python-mutation-testing

Mutation testing introduces small deliberate bugs into source code one at a time and runs the test suite after each change. A mutation is killed if a test fails and survived if all tests pass, revealing gaps in your test suite.

The mutation score is killed mutations divided by total mutations, such as 6/8 = 75%. Higher scores indicate stronger tests; each survived mutation represents a behavior change your tests cannot detect and should prompt a new test.

Yes, the workflow detects pytest via pytest.ini, pyproject.toml configuration, or a tests directory. It confirms the suite passes before starting and runs it after each mutation to record killed or survived results.

Common mutations include negating conditions, changing comparison boundaries, swapping return values, deleting side effects or guard clauses, changing arithmetic operators, modifying default arguments, and swapping argument order in non-commutative calls.

A clean working tree ensures mutations can be safely reverted with git checkout after each test run. Uncommitted changes would be lost or mixed with mutations, so you must commit or stash before starting.

Mutation testing only mutates production code, never test files, imports, type annotations, or docstrings. It applies 3-8 mutations per file for quality over quantity, so it samples behavior rather than exhaustively testing every possible change.