@redproof/stryker

Redproof adapter for Stryker mutation testing. It turns mutation score and undetected mutants into Redproof Rules, so you can prove your tests really catch broken code.

Part of Redproof. Redproof does not just run your guardrails. It proves they can fail.

Install

npm install --save-dev @redproof/stryker redproof @stryker-mutator/core

Use

// gates/test-strength.ts
import { stryker } from '@redproof/stryker';
import { defineGate } from 'redproof';

const adapter = stryker({
  // Optional: run Stryker from a package inside the Gate root.
  cwd: 'packages/parser',
  configFile: 'stryker.config.mjs',
  rules: {
    mutantsDetected: true,
    mutationScore: { minimum: 100 },
  },
});

export default defineGate({ id: 'test-strength', adapter });

cwd is relative to the Gate root and cannot resolve outside it, including through a symbolic link. configFile, acceptedMutantsFile, and Stryker's own relative paths are resolved from that working directory. A cwd that does not exist is reported as stryker-unavailable.

If a mature project intentionally carries surviving or uncovered mutants, use an accepted-mutant baseline instead of weakening the score until unrelated regressions fit under it. Select noNewUndetectedMutants in place of mutantsDetected. With both selected, every accepted mutant still breaches mutantsDetected.

const adapter = stryker({
  configFile: 'stryker.config.mjs',
  rules: {
    noNewUndetectedMutants: {
      acceptedMutantsFile: 'accepted-mutants.json',
    },
  },
});

[
  {
    "fileName": "src/parser.ts",
    "mutatorName": "EqualityOperator",
    "replacement": ">",
    "location": {
      "start": { "line": 4, "column": 10 },
      "end": { "line": 4, "column": 12 }
    },
    "reason": "Equivalent for the supported input domain."
  }
]

The baseline is a JSON array. Redproof resolves the file from the working directory and refuses a path that leaves the Gate root. Each fileName is relative to the working directory. The identity of a mutant is its file, its full start and end location, its mutator, and its replacement. This is the key Stryker uses in stryker-incremental.json, so you can copy an entry from that report or from the JSON reporter output. A different undetected mutant breaches the Rule even when the survivor count is unchanged. The optional reason is documentation and does not take part in matching.

Three cases REFUSE. A malformed, duplicate, or escaping entry refuses before Stryker runs. An accepted mutant that the tests now detect refuses, because the baseline would pass that mutant if it survived again. Remove the entry. An undetected mutant without a stable identity refuses, and no other Rule is evaluated in that run. The replacement text comes from Stryker's code generator, so a Stryker upgrade can change it and make entries stale.

The Adapter disables Stryker's reporters and logs while it runs. This keeps Redproof's text, JSON, and SARIF output as the only reporting stream. A Stryker failure is returned as REFUSE detail; rerun Stryker directly when you need its native debug log or mutation report. Mutant and baseline locations are reported relative to the Gate root, even when Stryker runs from a nested cwd.

Then run:

npx redproof check   # do the rules hold right now?
npx redproof prove   # can each rule actually fail?

Documentation

See the Redproof documentation.

License

Apache-2.0