moai-ref-testing-pyramid

The recommended distribution is 70% unit tests, 20% integration tests, and 10% E2E tests. Unit tests run fast and cover pure logic, integration tests verify API and database boundaries, and E2E tests cover only critical user journeys.

The AAA pattern divides each test into Arrange, Act, and Assert phases. Arrange sets up test data and preconditions, Act executes the function under test, and Assert verifies the expected outcome.

Coverage targets depend on context: 95%+ for critical business logic, 90%+ for API endpoints, 85%+ for utilities, 80%+ for UI components, and 60%+ for configuration code. Generated code requires no coverage.

Unit tests should mock external dependencies like databases and APIs at boundaries to stay fast and deterministic. Reserve real dependencies for integration tests using tools like testcontainers or in-memory databases.

Coverage measures execution, not correctness. A test that executes code without meaningful assertions provides zero value, so pair coverage targets with mutation scores and behavior-focused assertions.

Flaky tests erode trust in the entire suite and should stay below a 1% rate. Fix the flakiness immediately or quarantine the test with a tracking issue, and verify stability by running the suite three times.