Testing & CI¶

Test suite¶

345 test functions across 20 files.

uv run pytest                    # full suite
uv run pytest -n auto            # parallel via pytest-xdist
uv run pytest tests/test_graph.py -v
uv run pytest -k "federated"     # by keyword

Configuration in pyproject.toml:

[tool.pytest.ini_options]
pythonpath = ["src/fed_maxfuse", "src/cli", "src/shared"]
testpaths = ["tests"]
addopts = "-p no:warnings"

Coverage by area¶

Test file	Covers
`test_pre_processor.py`	Preprocessing pipeline
`test_graph.py`	k-NN graph construction, Leiden clustering
`test_matching_utils.py`	Matching helpers
`test_correlation.py`	Pearson correlation
`test_federated_params.py`	Federated parameter parsing
`test_distributed_params.py`	Distributed parameter parsing
`test_global_refinement_loop_params.py`	Refinement-loop configuration
`test_evaluation_cluster.py`	Cluster metrics
`test_zadu_abbreviations.py`	Metric name mapping
`test_anndata_loader.py`	AnnData ingestion
`test_cli_generate_utils.py`, `test_cli_utils.py`	Configuration generation
`test_data_utils.py`, `test_deep_dict.py`, `test_verbosity.py`	Utilities

Supporting libraries: parameterized (table-driven cases), pyfakefs (filesystem isolation without touching disk), psutil (resource checks), pytest-xdist (parallelism).

Testing is a differentiator here

The upstream MaxFuse reference implementation ships no tests at all. Its two long-standing latent defects survived a rename and a repackaging precisely because nothing checked behaviour. A test suite is the cheapest protection against that, and it is worth maintaining as this codebase evolves.

Linting and static analysis¶

Four tools are configured:

uv run ruff check src tests      # lint
uv run ruff format src tests     # format
uv run pylint src                # deeper static analysis
uv run mypy src                  # type checking

pyproject.toml configures [tool.ruff] (with lint.pydocstyle, lint.pylint, lint.mccabe, lint.flake8-tidy-imports), [tool.pylint.*], [tool.mypy], [tool.black], and [tool.isort].

Continuous integration¶

No CI is currently configured

There are no workflow definitions in the repository. Tests, linting, and type checking must be run manually before committing.

A minimal workflow would be a worthwhile addition:

# .github/workflows/ci.yml
name: CI
on: [push, pull_request]
jobs:
  test:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: astral-sh/setup-uv@v3
      - run: uv sync --group dev
      - run: uv run ruff check src tests
      - run: uv run mypy src
      - run: uv run pytest -n auto

Before committing¶

uv run ruff format src tests
uv run ruff check src tests
uv run mypy src
uv run pytest -n auto

Testing numerical code¶

Two practices worth following in this codebase specifically:

Seed anything stochastic. Leiden clustering, k-NN construction, and CCA noise initialisation are all random by default. A test that does not fix seeds will be flaky. Set leiden_seed and nn_graph_seed explicitly in test configurations.

Test invariants, not just values. Exact floating-point equality is brittle across platforms. Prefer assertions on properties that must hold: matrix shapes, index ranges, monotonicity, symmetry, normalisation (unit norms, row sums), and tolerance-based comparisons via np.testing.assert_allclose.