Python Testing

Philosophy

• Tests document expected behavior—prioritize clarity over cleverness
• Test behavior at boundaries, not internal implementation
• Ask before writing: "What bug would this test catch?" If no clear answer, skip it

Rules

• Must follow AAA pattern: Arrange, Act, Assert (clearly separated)
• Should use @pytest.mark.parametrize for testing multiple inputs
• Should use markers (@pytest.mark.slow, @pytest.mark.integration) for categorization

Tooling

Need	Use	Avoid
Mocking	pytest-mock (mocker fixture)	unittest.mock.patch directly
HTTP mocking	respx (for httpx)	Manual response mocks
Async testing	pytest-asyncio with asyncio_mode = "auto"	Manual event loop management
DB testing	In-memory SQLite (aiosqlite) or testcontainers	Mocking the ORM

Fixtures

• Must use fixtures for reusable setup—extract if same mock appears in 2+ tests
• Must use httpx.AsyncClient with ASGITransport for FastAPI test client
• Should scope fixtures appropriately: function default, module for expensive setup
• Avoid autouse=True except for truly universal setup

Test Organization

tests/
├── conftest.py      # Shared fixtures
├── unit/            # Fast, isolated, mock all externals
└── integration/     # Real dependencies, slower

What to Test

• Do test: Business logic, edge cases, error paths, API contracts, regressions
• Avoid testing: Framework behavior, library internals, trivial getters/setters
• Never write tests just to hit coverage numbers