api-review-baselines.yml

# This workflow finds changed `*.baseline.json` files in a merged PR,
# labels the PR for API review, generates ApiChief diffs, and posts them as PR comments.

name: Comment API baseline deltas on PRs

on:
  pull_request_target:
    types: [closed]
    branches:
      - main
      - release/**
  workflow_dispatch:
    inputs:
      pr-number:
        description: Pull request number to process
        required: true
        type: number

permissions:
  contents: read
  issues: write
  pull-requests: write

jobs:
  api-review:
    if: github.event_name == 'workflow_dispatch' || github.event.pull_request.merged == true
    runs-on: ubuntu-latest
    steps:
      - name: Detect changed baseline files and add label
        id: detect
        uses: actions/github-script@3a2844b7e9c422d3c10d287c895573f7108da1b3
        with:
          script: |
            const owner = context.repo.owner;
            const repo = context.repo.repo;
            const prNumberInput = context.eventName === 'workflow_dispatch'
              ? context.payload.inputs?.['pr-number']
              : context.payload.pull_request.number;
            const prNumber = Number(prNumberInput);

            if (!Number.isInteger(prNumber) || prNumber <= 0) {
              core.setFailed(`Invalid PR number: ${prNumber}`);
              return;
            }

            const { data: pullRequest } = await github.rest.pulls.get({
              owner,
              repo,
              pull_number: prNumber
            });

            let baseSha = pullRequest.base.sha;
            const targetSha = pullRequest.merge_commit_sha ?? pullRequest.head.sha;

            // pull_request_target's `base.sha` is the original base from when the PR was opened, so it
            // includes anything pushed to the target branch while the PR was open. The merge commit's
            // first parent is the target tip immediately before the merge — that gives a diff that
            // reflects only what the PR introduced.
            if (pullRequest.merged && pullRequest.merge_commit_sha) {
              try {
                const { data: mergeCommit } = await github.rest.git.getCommit({
                  owner,
                  repo,
                  commit_sha: pullRequest.merge_commit_sha
                });
                if (mergeCommit.parents.length > 0) {
                  baseSha = mergeCommit.parents[0].sha;
                  console.log(`Using merge commit's first parent ${baseSha} as base.`);
                }
              } catch (err) {
                console.log(`Failed to fetch merge commit; falling back to PR base.sha: ${err.message}`);
              }
            }

            const files = await github.paginate(github.rest.pulls.listFiles, {
              owner,
              repo,
              pull_number: prNumber,
              per_page: 100
            });

            const baselineFiles = files
              .filter(file => file.filename.endsWith('.baseline.json'))
              .map(file => ({
                filename: file.filename,
                status: file.status,
                previous_filename: file.previous_filename ?? null
              }));

            core.setOutput('pr_number', String(prNumber));
            core.setOutput('base_sha', baseSha);
            core.setOutput('target_sha', targetSha);
            core.setOutput('has_baselines', String(baselineFiles.length > 0));
            core.setOutput('files_json', JSON.stringify(baselineFiles));

            if (!pullRequest.merged) {
              console.log(`PR #${prNumber} is not merged; using head SHA ${pullRequest.head.sha}.`);
            }

            if (baselineFiles.length === 0) {
              console.log(`No changed baseline files detected on PR #${prNumber}.`);
              return;
            }

            await github.rest.issues.addLabels({
              owner,
              repo,
              issue_number: prNumber,
              labels: ['api-review']
            });

            console.log(`Detected ${baselineFiles.length} changed baseline file(s) on PR #${prNumber}.`);

      - name: Check out selected commit
        if: steps.detect.outputs.has_baselines == 'true'
        uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0
        with:
          ref: ${{ steps.detect.outputs.target_sha }}
          fetch-depth: 1

      - name: Restore repo-local .NET SDK
        if: steps.detect.outputs.has_baselines == 'true'
        shell: bash
        run: ./restore.sh

      - name: Build ApiChief and compute review diffs
        if: steps.detect.outputs.has_baselines == 'true'
        id: delta
        shell: bash
        env:
          FILES_JSON: ${{ steps.detect.outputs.files_json }}
          OWNER: ${{ github.repository_owner }}
          REPO: ${{ github.event.repository.name }}
          BASE_SHA: ${{ steps.detect.outputs.base_sha }}
          TARGET_SHA: ${{ steps.detect.outputs.target_sha }}
        run: |
          set -euo pipefail

          DOTNET="$PWD/.dotnet/dotnet"
          APICHIEF_PROJECT="$PWD/eng/Tools/ApiChief/ApiChief.csproj"
          WORKDIR="$PWD/artifacts/tmp/api-review"
          mkdir -p "$WORKDIR"

