fix: close remaining #271 type-check gaps in swrl.py and owl_data_ranges.py

[Demirrr]

Summary

Follow-up to #272/#273/#274, which audited every OWL-construct constructor for Python-side type checks per #271 but explicitly left two areas open, plus a broader sweep of a few more constructors that share the same failure pattern.

Closing out #271's explicitly-tracked scope

• owlapy/swrl.py had no type checks at all: Variable/IVariable/DVariable, ClassAtom, DataRangeAtom, ObjectPropertyAtom, DataPropertyAtom, SameAsAtom, DifferentFromAtom, BuiltInAtom, Rule. All now validate their arguments and raise TypeError on the wrong type.

  • Adding a check to BuiltInAtom surfaced a latent bug in Atom.from_string's built-in-predicate branch: the return sat inside the arg-conversion loop, so only the first argument of a multi-arg built-in (e.g. greaterThanOrEqual(?x, 40)) was ever converted to a DVariable/OWLLiteral — every later argument leaked through as a raw, untyped string. The non-variable branch also wrapped values as OWLNamedIndividual instead of OWLLiteral, contradicting the constructor's own type annotation. Fixed both; updated the one test whose hardcoded repr() string baked in the old buggy output.

• owlapy/owl_data_ranges.py's OWLNaryDataRange (backs OWLDataIntersectionOf/OWLDataUnionOf) and OWLDataComplementOf were deliberately left unchecked in fix: close remaining #271 type-check gaps in class_expression.restriction #274, because owlapy.utils.nnf.NNF reuses these constructors to wrap data-side class expressions (e.g. OWLDataSomeValuesFrom) during negation, not just genuine OWLDataRange instances — a strict isinstance(x, OWLDataRange) check breaks test_owlapy_nnf.py. Validated against OWLPropertyRange instead: the actual common base of OWLDataRange and OWLClassExpression, already defined in owl_data_ranges.py (no circular import needed). This accepts the intentional NNF reuse pattern while still rejecting genuinely wrong types (str, int, OWLIndividual, ...).

This resolves the two things #274 flagged as needing "a dedicated design discussion" / being "a separate, larger piece of work."

Broader sweep beyond #271's tracked scope

A follow-up audit of the rest of the codebase for the same silent-Python-success-then-opaque-JVM-failure pattern turned up a few more constructors worth closing, all bridged to Java via owlapi_mapper.py:

• owlapy/iri.py — IRI.__init__/IRI.create: replaced several bare, -O-strippable asserts (which also raised confusing IndexError/TypeError on bad input instead of a clear message) with proper TypeError/ValueError checks on namespace/remainder/iri. IRI backs the identity of virtually every OWL entity, so this is the highest-value item in the sweep.
• owlapy/owl_individual.py — OWLAnonymousIndividual.__init__: node_id is now validated instead of failing with an unclear AttributeError deep inside NodeID.get_node_id.
• owlapy/owl_ontology.py — OWLOntologyID.__init__: ontology_iri/version_iri are now validated instead of only failing once mapped to Java's OWLOntologyID.

(A few private _OWLLiteralImpl* classes in owl_literal.py also use bare asserts, but were left alone — they're internal implementation details reached only via the already-validated public OWLLiteral(), and misuse there fails immediately in pure Python rather than silently succeeding until it hits the JVM, so they're a different, lower-priority category of cleanup than #271's actual motivating failure mode.)

Test plan

• ruff check on all changed files, --line-length=200
• PYTHONPATH=. pytest tests/test_swrl.py tests/test_owl_data_ranges_type_checks.py tests/test_owlapy_nnf.py tests/test_iri.py tests/test_owl_anonymous_individual.py tests/test_owl_ontology_id.py -p no:warnings -q (all pass — confirms the OWLPropertyRange check doesn't break NNF's reuse pattern)
• PYTHONPATH=. pytest --ignore=tests/test_z_do_last_ebr_retrieval.py -p no:warnings -q (1327 passed, 25 skipped; the only failures are 3 pre-existing environment issues unrelated to this change — a missing fixture file (test_ontology_justification.py, test_reasoner_timeout.py) and one already-known flaky timing test test_rdflib_reasoner_regression.py, same one test: regression coverage + changelog for #271/#272 type checks #273's own test plan noted)

Closes #271.

🤖 Generated with Claude Code

https://claude.ai/code/session_01M1RdTi68z3rkZzTohwSZ5F