mutation-testing

A Python mutation testing framework that validates test suite quality by injecting runtime mutations and checking whether your tests catch them.

Mutations are applied at runtime using AST pattern matching — source files are never modified.

How it works

1. You define mutations: intentional code changes like replacing + with - or > with >=
2. The framework injects each mutation into the running code by swapping the function's __code__ object
3. Your test suite runs against the mutated code
4. If a test fails, the mutation was killed (good — your tests caught the bug)
5. If all tests pass, the mutation survived (bad — your tests have a gap)

The mutation score is the percentage of mutations killed. A higher score means stronger tests.

Installation

Requires Python 3.10+.

pip install mutation-testing

# Or with uv
uv add mutation-testing

For development:

git clone https://github.com/YeahWick/mutation-testing.git
cd mutation-testing
pip install -e ".[dev]"

Quick start

1. Define mutations in YAML

Create a mutations.yaml file:

version: "1.0"

settings:
  timeout: 30

targets:
  - module: "calculator"
    file: "src/calculator.py"
    mutations:
      - id: "add-001"
        function: "add"
        description: "Replace + with -"
        original: "return a + b"
        mutant: "return a - b"

      - id: "pos-001"
        function: "is_positive"
        description: "Replace > with >="
        original: "return x > 0"
        mutant: "return x >= 0"

2. Write a test runner

from mutation_testing import MutationRunner

def run_tests() -> bool:
    """Return True if all tests pass."""
    try:
        assert calculator.add(2, 3) == 5
        assert calculator.add(-1, 1) == 0
        assert calculator.is_positive(5) is True
        assert calculator.is_positive(-1) is False
        return True
    except AssertionError:
        return False

runner = MutationRunner(run_tests)
report = runner.run_from_config("mutations.yaml")

3. Run it

cd example
python run_mutations.py

Output:

============================================================
MUTATION TESTING
============================================================

[✓] [add-001] Replace + with -: KILLED
[✓] [add-002] Replace + with *: KILLED
[✓] [sub-001] Replace - with +: KILLED
[✗] [pos-001] Replace > with >=: SURVIVED
[✓] [pos-002] Replace > with <: KILLED
[✓] [clamp-001] Off-by-one in lower bound: KILLED
[✗] [clamp-002] Off-by-one in upper bound: SURVIVED

============================================================
SUMMARY
============================================================
Total mutations:  7
Killed:           5
Survived:         2
Mutation Score:   71.4%
============================================================

SURVIVING MUTATIONS (improve your tests!):
  - [pos-001] Replace > with >=
  - [clamp-002] Off-by-one in upper bound

The surviving mutations tell you exactly where your tests are weak — in this case, missing boundary checks for is_positive(0) and clamp at its upper bound.

API

Defining mutations explicitly

Instead of YAML, you can define mutations in code:

from mutation_testing import MutationRunner, Mutation

runner = MutationRunner(run_tests)
report = runner.run(
    mutations=[
        Mutation(
            id="add-001",
            function="add",
            original="return a + b",
            mutant="return a - b",
            description="Replace + with -",
        ),
    ],
    module_name="calculator",
)

print(f"Score: {report.score:.1%}")
print(f"All killed: {report.all_killed}")

Core classes

Low-level API

For direct control over injection:

from mutation_testing import MutationInjector

injector = MutationInjector()

# Inject a mutation
injector.inject("calculator", "add", "return a + b", "return a - b")

# Run your tests here...

# Restore the original
injector.restore("calculator", "add")

# Or restore everything at once
injector.restore_all()

Mutation coverage report

The coverage report tracks which of your test functions have corresponding mutations defined. This helps you ensure every test is validated by at least one mutation, and can fail CI when coverage drops below a threshold.

Configure in YAML

Add a coverage section to your mutations.yaml:

coverage:
  threshold: 100.0        # minimum percentage of tests that must have mutations
  fail_under: true         # exit non-zero if threshold not met
  test_paths:
    - "tests"
  test_mappings:           # optional: override the default naming convention
    test_boundary_check:
      - "clamp"
      - "is_positive"

By default, tests are mapped to source functions by naming convention — test_add maps to add, test_is_positive maps to is_positive. Use test_mappings when a test name doesn't match its source function.

Generate the report

from mutation_testing import MutationRunner

runner = MutationRunner(run_tests)
coverage = runner.coverage_from_config("mutations.yaml")

# Check results programmatically
if not coverage.meets_threshold:
    sys.exit(1)

Or use the standalone function:

from mutation_testing import generate_coverage_report, print_coverage_report

report = generate_coverage_report(
    test_paths=["tests"],
    mutation_config_path="mutations.yaml",
    threshold=100.0,
)
print_coverage_report(report)

# JSON output for CI integration
print(report.to_json())

Output

============================================================
MUTATION COVERAGE REPORT
============================================================

[✓] test_add
      functions: add
      mutations: 2 (add-001, add-002)
[✓] test_subtract
      functions: subtract
      mutations: 1 (sub-001)
[✗] test_multiply
      functions: multiply
      mutations: 0

============================================================
COVERAGE SUMMARY
============================================================
Total tests:     3
Covered:         2
Uncovered:       1
Coverage:        66.7%
Threshold:       100.0%
Status:          FAIL
============================================================

UNCOVERED TESTS (add mutations for these!):
  - test_multiply -> multiply

CI usage

Use the exit code to fail CI pipelines:

coverage = runner.coverage_from_config("mutations.yaml")
sys.exit(0 if coverage.meets_threshold else 1)

Coverage classes

Supported mutation patterns

The AST pattern matcher supports any valid Python expression or statement:

• Arithmetic operators: + ↔ -, *, /, //
• Comparison operators: > ↔ >=, < ↔ <=, == ↔ !=
• Boolean operators: and ↔ or
• Return values: return x → return None, return -x
• Any expression parseable as Python

Patterns are matched structurally via AST, so whitespace and formatting differences are ignored.

Project structure

mutation_testing/
├── __init__.py     # Public API exports
├── core.py         # AST pattern matching, injection engine, Mutation/MutationResult
├── config.py       # YAML configuration loader
├── coverage.py     # Mutation coverage reporting
└── runner.py       # MutationRunner and MutationReport

example/
├── src/calculator.py       # Sample module under test
├── tests/test_calculator.py  # Sample test suite
├── mutations.yaml          # Sample mutation config
└── run_mutations.py        # Example runner script

Key concepts

License

MIT