fix(validation): count a scoped @context term only where it is a property

[simontaurus]

Fixes OO-LD/oold-schema#149. Repro from the issue now passes: 30 ok, 0 failed, 0 warning(s).

Cause

embedded_properties (src/oold/validation/frame.py) took every @context term carrying a scoped @context, without checking that the schema declares a property of that name.

OOLD-CMP-b926 requires a schema to reflect every $ref in its @context, including $refs reached from $defs. In the repro, amount lives in $defs/Component, so its term is in the root context by that rule, but the root object has no amount property and no instance of it can. Counting the term had two effects, both from the same call:

1. roundtrip() chose framed over compacted
2. schema_to_frame() emitted "amount": {}

A frame property is a match filter, so the frame selected nodes carrying amount. The generated instance {"type": ["Composition"]} carries none, framing returned nothing, and type was reported lost. type was never involved; it was just the only property present.

This is why the two branches in the issue looked contradictory: removing the scoped @context removed the phantom embed, so the round-trip fell back to compaction and passed, while OOLD-CMP-5266 returned. The rule and the check never actually disagreed.

Fix

Require a scoped term to name a property the schema declares. schema here is dereferenced, so collect_composed_properties sees properties inherited through allOf as well.

No spec change: OOLD-CMP-5266 recommends exactly what the round-trip supports, and the three warnings in the materials schemas can now be acted on.

Verification

• repro from the issue: FAIL -> PASS
• pytest: 558 passed, 9 skipped, plus a regression test for the $defs case
• oold-schema/examples: 380 ok, 0 failed, unchanged warnings
• parity against validate.mjs: 6 passed

scripts/schema_to_frame.mjs in oold-schema has the same defect; it is not exercised by the parity fixtures. Fixed there in OO-LD/oold-schema#151 so the reference implementation carries it into oold-js.

[github-actions]

Release preview

Merging this PR would release v0.18.1 (current: v0.18.0).

Changelog preview (truncated)

## v0.18.1 (2026-08-27)

### Bug Fixes

- **validation**: Count a scoped @context term only where it is a property
  ([`4f4c39c`](https://github.com/OO-LD/oold-python/commit/4f4c39c0723a423d49df54ff742abb389d239f43))

### Code Style

- Apply ruff format to the new test
  ([`c23383d`](https://github.com/OO-LD/oold-python/commit/c23383dad981c03c46960a362094090b333831c2))

### Continuous Integration

- Gate the parity check against the reference
  ([`fdb9007`](https://github.com/OO-LD/oold-python/commit/fdb9007377aab115da3b2661e59834090af79314))

### Documentation

- **validation**: State invariants in comments, not prior behaviour
  ([`4baf212`](https://github.com/OO-LD/oold-python/commit/4baf212f85a6f97fc9f6f16eee07ab2a0bcd42d5))

_{Preview via python-semantic-release and conventional commits.}

[github-actions]

📊 Benchmark Results

Click to see benchmark comparison

📊 Benchmark Comparison (threshold: 1.3x)
============================================================

➖ Unchanged (within threshold):
  ➖ test_simple_dict_document_store: 0.0017s → 0.0017s (+0.5%)
  ➖ test_sqlite_document_store: 0.0019s → 0.0019s (+0.3%)
  ➖ test_local_sparql_store: 0.0380s → 0.0389s (+2.3%)
  ➖ test_oneof_subschema: 0.0589s → 0.0587s (-0.4%)
  ➖ test_enum_docstrings: 0.0494s → 0.0496s (+0.4%)
  ➖ test_subclass_inheritance: 0.0528s → 0.0530s (+0.4%)
  ➖ test_class_hierarchy: 0.0502s → 0.0507s (+0.9%)
  ➖ test_core[v1]: 0.0376s → 0.0384s (+2.1%)
  ➖ test_core[v2]: 0.0433s → 0.0452s (+4.4%)
  ➖ test_schema_generation[v1]: 0.0017s → 0.0017s (+0.1%)
  ➖ test_schema_generation[v2]: 0.0028s → 0.0028s (+0.7%)
  ➖ test_simple_json: 0.0007s → 0.0007s (+0.7%)
  ➖ test_complex_graph: 0.0016s → 0.0016s (+0.7%)

============================================================
Summary: 0 regressions, 0 improvements, 13 unchanged
============================================================

✅ No significant performance regressions

Threshold: 1.3x (30% slower triggers a regression warning)

Note: Benchmarks are informational only and won't fail the build.

_{💡 Tip: Download the benchmark-results artifact for detailed JSON data}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[github-actions]

📊 Benchmark Results

Click to see benchmark comparison

📊 Benchmark Comparison (threshold: 1.3x)
============================================================

➖ Unchanged (within threshold):
  ➖ test_simple_dict_document_store: 0.0019s → 0.0019s (-0.7%)
  ➖ test_sqlite_document_store: 0.0020s → 0.0019s (-6.3%)
  ➖ test_local_sparql_store: 0.0450s → 0.0429s (-4.7%)
  ➖ test_oneof_subschema: 0.0671s → 0.0657s (-2.1%)
  ➖ test_enum_docstrings: 0.0561s → 0.0560s (-0.1%)
  ➖ test_subclass_inheritance: 0.0619s → 0.0607s (-2.1%)
  ➖ test_class_hierarchy: 0.0567s → 0.0587s (+3.5%)
  ➖ test_core[v1]: 0.0421s → 0.0414s (-1.6%)
  ➖ test_core[v2]: 0.0496s → 0.0488s (-1.5%)
  ➖ test_schema_generation[v1]: 0.0017s → 0.0017s (-2.2%)
  ➖ test_schema_generation[v2]: 0.0029s → 0.0029s (-2.0%)
  ➖ test_simple_json: 0.0007s → 0.0007s (+1.5%)
  ➖ test_complex_graph: 0.0016s → 0.0017s (+1.1%)

============================================================
Summary: 0 regressions, 0 improvements, 13 unchanged
============================================================

✅ No significant performance regressions

Threshold: 1.3x (30% slower triggers a regression warning)

Note: Benchmarks are informational only and won't fail the build.

_{💡 Tip: Download the benchmark-results artifact for detailed JSON data}