Stop the missing-token check reporting a header's own Number=.

[ecrum19]

Problem

preflight_missing_token_conformance flags bare "." literals that should have been typed vcfc:Null. It already excluded vcfc:fieldNumber and vcfc:genotypeString, where the vocabulary requires a dot.

The same token reaches the graph a third way. The structured-header layer carries every declaration's attributes verbatim as vcfc:HeaderAttribute resources, so Number=. becomes a vcfc:attributeValue of "." on the attribute keyed Number. The core shapes require that value to be exactly xsd:string, so it can never be vcfc:Null.

On the v3.1.0 benchmark (vcf-bench-1, 13_query_cost, 100,000-record HG005_GRCh38 slice) the check still returned ten rows, one per Number=. INFO declaration in the header. It reported them as EXPECTED_CONFORMANCE_FAILURE on all nine validations of that input.

Change

• Both the sample query and its _count form now exclude only the Number attribute's value. A bare dot on any other attribute or field is still reported.
• The filter uses OPTIONAL + BOUND rather than NOT EXISTS. OPTIONAL and BOUND are already used by queries (q05, q07) that pass on all four validation engines; NOT EXISTS appears nowhere in the suite.
• docs/conversion.md §7 documents the three places a bare dot is conformant.

Tests

test/test_missing_token_scope_unit.py (new, 5 tests) runs both queries with rdflib over a graph produced by the real header emitter:

• Number=. declarations are not reported, and the count is 0.
• A bare dot on a record field and on a non-Number header attribute are both still reported, and the count is 2.
• A premise guard checks that the emitter really writes the "." the test is about.
• The three behavioural tests fail against the previous queries.

Full suite: 865 tests, 864 pass. The one failure, test_shacl_default_unit (the vendored vocabulary vs a sibling vocabulary checkout), also fails on an untouched checkout of main: the local vcf-rdfizer-vocabulary checkout is at d46fbdf, ahead of the vendored snapshot 5bfd19d. It is unrelated to this change.

Not yet checked on a live QLever/Comunica/HDT/COTTAS run; the next benchmark validation of the HG005 slice should show PASS for this preflight.

🤖 Generated with Claude Code

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

❌ Patch coverage is 98.30508% with 1 line in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
test/test_missing_token_scope_unit.py	98.30%	1 Missing ⚠️

📢 Thoughts on this report? Let us know!