Validation And Tests

Run semantic production-data validation before committing data updates:

astromol-validate

or from a source checkout:

python -m astromol.validation

Validation errors should block release and data-update commits. Validation warnings are allowed for known unresolved curation follow-ups, such as inherited legacy * dipole placeholders.

Production inventory counts, known validation warnings, and curation-sensitive 2026 regression expectations are also checked against the committed baseline at tests/baselines/production_data.json. Refresh that file only after a reviewed data update:

python scripts/update_data_baseline.py
python scripts/check_curation.py

Do not repair curation-driven assertion failures by editing scattered numeric literals in regression scripts. Add the generated expectation to scripts/update_data_baseline.py, regenerate tests/baselines/production_data.json, and inspect the baseline diff with the curation change.

scripts/check_curation.py expands to production-data validation plus the load, validation-baseline, and output regression tests. Use it after applying new molecule or detection records.

Run the regression suite with:

python -m pytest

The current pytest suite includes:

• native pytest checks for database loading and validation

• a wrapped regression-script harness for the migration-era table, figure, and slide checks

The wrapper is intentional for now: it preserves the audited migration checks while making them visible to pytest. Once the public API, documentation examples, and CI workflow stabilize, the highest-value checks should be converted incrementally into conventional unit and integration tests.