python-testing

Python Testing

Modern testing patterns with pytest for reliable, maintainable test suites.

Core Principles

1. Use pytest over unittest - Better fixtures, assertions, parametrization
2. Arrange-Act-Assert pattern - Clear test structure
3. Test behavior, not implementation - Focus on public APIs
4. Fixtures for setup - Reusable, composable test state

Basic Test Structure

def test_process_data_filters_correctly():
    # Arrange
    data = pd.DataFrame({"score": [1, 5, 10]})
    
    # Act
    result = process_data(data, threshold=5)
    
    # Assert
    assert len(result) == 1
    assert result["score"].iloc[0] == 10

Fixtures

Use fixtures for reusable test setup

import pytest

@pytest.fixture
def sample_data():
    """Provide sample DataFrame for tests."""
    return pd.DataFrame({
        "id": [1, 2, 3],
        "value": [10, 20, 30]
    })

def test_with_fixture(sample_data):
    result = process_data(sample_data)
    assert len(result) == 3

Fixture scopes:

• scope="function" (default) - Per test
• scope="class" - Per test class
• scope="module" - Per file
• scope="session" - Once per test run

See fixture-patterns.md for:

• Fixture composition
• Parametrized fixtures
• Fixture cleanup (yield pattern)
• conftest.py organization

Parametrization

Test multiple inputs efficiently

@pytest.mark.parametrize("input,expected", [
    (5, 25),
    (0, 0),
    (-3, 9),
])
def test_square(input, expected):
    assert square(input) == expected

# Multiple parameters
@pytest.mark.parametrize("threshold", [0.5, 0.8])
@pytest.mark.parametrize("method", ["linear", "cubic"])
def test_interpolation(threshold, method):
    result = interpolate(data, threshold, method)
    assert result is not None

See parametrization-examples.md for:

• Complex parameter combinations
• Parametrizing fixtures
• Using pytest.param for IDs
• Indirect parametrization