tests.yml

name: Tests

# Runs on pushes to main as well as PRs. This used to be pull_request-only,
# which meant it silently stopped running once work started going straight to
# main -- it last ran in April while main kept moving.
on:
  push:
    branches:
      - main
  pull_request:

permissions:
  contents: read

jobs:
  unit:
    name: Unit tests
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v7

      - uses: actions/setup-node@v7
        with:
          node-version: 24
          cache: npm

      - name: Install dependencies
        run: npm ci --legacy-peer-deps

      # No build step needed: no remaining spec imports the workspace libraries
      # from dist/, so this passes on a clean checkout.
      - name: Test
        run: npm test

      # The type-aware lint rules and knip both need the workspace libraries to
      # resolve. Without dist/ they report against a different graph than they do
      # locally, so the baselines drift -- which already failed one run with
      # "Cannot find module 'reactome-gsa-form'". Build them first so CI and a
      # developer's machine are looking at the same code.
      - name: Build workspace libraries
        run: npm run build:libs

      # The app build is the only thing that runs Angular's template type
      # checking, and strictTemplates is on: a template referencing a member
      # that does not exist, binding an input on a component it never imported,
      # or using a pipe that is not in scope fails here. None of that shows up
      # in the unit tests -- a merge once landed a dead @if block that compiled
      # and passed every test. Without this step the only build was the deploy
      # job, so a broken template reached main and was found by a person.
      #
      # content-dist is a gitignored asset input, so stage:content has to run or
      # ng build cannot find it. The libraries no longer need a step of their
      # own: `npm run build` builds them, because a build that does not produce
      # what it links against is a build that can pass here and ship something
      # else -- a fix to reactome-cytoscape-style reached CI and never reached
      # the dev server, which builds the app without building the library.
      - name: Compile CMS content
        run: npm run stage:content

      - name: Build the app
        run: npm run build

      # Lint errors must be zero -- eslint.config.js only marks a rule "error"
      # once the codebase is clean of it, so an error is always a new violation.
      # Warnings are the ~1600 pre-existing findings, held to a baseline so they
      # can only go down.
      # Formatting is checked, not applied. The previous workflow ran prettier
      # on PR branches and pushed a commit back -- it used yarn in an npm repo
      # and prettier was never a dependency, so it silently formatted nothing
      # for months, and it could not run at all once work went straight to main.
      - name: Check formatting
        run: npm run format:check

      - name: Lint
        run: npm run check:lint

      # eslint sees unused symbols within a file; knip sees whole files and
      # exports nothing imports, which was the bigger pile. Also baselined,
      # because some findings need someone who knows the data model to judge --
      # the graph classes are referenced by schemaClass name, not by import.
      - name: Check unreachable code has not increased
        run: npm run check:dead

  e2e:
    # The code suite only: "is the code right?". The release checks -- every
    # top-level pathway draws, every download link resolves, the version and news
    # and statistics match what is being served -- depend on generated release
    # data this runner does not have, and live in release-verification.yml.
    #
    # Still sharded: playwright splits by test, so coverage is unchanged and the
    # wall clock stays a few minutes rather than the 30 that got a job killed.
    name: End-to-end (${{ matrix.shard }}/4)
    runs-on: ubuntu-latest
    timeout-minutes: 25
    strategy:
      fail-fast: false
      matrix:
        shard: [1, 2, 3, 4]
    steps:
      - uses: actions/checkout@v7

      - uses: actions/setup-node@v7
        with:
          node-version: 24
          cache: npm

      - name: Install dependencies
        run: npm ci --legacy-peer-deps

      # src/styles.scss @use's all three of these from dist/, and the
      # pathway-browser build copies reactome-cytoscape-style's assets, so the
      # app cannot be served until they exist.
      - name: Build workspace libraries
        run: npm run build:libs

      # site-search-index.json is generated, not committed, and the search
      # specs depend on it.
      - name: Generate content indices and compile content
        run: |
          npm run generate:indices
          npm run stage:content

      - name: Install Playwright browser
        run: npx playwright install --with-deps chromium

      # playwright.config.ts starts `ng serve` itself and waits for :4200.
      # Backend calls go through proxy.conf.js, which defaults to the Tomcat on
      # localhost:8080 that the Reactome dev host runs. A CI runner has no such
      # backend, so point it at production or every data-dependent test hangs
      # until the job's 30 minute limit -- which is exactly what happened when
      # the default was switched to localhost.
      #
      # Production lacks the content-page endpoints that only exist on the dev
      # host's unreleased branch; the specs needing those skip themselves rather
      # than fail here.
      - name: Run e2e
        env:
          REACTOME_BACKEND: https://reactome.org
        run: npm run e2e -- --shard=${{ matrix.shard }}/4

      - name: Upload Playwright report
        uses: actions/upload-artifact@v7
        if: ${{ !cancelled() }}
        with:
          # One name per shard: artifacts with the same name collide and the
          # upload fails, losing the report exactly when it is wanted.
          name: playwright-report-${{ matrix.shard }}
          path: playwright-report/
          retention-days: 7