[stevevanhooser]

Summary

Brings the port to parity with ndi-ontology-matlab at 1fbd38b (its PR #10): OM now reads its units from OM's own RDF file, and seven new ontologies are added.

Changes

OM reads units from OM's own file

OM's units are individuals (om:metre is an om:Unit), which the EBI OLS index does not hold. So OMProvider, like OM.m since 44efcec, now reads om-2.0.rdf first and falls back to the existing OLS label search only when the file can't be read or has no match. That fallback is still how OM's classes are found, e.g. MolarVolumeUnit.

• fileIndex(reset=False) downloads and indexes the file once per process. reset=True forgets the index, and a failed download caches nothing.
• parseRDF(txt) uses MATLAB's regex to index every top-level entry, merging blocks that share an identifier. It keeps the English label, comment and alternative labels.
• findInIndex(idx, term) tries five rules in order (identifier exact, identifier ignoring case, label, alternative label, camel-case heuristic). The first rule that names exactly one entry wins.
• Purely numeric terms skip the file, as in MATLAB. An unreadable file raises a warning and falls back to OLS.
• clearCache() now also resets the OM file index, matching MATLAB's clearCache.

New providers

• OLS-backed: MICRO, MCO, OBI, ZECO, WBls. Each is a minimal OLSProvider subclass that sets only the OLS ontology id and the prefix, as the MATLAB classes do.

• Fixed lists held in the provider (no lookup service indexes them):

  • OWLTime (time:) holds OWL-Time's 15 interval relations.
  • CREDIT (credit:) holds the 14 CRediT contributor roles. Ids are slugs (credit:data-curation), and CRO's older numbering (CREDIT:00000002) is accepted as an alias.

  Both lists were copied verbatim from MATLAB and checked equal row for row. A term that isn't on the list raises.

Data, docs and the bridge contract

• ontology_list.json is an exact copy of MATLAB's at 2349314 (30 prefixes).
• The README's supported-ontologies table and shared-case counts are updated.
• ndi_matlab_python_bridge.yaml has new entries for the seven providers, OM and ontology.m are re-examined, and the MATLAB hashes are recorded. It also records a gap that predates this PR: Python's clearCache() does not reset EMPTY's cached download, whereas MATLAB's clears its OBO file cache.

Tests

• Ported TestOMFile.m (offline parsing and matching against an RDF excerpt).
• Added unit tests for OM's file lookup, its fallback and cache reset, the new OLS providers' parameters, and CREDIT and OWLTime lookup and not-found behaviour.
• MATLAB's shared case table (ontology_lookup_tests.json, read from the MATLAB checkout, not vendored here) now has 114 cases across 23 ontologies, up from 66. MATLAB's PR added the new cases; this PR supplies the providers they run against. Every case now has a ported provider behind it. CREDIT and OWLTime cases run offline alongside NDIC's.
• All CI checks pass, including the live shared-case-table job.

🤖 Generated with Claude Code

https://claude.ai/code/session_017iFBFMQmEZaz21d9ZFJbiH