How do pytest fixtures improve test setup/teardown compared to unittest's setUp/tearDown?
Fixtures are composable, scope-aware dependency injection instead of fixed per-class hooks
unittest setUp and tearDown are fixed hooks tied to a TestCase class. They run for every test method, they cannot be parameterized per test, and sharing them across classes requires inheritance. pytest fixtures are functions that a test requests by name, so setup is composed by dependency injection rather than by inheritance. A fixture can be function, class, module, package, or session scoped, which means an expensive resource like a database container can be created once per session and injected everywhere. Fixtures can depend on other fixtures, so a db_session can depend on a connection which depends on a container. Teardown is expressed as code after the yield inside the fixture, and pytest guarantees it runs even if the test fails.
- 1
Request fixtures by naming them in the test signature; pytest resolves the dependency graph.
- 2
Scope controls lifetime: function (default), class, module, package, or session.
- 3
Fixtures can be parameterized, which generates one test per parameter set without duplicating code.
- 4
conftest.py exposes fixtures to an entire directory tree without imports.
- 5
autouse fixtures apply implicitly when you need global setup such as environment variables.
- 6
Trade-off: session-scoped fixtures are fast but shared state can leak between tests. Function-scoped fixtures are safer but slower.
- 7
Common mistake: putting mutable state in a session fixture and having one test corrupt another.
- 8
Common mistake: calling teardown code after yield that can itself raise, masking the original test failure.
- 9
Version note: fixtures have been core to pytest since early versions. pytest 7+ and 8+ refined scoping and warning behavior but the API is stable.
Scenario Questions
0-2 years experience
- 1How does a test tell pytest which fixture to use?
- 2Where does teardown code live in a pytest fixture?
2-5 years experience
- 1You have a slow database container and want to start it once for the whole suite. How do you do it with fixtures?
- 2You need the same test to run against three different configurations. How do you parameterize a fixture?
5-8 years experience
- 1A session-scoped fixture leaks state between tests and one test corrupts another. How do you diagnose and fix it?
- 2You need a fixture that yields a resource and guarantees cleanup even when the test fails inside a nested context. How do you structure it?
8+ years experience
- 1Design a fixture hierarchy for a large test suite with layered resources (process, connection, transaction, data) that keeps isolation while staying fast.
- 2Explain how fixture finalization, scope mismatches, and pytest-asyncio interact, and how to avoid resource leaks in async tests.
Follow-up Questions
- When is a session-scoped fixture the wrong choice?
- How do you override a fixture for a single test or module?