          "$DOTNET" build "$APICHIEF_PROJECT" --configuration Release --nologo --verbosity minimal
          APICHIEF_PATH="$($DOTNET msbuild "$APICHIEF_PROJECT" --getProperty:TargetPath -p:Configuration=Release --nologo | tr -d '\r')"

          export DOTNET APICHIEF_PATH WORKDIR

          python3 - <<'PY'
          import json
          import os
          import pathlib
          import subprocess
          import urllib.error
          import urllib.parse
          import urllib.request

          files = json.loads(os.environ['FILES_JSON'])
          owner = os.environ['OWNER']
          repo = os.environ['REPO']
          base_sha = os.environ['BASE_SHA']
          target_sha = os.environ['TARGET_SHA']
          dotnet = os.environ['DOTNET']
          apichief = os.environ['APICHIEF_PATH']
          workdir = pathlib.Path(os.environ['WORKDIR'])
          workdir.mkdir(parents=True, exist_ok=True)

          def download_file(ref: str, path: str, destination: pathlib.Path) -> bool:
              encoded_path = urllib.parse.quote(path, safe='/')
              url = f'https://raw.githubusercontent.com/{owner}/{repo}/{ref}/{encoded_path}'
              request = urllib.request.Request(url, headers={'User-Agent': 'efcore-api-review-workflow'})

              try:
                  with urllib.request.urlopen(request) as response:
                      destination.write_bytes(response.read())
                  return True
              except urllib.error.HTTPError as error:
                  if error.code == 404:
                      return False
                  raise

          max_comment_length = 60000
          comment_bodies: list[str] = []
          current_body = ''

          for index, file in enumerate(files):
              filename = file['filename']
              old_path = workdir / f'{index}.old.baseline.json'
              new_path = workdir / f'{index}.new.baseline.json'
              review_dir = workdir / f'{index}.review'

              old_exists = download_file(base_sha, filename, old_path)
              new_exists = download_file(target_sha, filename, new_path)

              section = ''
              if old_exists and new_exists:
                  result = subprocess.run(
                      [dotnet, apichief, str(new_path), 'emit', 'delta', str(old_path), '--diff', '-o', str(review_dir)],
                      capture_output=True,
                      text=True,
                      check=False)

                  if result.returncode == 2:
                      continue

                  if result.returncode != 0:
                      raise RuntimeError(
                          f'ApiChief delta diff failed for {filename} with exit code {result.returncode}:\n'
                          f'{result.stdout}\n{result.stderr}')

                  review_files = sorted(review_dir.rglob('*.md'))
                  if not review_files:
                      continue

                  section = (
                      f"## API review baseline changes for `{filename}`\n\n"
                      "<details>\n<summary><strong>Show diff</strong></summary>\n\n"
                      'The diff below was generated by `ApiChief` between the base and the PR.\n\n')

                  for review_file in review_files:
                      section += (
                          f"{review_file.read_text(encoding='utf-8').rstrip()}\n\n")

                  section = section.rstrip() + '\n\n</details>'
              elif new_exists:
                  section = (
                      f"## API review baseline changes for `{filename}`\n\n"
                      'This baseline file was **added** in the PR.')
              elif old_exists:
                  section = (
                      f"## API review baseline changes for `{filename}`\n\n"
                      'This baseline file was **removed** in the PR.')

              if not section:
                  continue

              if len(section) > max_comment_length:
                  section = (
                      f"## API review baseline changes for `{filename}`\n\n"
                      ':warning: The generated diff for this baseline file exceeds the GitHub comment size limit '
                      'and cannot be displayed. Review the baseline changes locally.')

              if not current_body:
                  current_body = section
              elif len(current_body) + 2 + len(section) > max_comment_length:
                  comment_bodies.append(current_body)
                  current_body = section
              else:
                  current_body += '\n\n' + section

          if not current_body:
              current_body = (
                  '## API review baseline changes\n\n'
                  'No API deltas were produced for the modified baseline files.')

          if current_body:
              comment_bodies.append(current_body)

          comments_dir = workdir / 'comments'
          comments_dir.mkdir(parents=True, exist_ok=True)

          for index, body in enumerate(comment_bodies, start=1):
              comment_path = comments_dir / f'{index:03}.md'
              comment_path.write_text(body.rstrip() + '\n', encoding='utf-8')

          with open(os.environ['GITHUB_OUTPUT'], 'a', encoding='utf-8') as output:
              output.write(f'comments_dir={comments_dir}\n')
          PY

      - name: Create PR comment with diffs
        if: steps.detect.outputs.has_baselines == 'true'
        uses: actions/github-script@3a2844b7e9c422d3c10d287c895573f7108da1b3
        env:
          COMMENTS_DIR: ${{ steps.delta.outputs.comments_dir }}
          PR_NUMBER: ${{ steps.detect.outputs.pr_number }}
        with:
          script: |
            const fs = require('fs');
            const path = require('path');
            const owner = context.repo.owner;
            const repo = context.repo.repo;
            const issue_number = Number(process.env.PR_NUMBER);
            const commentsDir = process.env.COMMENTS_DIR;

            const commentFiles = fs.readdirSync(commentsDir)
              .filter(file => file.endsWith('.md'))
              .sort((a, b) => a.localeCompare(b));

            for (const file of commentFiles) {
              const body = fs.readFileSync(path.join(commentsDir, file), 'utf8');
              await github.rest.issues.createComment({
                owner,
                repo,
                issue_number,
                body
              });
            }

            console.log(`Created ${commentFiles.length} API review comment(s) for PR #${issue_number}.`);