Testing

Testing#

Testing guidelines for brainmass.

Running Tests#

Run all tests:

pytest tests/

Run specific test file:

pytest tests/test_models.py

Run with coverage:

pytest --cov=brainmass tests/

Test Structure#

tests/
├── test_models.py          # Neural mass model tests
├── test_noise.py           # Noise process tests
├── test_coupling.py        # Coupling mechanism tests
├── test_forward.py         # Forward model tests
└── test_utils.py           # Utility function tests

Writing Tests#

Basic Test#

import pytest
import jax.numpy as jnp
import brainmass

def test_hopf_oscillator():
    # Create model
    model = brainmass.HopfOscillator(in_size=10, omega=10)

    # Initialize
    model.init_all_states()

    # Test update
    output = model.update()

    # Assertions
    assert output.shape == (10,)
    assert not jnp.any(jnp.isnan(output))

Parametrized Tests#

@pytest.mark.parametrize("in_size", [1, 10, 100])
def test_model_shapes(in_size):
    model = brainmass.WilsonCowanModel(in_size=in_size)
    model.init_all_states()

    output = model.update(rE_inp=0.5, rI_inp=0.2)
    assert output.shape == (in_size,)

Batch Testing#

@pytest.mark.parametrize("batch_size", [None, 1, 32])
def test_batching(batch_size):
    model = brainmass.HopfOscillator(in_size=5, omega=10)
    model.init_all_states(batch_size=batch_size)

    output = model.update()

    if batch_size is None:
        assert output.shape == (5,)
    else:
        assert output.shape == (batch_size, 5)

Test Best Practices#

  1. Test one thing per test

  2. Use descriptive test names

  3. Check shapes, values, and edge cases

  4. Test with and without batch dimension

  5. Test with units where applicable

Coverage#

Aim for >80% code coverage:

pytest --cov=brainmass --cov-report=html tests/
# Open htmlcov/index.html

Continuous Integration#

Tests run automatically on:

  • Every pull request

  • Pushes to main branch

  • Nightly builds

Ensure tests pass before submitting PR.

See Also#