Run one pytest file, test, or class

Pass a test file to run that file, or append :: components to select a function, class, or method. Use the paths and names pytest actually collects.

python -m pytest tests/test_math.py
python -m pytest tests/test_math.py::test_addition
python -m pytest tests/test_math.py::TestMath
python -m pytest tests/test_math.py::TestMath::test_addition

Paths can be absolute or relative to the current working directory. Running from a consistent project directory is convenient, but pytest can discover applicable parent conftest.py files without being launched precisely at the root.

Filter by a keyword expression

python -m pytest -k login
python -m pytest -k "login and not slow"

-k performs case-insensitive matching against collected names and related keywords, including parent names; it is not just an exact function-name comparison. Use a node ID for precision. Use -m for an intentional marker expression rather than treating keyword matching as a substitute for a marker policy.

Add useful run controls

python -m pytest tests/test_math.py -v
python -m pytest tests/test_math.py -x
python -m pytest tests/test_math.py -s
python -m pytest --lf

-v shows more result detail, -x stops after the first failure, and -s disables standard output/error capture in a normal serial run. --lf uses the local cache of previous failures; when there is no such cache, the default is to run the selected tests normally rather than to discover an imaginary prior failure set.

FAQ

Can I select more than one file?

Yes: python -m pytest tests/test_a.py tests/test_b.py. Additional selectors and filters apply within that selection.

How do I select one parametrized case?

Inspect IDs with python -m pytest --collect-only -q, then quote the complete node ID, for example "tests/test_math.py::test_addition[positive]". Quoting prevents shell interpretation of brackets. The parameter ID must match the collected case.

What should I check in an IDE?

Use the test explorer's single-test action, then verify the selected interpreter, working directory, and pytest configuration. An IDE may add its own runner arguments; a similar-looking test label alone does not guarantee the same environment as a terminal run.

References: Running pytest, pytest cache and previous failures.