fix(shaclgen): emit sh:pattern for pattern constraints inside any_of

[jdsika]

Summary

The SHACL generator (gen-shacl) drops pattern constraints declared inside any_of branches,
producing empty blank nodes [ ] instead of the intended [ sh:pattern "..." ].

Root Cause

In shaclgen.py, the for any in s.any_of: loop (line ~202) dispatches exclusively on
any.range — resolving it as a class, type, enum, or simple datatype. The pattern attribute
of each AnonymousSlotExpression is never read.

When a branch specifies pattern: without range:, any.range is None, so
add_simple_data_type(func, None) finds no matching datatype → emits an empty blank node.

The top-level s.pattern check (line ~267) only runs in the else branch when any_of is absent.

Fix

After the range dispatch loop appends a BNode to range_list, propagate the pattern:

if any.pattern:
    g.add((range_list[-1], SH.pattern, Literal(any.pattern)))

This correctly handles:

• Pattern-only branches (no range): node gets only sh:pattern
• Range + pattern branches: node gets both sh:datatype and sh:pattern
• Range-only branches (no pattern): unchanged behaviour

Motivation: SPDX LicenseRef-

The SPDX specification v2.3 (Annex D) defines custom license identifiers:

license-ref = ["DocumentRef-"(idstring)":"]"LicenseRef-"(idstring)
idstring    = 1*(ALPHA / DIGIT / "-" / ".")

Schemas modelling SPDX-compliant license fields need any_of with:

1. An enum of standard SPDX identifiers
2. A URI range (for externally-defined licenses)
3. A pattern ^LicenseRef-[a-zA-Z0-9\-\.]+$ for custom identifiers

Without this fix, branch (3) generates [ ] which is trivially satisfied by ANY value,
making the constraint weaker than intended.

Real-World Validation

This fix has been validated end-to-end with the Gaia-X Trust Framework ontology
(gx:VirtualResourceShape license constraint), where the pattern branch is used to accept
LicenseRef-Custom-Commercial-Agreement for proprietary simulation assets in the ENVITED-X
data space.

The corresponding upstream model change that depends on this fix:

• ASCS-eV/service-characteristics fix/spdx-license-ref-pattern —
  adds pattern: "^LicenseRef-[a-zA-Z0-9\\-\\.]+$" as a third any_of branch in the
  license slot definition (GitLab MR pending for upstream service-characteristics).

After regenerating GX artifacts with this fix, all 14 gx:license property shapes now
produce [ sh:pattern "^LicenseRef-[a-zA-Z0-9\\-\\.]+$" ] instead of [ ], and 45+
real-world simulation assets pass SHACL validation.

Test Coverage

New test schema (any_of_pattern.yaml) exercises all three cases with assertions on:

• Generated RDF triples (sh:pattern present/absent on correct branch nodes)
• pyshacl validation (conforming and non-conforming data)

All existing tests pass unchanged (23→24 in test_shaclgen.py).

Specification References

• SPDX v2.3 Annex D — LicenseRef ABNF
• W3C SHACL §4.7.2 — sh:pattern
• LinkML any_of — boolean slot constraints

AI Disclosure

This fix was developed with AI assistance (GitHub Copilot) for code analysis, test generation,
and commit message drafting. All changes were reviewed, validated with pyshacl against the
Gaia-X Trust Framework model, and are fully understood by the author.