Testing and coverage

DELi uses pytest for automated tests and coverage.py (via pytest-cov) to measure line coverage of the deli package.

Run the full suite from a development install:

uv sync --dev --extra ml
uv run pytest

GNN tests are marked ml and require the optional ML dependencies above. To skip them:

uv run pytest -m "not ml"

Generate a coverage report

uv run pytest --cov=deli --cov-report=term-missing --cov-report=html

This runs all tests, measures which lines in src/deli were executed, and writes:

Coverage counts statements (executable lines) in src/deli. It does not include tests, docs, or example scripts.

Figures below were generated on 2026-07-07 with 199 passing tests (Python 3.13+, including ML/GNN tests with deli-chem[ml] installed).

Total coverage
Metric	Covered	Total	Percent
All `deli` modules	4,171	5,385	77.5%

Coverage by module
Module	Covered	Statements	Coverage
`deli.decode`	1,137	1,384	82.2%
`deli.dels`	859	1,008	85.2%
`deli.enumeration`	330	409	80.7%
`deli.analysis`	978	1,329	73.6%
`deli.dna`	96	99	97.0%
`deli` (package root: CLI, configure, etc.)	718	1,044	68.8%
`deli.utils`	53	73	72.6%
`deli.design`	0	39	0.0%

`deli.analysis` submodules
File	Covered	Statements	Coverage
`analysis/cube_class.py`	569	797	71.4%
`analysis/poly_o.py`	95	99	96.0%
`analysis/gnn.py`	64	73	87.7%
`analysis/analysis_parse.py`	125	166	75.3%
`analysis/analysis_report_gen.py`	62	85	72.9%
`analysis/analyzers/gnn_analyzer.py`	63	109	57.8%

Higher coverage in decode and dels reflects unit tests around library definitions, barcodes, and decoding.
Analysis coverage exercises cube enrichment, disynthon plots, PolyO, Random Forest baselines, and GNN helpers; the GNN training loop is partially covered via mocked tests.
`deli.design`` is not yet covered by automated tests.

In publications, we cite the total coverage percentage, test count, Python version, and DELi version used when the figures were generated.