps the previous minor and bumps patch.
• Tag or crates.io version collisions also bump patch.

Algorithm is #[non_exhaustive]: new upstream languages are a minor change, not a major break.

Algorithms

Every Snowball algorithm under algorithms/*.sbl is included (including curiosities such as lovins). Feature names match the algorithm ids (e.g. english, dutch_porter, earlymodernenglish).

Use Algorithm::all() for the set enabled in a given build.

Maintenance

Regenerate from upstream (requires a C toolchain and make):

python3 scripts/sync_from_snowball.py
cargo test

Options:

--snowball-dir DIR   use an existing checkout instead of .snowball-src
--major N            override major (default: keep current)
--minor YYYYMMDD     override minor (default: algorithms/ gate)
--patch N            override patch (default: algorithms/ gate)

scripts/sync_from_snowball.py applies the algorithms gate locally and in CI: if upstream algorithms/ changed vs upstream-pin.toml, minor becomes the commit’s UTC YYYYMMDD and patch is 0; otherwise minor is kept and patch increments. Daily CI only adds major bumps from cargo-semver-checks and extra patch bumps for tag/crates.io collisions.

CI runs a daily sync against main, publishes to crates.io when the upstream commit changes (Trusted Publishing / OIDC), and creates a GitHub Release whose notes summarize Snowball upstream commits and frostem changes since the previous tag.

Trusted Publishing

Publishing from CI uses Trusted Publishing (OIDC via rust-lang/crates-io-auth-action). No CARGO_REGISTRY_TOKEN secret is required for routine releases.

Configured for this repo as:

cargo install cargo-trustpub
cargo trustpub add --publisher github --owner Xuanwo --repo frostem --pipeline daily.yml
cargo trustpub status

(daily.yml must match .github/workflows/daily.yml. No GitHub Environment is used.)

After a successful CI publish, you may optionally require Trusted Publishing for all future versions:

cargo trustpub set --trustpub-only true

If you rename the workflow file or add a GitHub Environment, update the config with cargo trustpub (or the crates.io UI) so it matches exactly.

License

BSD-3-Clause. Stemming algorithms and the Snowball runtime are from the Snowball project; see LICENSE and LICENSE-SNOWBALL.