Migrate frontend test framework from Karma to Vitest

[Yicong-Huang]

Task Summary

Replace Karma + Jasmine with Vitest for the Angular 21 frontend. Removes a deprecated runner, modernizes the test setup, and keeps texera on the path Angular's tooling team is investing in.

Why now, why Vitest

Karma is deprecated. npm-published deprecation since April 2023; Angular's official builder treats @angular-devkit/build-angular:karma as end-of-life. We are still on karma@6.4.4 + karma-jasmine@5.1.0.

Three options were considered:

1. Stay on Karma. Increasingly broken on each Angular major (we have already shipped one workaround in karma.conf.js for the Angular --code-coverage injection). Defers the inevitable.
2. Migrate to Jest. PR Migrate frontend test framework from Karma to Jest #2995 (Nov 2024, closed unmerged) attempted this. Path works, but Angular's Jest support is officially "experimental" and frozen — no further investment from the team. The PR's own checklist captured one pain point that never got resolved: "ignore transformation of ESM modules" — Jest needs transformIgnorePatterns per ESM dependency in node_modules, a recurring tax on every dependency bump.
3. Migrate to Vitest. Angular's actively-developed test runner path (Angular 19+); ESM-native so the Jest pain point doesn't apply; Vite-powered for fast watch; expect API is Jest-compatible so most matcher / spy code reads identically.

We are choosing option 3 — Vitest — because picking Jest in 2026 inherits Karma's fate in 2-3 years.

Known friction points

• frontend/custom-webpack.config.js is webpack-specific. Any test-time aliases / loaders / env-var injection there have to be re-expressed for Vite.
• We use Nx (nx.json, @nx/angular) as a thin cache wrapper. The @nx/vitest executor is younger than @nx/jest. Worst case: bypass Nx for the test target and use ng test --use=vitest directly.
• 73 *.spec.ts files to migrate. Most should be a near-mechanical sweep:
  • jasmine.createSpy / jasmine.Spy → vi.fn() / Mock from vitest
  • spyOn(obj, 'method').and.returnValue(...) → vi.spyOn(obj, 'method').mockReturnValue(...)
  • jasmine.clock() → vi.useFakeTimers()
  • expect(x).toEqual(y) etc. unchanged
• Coverage: switch karma-coverage (lcov via Karma) to Vitest's coverage (v8 by default). The Codecov upload glob already targets frontend/coverage/**/lcov.info, which Vitest's coverage reporter can produce when configured.

Recoverable from PR #2995

Most of #2995's 3,959 line diff is reusable: removing karma packages, fixing wrong test-file imports (e.g., HttpClient), deleting empty test files. Only the runner config and jest.fn → vi.fn swap differ. Roughly 70–80% of the work transfers.

Plan

Land in stages, smallest first:

1. Scaffolding PR (this issue's first deliverable): add Vitest deps + config, switch the ng test target, remove Karma deps and karma.conf.js. Goal: existing 73 spec files run under Vitest, even if some need shims.
2. Mechanical migration sweeps: batch updates for jasmine.* → vi.* patterns, in chunks small enough to review.
3. Cleanup PR: delete empty / dead spec files surfaced during the sweep, fix the transformIgnorePatterns-equivalent ESM issues that PR Migrate frontend test framework from Karma to Jest #2995 left open.
4. CI rewire: update frontend/karma.conf.js references in .github/workflows/build.yml and Codecov upload paths if needed.

References

• PR Migrate frontend test framework from Karma to Jest #2995 — closed Karma → Jest attempt (preserved for diff context)
• Karma deprecation: https://www.npmjs.com/package/karma

Task Type

• Refactor / Cleanup
• DevOps / Deployment / CI
• Testing / QA
• Documentation
• Performance
• Other

[Yicong-Huang]

I will handle this issue as it is a large framework change.