python-testing

Follow the red-green-refactor cycle: write a failing test for the desired behavior first, implement the minimal code to make it pass, then refactor while keeping tests green. Each test should verify a single behavior with a descriptive name.

Use unittest.mock's @patch decorator targeting the import path of the external call, then set return_value or side_effect on the mock. Verify interactions with assert_called_once or assert_awaited_once for async functions.

Yes, with pytest-asyncio you mark tests with @pytest.mark.asyncio and use async def. Async fixtures can yield resources like async HTTP clients, and async mocks are verified with assert_awaited_once.

Mark slow tests with @pytest.mark.slow, register the marker in pytest.ini or pyproject.toml, then run pytest -m "not slow". You can combine markers, such as pytest -m "unit and not slow".

Target 80%+ overall coverage with 100% on critical paths. Measure it with pytest --cov=mypackage --cov-report=term-missing to see uncovered lines, and generate HTML reports for detailed inspection.

Shared state makes tests order-dependent and produces flaky, hard-to-debug failures. Use function-scoped fixtures so each test gets fresh setup, and rely on teardown or tmp_path for automatic cleanup.