Skip to content

Golden Tests

Jump to bottom
Chris Day edited this page Feb 8, 2026 · 10 revisions

Golden Tests – Current Behavior and Suggested Expansion

This document explains the current test coverage for uml2semantics-python related to golden outputs, and outlines a realistic path to expand it.

1. Current Golden-Style Coverage

The current repository contains a lightweight golden-style test in tests/test_golden.py that:

  • runs the CLI against the examples/ TSV bundle
  • parses the generated Turtle
  • asserts the ontology IRI exists
  • asserts the graph is non-empty
  • asserts at least one owl:Class and one owl:AnnotationProperty exist

This confirms the end-to-end pipeline works and produces a valid RDF graph, but it does not compare full outputs against a canonical expected.ttl.

2. What Is Not Implemented Yet

The following golden-test features are not present in the codebase today:

  • canonical Turtle normalisation
  • deterministic blank-node ordering
  • full file-to-file diffs between expected.ttl and actual output
  • a suite of per-feature golden cases

3. Recommended Expansion (If Desired)

If you want true golden regression testing, a practical next step is:

  1. Create a tests/golden/ directory with per-case TSV bundles.
  2. Generate expected.ttl once per case.
  3. Add a normalisation step (or accept stable rdflib output with sorted triples).
  4. Diff actual output against expected.ttl in the test.

This would allow detection of changes in:

  • choice semantics
  • datatype facets
  • enumeration individuals
  • annotations
  • prefix handling

4. Navigation

Clone this wiki locally