Skip to content

Latest commit

 

History

History
62 lines (37 loc) · 3.62 KB

README.md

File metadata and controls

62 lines (37 loc) · 3.62 KB

Tests

Tests should automatically be run on every PR via Github CI, but it's recommended to also run them locally.

Running Tests

First you must configure your development environment per ../CONTRIBUTING.md

Then run:

npm test

Note that for Linux you'll need to prefix test commands with xvfb-run -a for example xvfb-run -a npm run test Note that some tests rely on the built project and will not pass until you have run npm run build-dist.

To run individual tests:

  • Unit tests: npx jest path/to/file.test.ts (e.g. npx jest src/style/style_layer.test.ts)
  • Render tests: npm run test-render -- render-test-name (e.g. npm run test-render -- text-rotation-alignment)

To run folders in watch mode, meaning they will run continuously as you make changes to relevant code, (i.e. for test driven development): use npm run test-watch-roots *folder1* [*folder2*...] (e.g. npm run test-watch-roots ./src/ui/control)

Debugging Tests in VSCode

Steps to use Visual Studio Code:

  • Install VS Code.

  • Open MapLibre root folder with VS Code UI.

  • Start debugging:

    Start debugging Before debugging, you might want to set break points in sources.

You may also run tests and view output with the plugin orta.vscode-jest:

VScode screenshots

If you do so, note that the initial test run may take a while (3 minutes or so), during which time the test results are not yet available.

Integration Tests

See test/integration/README.md.

Writing Unit Tests

  • You must not share variables between test cases. All test fixtures must be wrapped in create functions. This ensures each test is run in an isolated environment.
  • You should not mock any internal domain objects. Internal domain objects include Style, Map, Transform, and Dispatcher. If this is difficult because of some interface, refactor that interface. This ensures that tests accurately exercise the code paths used in production.
  • You should test one return value or side effect per test case. Feel free to pull shared logic into a function. This ensures that tests are easy to understand and modify.
  • You should only test the return values and global side effects of methods. You should not not test internal behavior, such as that another method is called with particular arguments. This ensures that method implementations may change without causing test failures.
  • You must not make network requests in test cases. This rule holds in cases when result isn't used or is expected to fail. You may use fakeServer.create() per the nise (Sinon) API to simulate network requests. This ensures that tests are reliable, able to be run in an isolated environment, and performant.
  • You should use clear input space partitioning schemes. Look for edge cases! This ensures that tests suites are comprehensive and easy to understand.
  • Before submit or modify tests, suggest to run the tests and verify your results on both Windows and Linux CI (e.g. WSL).

Spies, Stubs, and Mocks

The test object is augmented with methods from Sinon.js for spies, stubs, and mocks. For example, to use Sinon's spy API, call t.spy(...) within a test.

The test framework is set up such that spies, stubs, and mocks on global objects are restored at the end of each test.