main.yml

---
name: CI & Release

# Workflow name based on selected inputs. Fallback to default Github naming when expression evaluates to empty string
run-name: >-
  ${{
    inputs.release && inputs.test && format('Build {0} ➤ Test ➤ Publish to NPM', github.ref_name)  ||
    inputs.release && !inputs.test && format('Build {0} ➤ Skip Tests ➤ Publish to NPM', github.ref_name) ||
    github.event_name == 'workflow_dispatch' && inputs.test && format('Build {0} ➤ Test', github.ref_name) ||
    github.event_name == 'workflow_dispatch' && !inputs.test && format('Build {0} ➤ Skip Tests', github.ref_name) ||
    ''
  }}

on:
  # Build on pushes branches that have a PR (including drafts)
  pull_request:
  # Build on commits pushed to branches without a PR if it's in the allowlist
  push:
    branches: [main]
  # https://docs.github.com/en/actions/managing-workflow-runs/manually-running-a-workflow
  workflow_dispatch:
    inputs:
      test:
        description: Run tests
        required: true
        default: true
        type: boolean
      release:
        description: Release new version
        required: true
        default: false
        type: boolean

permissions:
  contents: read # for checkout

concurrency:
  # On PRs builds will cancel if new pushes happen before the CI completes, as it defines `github.head_ref` and gives it the name of the branch the PR wants to merge into
  # Otherwise `github.run_id` ensures that you can quickly merge a queue of PRs without causing tests to auto cancel on any of the commits pushed to main.
  group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }}
  cancel-in-progress: true

jobs:
  build:
    runs-on: ubuntu-latest
    name: Lint & Build
    steps:
      - uses: actions/checkout@v3
      - uses: actions/setup-node@v3
        with:
          cache: npm
          node-version: lts/*
      - run: npm ci
        # Linting can be skipped
      - run: npm run lint --if-present
        if: github.event.inputs.test != 'false'
        # But not the build script, as semantic-release will crash if this command fails so it makes sense to test it early
      - run: npm run prepublishOnly --if-present

  test:
    needs: build
    # The test matrix can be skipped, in case a new release needs to be fast-tracked and tests are already passing on main
    if: github.event.inputs.test != 'false'
    runs-on: ${{ matrix.os }}
    name: Node.js ${{ matrix.node }} / ${{ matrix.os }}
    strategy:
      # A test failing on windows doesn't mean it'll fail on macos. It's useful to let all tests run to its completion to get the full picture
      fail-fast: false
      matrix:
        # Run the testing suite on each major OS with the latest LTS release of Node.js
        os: [macos-latest, ubuntu-latest, windows-latest]
        node: [lts/*]
        # It makes sense to also test the oldest, and latest, versions of Node.js, on ubuntu-only since it's the fastest CI runner
        include:
          - os: ubuntu-latest
            # Test the oldest LTS release of Node that's still receiving bugfixes and security patches, versions older than that have reached End-of-Life
            node: lts/-2
          - os: ubuntu-latest
            # Test the actively developed version that will become the latest LTS release next October
            node: current
    steps:
      # It's only necessary to do this for windows, as mac and ubuntu are sane OS's that already use LF
      - name: Set git to use LF
        if: matrix.os == 'windows-latest'
        run: |
          git config --global core.autocrlf false
          git config --global core.eol lf
      - uses: actions/checkout@v3
      - uses: actions/setup-node@v3
        with:
          cache: npm
          node-version: ${{ matrix.node }}
      - run: npm i
      - run: npm test --if-present

  release:
    needs: [build, test]
    # only run if opt-in during workflow_dispatch
    if: always() && github.event.inputs.release == 'true' && needs.build.result != 'failure' && needs.test.result != 'failure' && needs.test.result != 'cancelled'
    runs-on: ubuntu-latest
    name: Semantic release
    permissions:
      contents: write # to be able to publish a GitHub release
      issues: write # to be able to comment on released issues
      pull-requests: write # to be able to comment on released pull requests
      id-token: write # to enable use of OIDC for npm provenance
    steps:
      - uses: actions/checkout@v3
        with:
          # Need to fetch entire commit history to
          # analyze every commit since last release
          fetch-depth: 0
      - uses: actions/setup-node@v3
        with:
          cache: npm
          node-version: lts/*
      - run: npm ci
        # Branches that will release new versions are defined in .releaserc.json
      - run: npx semantic-release
        # Don't allow interrupting the release step if the job is cancelled, as it can lead to an inconsistent state
        # e.g. git tags were pushed but it exited before `npm publish`
        if: always()
        env:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
          NPM_TOKEN: ${{ secrets.NPM_PUBLISH_TOKEN }}
        # Re-run semantic release with rich logs if it failed to publish for easier debugging
      - run: npx semantic-release --dry-run --debug
        if: failure()
        env:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
          NPM_TOKEN: ${{ secrets.NPM_PUBLISH_TOKEN }}