benchmark.yml

name: Performance Benchmark

on:
  pull_request:
    branches: [trunk]
  workflow_dispatch:

permissions:
  contents: read

concurrency:
  group: benchmark-${{ github.event.pull_request.number || github.ref }}
  cancel-in-progress: true

env:
  BENCHMARK_GATE_MODE: enforce

jobs:
  classify:
    name: classify
    runs-on: ubuntu-latest
    timeout-minutes: 5
    permissions:
      contents: read
      pull-requests: read
    outputs:
      applicable: ${{ steps.changed.outputs.applicable }}
    steps:
      - name: Classify benchmark changes
        id: changed
        uses: actions/github-script@3a2844b7e9c422d3c10d287c895573f7108da1b3 # v9.0.0
        with:
          script: |
            if (context.eventName === 'workflow_dispatch') {
              core.setOutput('applicable', 'true');
              return;
            }

            const files = await github.paginate(github.rest.pulls.listFiles, {
              owner: context.repo.owner,
              repo: context.repo.repo,
              pull_number: context.payload.pull_request.number,
              per_page: 100,
            });
            const applicable = files.some(({ filename }) =>
              /^packages\/[^/]+\/src\//.test(filename) ||
              filename.startsWith('tsconfig/') ||
              filename.startsWith('benchmarks/') ||
              filename.startsWith('ci-reports/benchmark/') ||
              filename === 'scripts/bench-threshold-check.mts' ||
              filename === 'scripts/benchmark-readiness-report.mts' ||
              filename === 'scripts/post-benchmark-comment.mjs' ||
              filename.startsWith('scripts/tests/') ||
              filename === 'package.json' ||
              filename === 'vitest.config.ts' ||
              filename === 'vitest.config.bench.ts' ||
              filename === 'tsconfig.json' ||
              filename === 'pnpm-lock.yaml' ||
              filename === '.github/workflows/benchmark.yml'
            );
            core.setOutput('applicable', String(applicable));

  benchmark:
    name: benchmark
    needs: classify
    if: needs.classify.outputs.applicable == 'true'
    runs-on: ubuntu-latest
    timeout-minutes: 20
    permissions:
      contents: read
    steps:
      - name: Checkout pull request revision
        uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
        with:
          ref: ${{ github.event.pull_request.head.sha || github.sha }}
          persist-credentials: false
      - uses: jdx/mise-action@c2a87611a18de5b3828c5652fe268e992400cb5c # v4.3.0
        with:
          version: "2026.9.13"
          cache: false
      - run: pnpm install --frozen-lockfile
      - name: Prepare benchmark dependencies
        run: pnpm bench:prepare
      - name: Run benchmarks with threshold check
        id: bench
        run: pnpm tracked-files:guard --recovery 'pnpm bench:update' -- pnpm bench:check --output-json=benchmark-result.json
        continue-on-error: true
        env:
          CI: true
      - name: Publish benchmark enforce-readiness report
        id: readiness
        if: always()
        continue-on-error: true
        shell: bash
        env:
          BENCHMARK_STEP_OUTCOME: ${{ steps.bench.outcome }}
          BENCHMARK_STEP_CONCLUSION: ${{ steps.bench.conclusion }}
        run: |
          set +e
          pnpm bench:readiness --input=benchmark-result.json --output=ci-reports/benchmark/summary.md --variance-evidence=ci-reports/benchmark/latest-five-green-runs.md
          readiness_exit=$?
          cat ci-reports/benchmark/summary.md >> "$GITHUB_STEP_SUMMARY" || true
          exit "$readiness_exit"
      - name: Upload benchmark readiness report
        if: always()
        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
        with:
          name: benchmark-readiness-report-${{ github.run_id }}-${{ github.run_attempt }}
          path: |
            benchmark-result.json
            ci-reports/benchmark/summary.md
            ci-reports/benchmark/latest-five-green-runs.md
          if-no-files-found: warn
      - name: Fail if benchmark enforcement failed
        if: env.BENCHMARK_GATE_MODE == 'enforce' && (steps.bench.outcome == 'failure' || steps.readiness.outcome == 'failure')
        run: exit 1

  benchmark-gate:
    name: benchmark-gate
    needs: [classify, benchmark]
    if: ${{ always() }}
    runs-on: ubuntu-latest
    timeout-minutes: 5
    permissions:
      contents: read
      pull-requests: read
    steps:
      - name: Require current pull request revision
        if: github.event_name == 'pull_request'
        uses: actions/github-script@3a2844b7e9c422d3c10d287c895573f7108da1b3 # v9.0.0
        with:
          script: |
            const { data: pullRequest } = await github.rest.pulls.get({
              owner: context.repo.owner,
              repo: context.repo.repo,
              pull_number: context.payload.pull_request.number,
            });
            if (pullRequest.head.sha !== context.payload.pull_request.head.sha) {
              core.setFailed(`Benchmark evidence targets stale revision ${context.payload.pull_request.head.sha}; current head is ${pullRequest.head.sha}.`);
            }
      - name: Enforce benchmark result
        shell: bash
        env:
          CLASSIFY_RESULT: ${{ needs.classify.result }}
          BENCHMARK_APPLICABLE: ${{ needs.classify.outputs.applicable }}
          BENCHMARK_RESULT: ${{ needs.benchmark.result }}
        run: |
          if [ "$CLASSIFY_RESULT" != "success" ]; then
            echo "::error::Benchmark change classification did not succeed."
            exit 1
          fi
          if [ "$BENCHMARK_APPLICABLE" != "true" ]; then
            echo "No benchmark-relevant files changed; benchmark gate passes as not applicable."
            exit 0
          fi
          if [ "$BENCHMARK_RESULT" != "success" ]; then
            echo "::error::Benchmark verification concluded $BENCHMARK_RESULT."
            exit 1
          fi
          echo "Benchmark verification passed for the current pull request revision."