Pytest Reference

A pytest fixture is a function decorated with `@pytest.fixture` that provides setup data or resources to test functions. Fixtures promote reuse — you define shared logic (like creating a database connection, a test client, or sample data) once and inject it into any test function by naming it as a parameter. Use fixtures whenever multiple tests share the same setup code. For cleanup, use `yield` inside the fixture: code before `yield` runs before the test, and code after `yield` runs as teardown.

Fixture scope controls how often a fixture is created and destroyed. The default scope is "function", meaning a new fixture instance is created for each test. "module" scope creates one instance shared across all tests in a module. "session" scope creates one instance shared across the entire test run. Use wider scopes for expensive resources like database connections or application servers, and narrow scopes (function) for anything that must be isolated between tests.

`@pytest.mark.parametrize` lets you run the same test function with multiple sets of inputs. You specify argument names as a comma-separated string and a list of value tuples: `@pytest.mark.parametrize("x,y,expected", [(1, 2, 3), (5, 5, 10)])`. pytest generates a separate test case for each tuple. You can add multiple `@parametrize` decorators to get the Cartesian product of all combinations. Use `pytest.param(..., id="name")` to give test cases human-readable IDs in the output.

`monkeypatch` is a pytest built-in fixture that temporarily replaces attributes, environment variables, dictionary entries, and file system paths for the duration of a single test, then automatically restores the originals after the test. It is simpler for straightforward attribute replacement. `unittest.mock.patch` (and `MagicMock`) is more powerful for creating full mock objects with return values, call tracking, and side effects. Both approaches work well in pytest; `pytest-mock` provides a `mocker` fixture that wraps `unittest.mock` in the pytest style.

Install the `pytest-cov` plugin (`pip install pytest-cov`) and run `pytest --cov=myapp --cov-report=html tests/`. This generates an HTML coverage report showing which lines are covered by your tests. You can also use `--cov-report=term-missing` for a terminal output that shows uncovered line numbers. Configure coverage settings in `pyproject.toml` or `.coveragerc` to exclude files, set a minimum coverage threshold, or control branch coverage.

Use `@pytest.mark.skip(reason="...")` to unconditionally skip a test. Use `@pytest.mark.skipif(condition, reason="...")` to skip based on a runtime condition, such as the OS platform or a missing optional dependency. Use `@pytest.mark.xfail(reason="...")` to mark a test as an expected failure — pytest will report it as "xfailed" if it fails (expected) or "xpassed" if it unexpectedly passes, which can be treated as an error with `xfail_strict = true` in your config.

Use `pytest --lf` (or `--last-failed`) to run only the tests that failed in the previous run. pytest stores results in a `.pytest_cache` directory. If no tests failed last time, `--lf` falls back to running the full suite. You can combine it with `-x` (stop on first failure) as `pytest --lf -x` to quickly iterate on a failing test without running the entire suite each time.

pytest can be configured in `pytest.ini`, `pyproject.toml` (under `[tool.pytest.ini_options]`), or `setup.cfg` (under `[tool:pytest]`). Common settings include `testpaths` to specify where tests live, `addopts` to set default CLI flags (like `-v --tb=short`), `markers` to register custom markers (required with `--strict-markers`), and `filterwarnings` to treat specific warnings as errors or ignore them. The `conftest.py` file is used to share fixtures and hook functions across multiple test files.