gen-shacl rules: translate equals_string on IRI- and typed-literal-valued ranges (follow-up to linkml/linkml#3989) #37

What

Since linkml#3989, and in linkml#3990, gen-shacl translates equals_string / equals_string_in in rules only on slots that hold strings or xsd:boolean. Strings means an enum, a type with datatype xsd:string, or no range. On any other range the rule is skipped with a warning (ShaclGenerator._value_terms), while gen-json-schema enforces it with const.

Reproduction (linkml#3990 head c89faac)

# schema.yaml
id: https://example.org/r
name: r
prefixes:
  ex: https://example.org/r/
  linkml: https://w3id.org/linkml/
default_prefix: ex
imports: [linkml:types]
classes:
  Thing:
    attributes:
      guard: {range: string}
      target: {range: uriorcurie}
    rules:
      - preconditions: {slot_conditions: {guard: {value_presence: PRESENT}}}
        postconditions: {slot_conditions: {target: {equals_string: "ex:a"}}}

# data.yaml
guard: g
target: ex:b

# both.yaml
plugins:
  JsonschemaValidationPlugin: {closed: true}
  linkml.validator.plugins.shacl_validation_plugin.ShaclValidationPlugin: {}

linkml-validate --config both.yaml -s schema.yaml -C Thing data.yaml

[ERROR] [data.yaml/0] 'ex:a' was expected in /target
Rule 1 of class 'Thing': skipped, because equals_string(_in) on slot 'target', whose range 'uriorcurie' is neither an enum nor a type with datatype xsd:string or xsd:boolean.

Other ranges behave the same way. The RDF term column shows rdflib_dumper / JSON-LD context / context with --xsd-anyuri-as-iri.

Contract (decide first)

• Metamodel equals_string: "the slot must have range string and the value of the slot must equal the specified value".
• Code: SchemaLoader._check_equals_string and gen-shacl's per-slot check both raise for any slot-level range other than literally string. That includes enums, curie, uriorcurie and typeof: string.
• Docs and tests use it on other ranges:
  • docs/howtos/model-property-graphs.md:421 (uriorcurie);
  • docs/faq/modeling.md:116,125 (enum, rule preconditions);
  • tests/linkml/test_compliance/test_boolean_slot_compliance.py:1054 (class range).
• array.yaml: equals_string used on enum-ranged slot in array.yaml linkml/linkml#3453 (enum range in array.yaml) is still open, although linkml-model replaced that use with ifabsent (Fix equals_string on enum-ranged array_linearization_order linkml/linkml-model#251, in 1.11.0-rc2).
• Other types: numeric ranges belong to equals_number. For booleans, see question 4 on feat(gen-shacl): translate presence-implies-value rules to SHACL-SPARQL linkml/linkml#3989.

Proposed approach (if in scope)

• Render each value as the term its range produces. Derive it from the range's type uri, not from a list of type names:
  • an expanded IRI for IRI-rendered ranges and class references;
  • "lex"^^datatype for typed literals, rejecting ill-typed values at generation time.
• Compare with _sparql_is_one_of (=), and decide which comparison the rule means.
  • SPARQL 1.1 §17.3 compares by value only numerics, strings, xsd:boolean and xsd:dateTime.
  • For other datatypes = is RDFterm-equal, which "produces a type error if the arguments are both literal but are not the same RDF term" (§17.4.1.7).
  • rdflib also compares xsd:date by value. JSON Schema compares the lexical string.
• Settle the xsd:anyURI rendering first. The dumper emits an IRI. The context emits a typed literal unless --xsd-anyuri-as-iri is set (feat(generators): add --xsd-anyuri-as-iri flag for cross-generator IRI consistency linkml/linkml#3292). Related: fix(rdf): give values of URI-like types one RDF term kind in every generator #32, Make it possible to map a URI to an RDF Node linkml/linkml#2495.
• One helper covers every translation: presence-implies-value, exclusive-value and the composed translation (feat(gen-shacl): add a compositional fallback for rule-to-SPARQL conversion linkml/linkml#3990) all share _value_terms.

Acceptance criteria

• For each newly supported range, SHACL and JSON Schema agree on the same instances. Extend _json_schema_and_shacl_verdicts (test_shaclgen.py) to load data through rdflib_dumper as well as through the JSON-LD context.
• Ill-typed values fail at generation time, and a non-canonical lexical form in the data is tested.
• Numeric ranges stay skipped unless the contract changes.