Patch Linter

Static analyzer that forbids unittest.mock.patch and pytest monkeypatch mutations in tests. Prefer real dependency injection over patching.

Why

Patching reaches into other modules and couples tests to implementation details. Refactoring becomes harder, and failures become brittle. Inject dependencies (fakes, stubs, or test doubles passed as arguments) instead.

What is reported

Not reported (by design): mocker (pytest-mock), bare Mock / MagicMock, freezegun, your own functions named patch, HTTP client.patch(...).

Code: PATCH001.

Installation

pip install patch-linter
# optional flake8 plugin:
pip install "patch-linter[flake8]"

Usage

CLI

patch-linter tests/
patch-linter tests/unit tests/test_foo.py

Output (clickable in Cursor / VS Code / PyCharm):

/abs/path/to/tests/test_foo.py:12:5: PATCH001 Patch usage in tests: @patch('module.function')

Exit code 1 if problems are found.

Flake8

flake8 --select=PATCH tests/

Suppress a line

@patch("module.function")  # noqa: PATCH001
def test_legacy(mock_fn):
    ...

Bare # noqa also suppresses the finding on that line.

Correct approaches

def process_data(repository):
    return repository.get_all()


def test_process_data():
    class FakeRepo:
        def get_all(self):
            return [1, 2, 3]

    assert process_data(FakeRepo()) == [1, 2, 3]

License

MIT