16. Property-Based Testing with Hypothesis

Architecture Decision Record: 16. Property-Based Testing with Hypothesis

16. Property-Based Testing with Hypothesis

Date: 2025-10-13

Status

Accepted

Category

Testing & Quality

Context

Traditional unit tests check specific examples:

assert calculate_discount(100, 0.1) == 90

Edge cases often missed:

• Negative numbers
• Zero values
• Very large numbers
• Unicode in strings

Decision

Use Hypothesis for property-based testing to discover edge cases automatically.

Example

from hypothesis import given
from hypothesis.strategies import integers, floats

@given(price=floats(min_value=0), discount=floats(min_value=0, max_value=1))
def test_discount_properties(price, discount):
    result = calculate_discount(price, discount)
    assert result >= 0  # Never negative
    assert result <= price  # Never more than original

Hypothesis generates 100+ test cases including edge cases.

Consequences

Positive

• Edge Case Discovery: Finds bugs unit tests miss
• Higher Confidence: Tests properties, not examples
• Less Test Code: One property test replaces many examples
• Regression Prevention: Generated cases saved on failure

Negative

• Slower Tests: 100 cases per test vs 1
• Learning Curve: Property thinking vs example thinking
• Flaky Tests: Random generation can cause intermittent failures

Usage

27+ property tests covering:

• LLM factory fallback logic
• Authentication providers
• Session management
• Role mapping

References

• Tests: tests/property/
• Dependency: pyproject.toml:66 - hypothesis>=6.100.0