# Run all tests
pytest tests/

# Run with coverage
pytest tests/ --cov=goodpoints --cov-report=html

# Run specific file or pattern
pytest tests/test_kt.py -v
pytest tests/ -k "kernel"

All fixtures use deterministic seeding for reproducibility.

make_X(n, d) - Standard normal test data (mean=0, std=1)

X = make_X(n=64, d=2)  # (64, 2) array, values typically in [-3, 3]
X = make_X(n=32, d=3)  # Each (n, d) pair produces same data every time

make_K_gauss(n) - Gaussian kernel matrices from standard normal data

K = make_K_gauss(n=64)  # (64, 64) symmetric PSD matrix
# K[i,j] = exp(-‖x_i - x_j‖²), diagonal is 1.0

gaussian_kernel_fn - Gaussian RBF kernel: k(y, x) = exp(-‖y - x‖²)

# Returns similarity scores in (0, 1], where 1 = identical points
result = kt.thin_X(X, m=1, split_kernel=gaussian_kernel_fn,
                   swap_kernel=gaussian_kernel_fn)

For testing patterns, see:

• test_properties.py - Property-based tests with Hypothesis
• test_kernels.py - Custom kernels (Laplacian, IMQ, Sobolev)
• test_integration.py - Cross-module workflows

Before submitting:

1. Run: pytest tests/
2. Check coverage: pytest tests/ --cov=goodpoints
3. Use factory fixtures (make_X, make_K_gauss)
4. Add docstrings to test functions