Known Limitations

CIMVocabCheck is a static validator: it reasons about a query or shapes graph against a schema, without data and without execution. That makes some checks impossible or intentionally conservative. This page is the honest list of what it does not catch, so you know where to apply judgement.

The guiding principle is silent-on-missing: when the schema doesn't carry the information a check needs, the check is skipped rather than guessed.

Dynamic predicates and classes

When a triple pattern uses a variable as the predicate or rdf:type object (e.g. ?s ?p ?o or ?s a ?cls), CIMVocabCheck cannot perform per-triple existence or domain/range checks. It emits a single UNSUPPORTED_DYNAMIC_PROPERTY (WARN) for the query to flag that static validation is incomplete — it does not try to enumerate what the variable might be. A variable in subject position is fine; only variable predicates and type-object positions are affected.

Exception — SHACL $PATH: in $this $PATH ?value patterns, $PATH is resolved to the enclosing property shape's sh:path before analysis, so no warning is emitted for those.

SERVICE blocks

SERVICE { } federated-query blocks are silently skipped — CIMVocabCheck has no access to the remote endpoint's schema, so no class/property checks run inside the SERVICE body. No diagnostic is emitted.

Incomplete schema — missing rdfs:domain / rdfs:range

All semantic checks follow the silent-on-missing policy:

• No rdfs:domain for a property → no PROPERTY_NOT_ALLOWED_FOR_CLASS or QUERY_IMPLIED_TYPE.
• No rdfs:range → no DATATYPE_MISMATCH.

So a query using a property against a completely wrong subject type produces no error if the profile omits rdfs:domain for it. The existence checks (UNKNOWN_CLASS / UNKNOWN_PROPERTY) still fire regardless.

Practical consequence for CIM users

The CGMES RDFS files published by ENTSO-E declare rdfs:domain/rdfs:range on most properties, so semantic checks are generally active. Proprietary or partial profile files may suppress them silently. rdfs:subClassOf traversal is transitive but limited to what the loaded profiles declare.

Path-chain checks

Forward property-path chains (ex:p1/ex:p2/ex:p3) are checked for range–domain compatibility between adjacent segments. These operators are not checked in chain context (each URI segment is still checked for existence):

• Inverse paths (^ex:p)
• Alternative paths (ex:p1 | ex:p2)
• Zero-or-more / one-or-more / zero-or-one (*, +, ?)

Datatype precision

Datatype range checks group XSD types into coarse compatibility buckets:

• All numeric XSD types (xsd:integer, xsd:float, xsd:decimal, …) are one bucket — using the wrong numeric subtype is not flagged.
• rdf:langString is treated as compatible with xsd:string.

Precision mismatches within a bucket do not produce a DATATYPE_MISMATCH.

Source positions

Line/column are located by searching the original query text for the IRI in three forms (<full-IRI>, prefix:local, and the a keyword), picking the occurrence nearest the offending triple's subject. Positions may be wrong or missing when:

• the IRI appears only inside a string literal or comment (a simple bracket/string tracker can be confused by complex multi-line strings);
• the IRI is a constant in a BIND / VALUES / FILTER expression — such constants are validated (existence, enumeration, vocabulary typos), but their reported source position may be approximate or absent;
• the finding is a SHACL shape-level annotation — the shapes graph is RDF, not text, so the position is resolved by searching the Turtle source for the offending term. Findings from embedded SPARQL are mapped back to their line/column in the Turtle source by all consumers (CLI, editors, and Code Quality reports alike).

SHACL: structural checks vs. SPARQL-validated

CIMVocabCheck makes two passes over a SHACL shapes graph.

Pass 1 — structural analysis validates shape declarations against the schema:

Predicate	What is checked
sh:targetClass / sh:class	CIM class IRI must exist in the selected profiles. Standard-vocab classes, classes the file declares itself, and constraint-component parameters are exempt; a closed-namespace typo (e.g. sh:Fooo) is reported as UNKNOWN_VOCABULARY_TERM
sh:path	Every URI segment must be a known property (standard-vocab terms, header extensions, properties the file declares itself, and constraint-component parameter paths are exempt); a closed-namespace typo (e.g. rdf:typ) is reported as UNKNOWN_VOCABULARY_TERM
sh:nodeKind + rdfs:range	NODE_KIND_INCOMPATIBLE_WITH_RANGE
sh:datatype / sh:class vs rdfs:range	DATATYPE_INCOMPATIBLE_WITH_RANGE / CLASS_INCOMPATIBLE_WITH_RANGE; a non-XSD-datatype sh:datatype is UNKNOWN_VOCABULARY_TERM
sh:minCount + sh:maxCount	INVALID_CARDINALITY when min > max; CARDINALITY_INCOMPATIBLE_WITH_MULTIPLICITY when incompatible with the property's cims:multiplicity
sh:in / sh:hasValue	Enumeration-member and existence checks on the listed/required values
sh:minInclusive / sh:maxInclusive / sh:minExclusive / sh:maxExclusive	INVALID_VALUE_RANGE on a contradictory bound
sh:targetSubjectsOf / sh:targetObjectsOf, sh:equals / sh:disjoint / sh:lessThan(OrEquals), sh:ignoredProperties	Referenced property IRIs must exist

Shapes carrying sh:deactivated true are skipped entirely.

Pass 2 — embedded SPARQL extracts and fully validates inline SPARQL (sh:sparql→sh:select, sh:target→sh:select, sh:validator→sh:ask, sh:rule→sh:construct), with $this typed as the shape's sh:targetClass. QUERY_IMPLIED_TYPE is suppressed inside embedded results (the intermediate bindings are transient and not expected to be type-annotated).

SHACL features not checked (they require data or boolean-shape reasoning): sh:pattern/sh:flags, sh:qualifiedValueShape/Min/MaxCount, sh:uniqueLang, sh:minLength/sh:maxLength, sh:or/sh:and/sh:not/sh:xone, and sh:targetNode. The shape-declared sh:severity is not honored either — findings use CIMVocabCheck's own ERROR/WARN mapping (adjust it globally with strictness) rather than the severity a shape declares.

owl:imports are not followed

When schema files are loaded from disk, owl:imports declarations are not resolved — every profile file that contributes terms must be passed explicitly (via schemas / schemasDirectory, --schema, or a schema-bearing endpoint). CGMES import references are namespace IRIs with no IRI→file catalog, so there is nothing to resolve them against automatically.

Roadmap

• Tighter path-chain checks (inverse / alternative operators).
• Remaining SHACL constraints (sh:pattern, qualified value shapes, logical constraints).
• Honor shape-declared sh:severity.
• SERVICE schema hints via config.
• Completion in full-IRI (<http://…) position.

See Validation checks for what is checked.