Plan, apply and roll back changes to a production ontology, with a blast radius report and a proof an auditor can re-check.
Rust ai-nativedescription-logicsformal-verificationinfrastructure-as-codeisabelle-languageknowledge-graphleanlean4mcpmcp-serverontologyowlowl2reasoningrustshaclsparqltheorem-proving
"""Closed-world vocabulary checking.
In the open world an undeclared IRI is unknown, not wrong, so an extractor that
invents `ex:hasProteinName` produces RDF that parses, loads and passes SHACL
without complaint. Closed-world checking is the only thing that separates a real
term from a plausible-looking one, and it is the check that has no equivalent
elsewhere in the Python RDF stack.
Mirrors the semantics of the Rust engine's `vocab_check`: police the namespaces
the ontology itself declares terms in, never instance IRIs, never the standard
vocabularies, and refuse to return a pass when there is no vocabulary to check
against.
"""
import pytest
pytest.importorskip("pyoxigraph")
from open_ontologies_lite.vocab_check import vocab_check # noqa: E402
ONTOLOGY = """
@prefix owl: <http://www.w3.org/2002/07/owl#> .
@prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#> .
@prefix ex: <http://example.org/onto#> .
ex:Person a owl:Class .
ex:Protein a owl:Class .
ex:hasName a owl:DatatypeProperty ; rdfs:domain ex:Person .
"""
CLEAN_DATA = """
@prefix ex: <http://example.org/onto#> .
<http://example.org/instances/ada> a ex:Person ; ex:hasName "Ada" .
"""
INVENTED_TERM = """
@prefix ex: <http://example.org/onto#> .
<http://example.org/instances/ada> a ex:Person ; ex:hasProteinName "Ada" .
"""
def test_clean_data_conforms():
report = vocab_check(ONTOLOGY, CLEAN_DATA)
assert report["conforms"] is True
assert report["undeclared_terms"] == []
assert report["ontology_terms"] == 3
def test_an_invented_predicate_is_flagged():
report = vocab_check(ONTOLOGY, INVENTED_TERM)
assert report["conforms"] is False
assert report["undeclared_terms"] == ["http://example.org/onto#hasProteinName"]
def test_instance_iris_are_not_policed():
"""Individuals live in the data, not the vocabulary; flagging them is noise."""
data = """
@prefix ex: <http://example.org/onto#> .
<http://example.org/onto#ada> a ex:Person ; ex:hasName "Ada" .
"""
report = vocab_check(ONTOLOGY, data)
assert report["conforms"] is True, report["undeclared_terms"]
def test_terms_outside_the_policed_namespaces_are_left_alone():
data = """
@prefix ex: <http://example.org/onto#> .
@prefix dct: <http://purl.org/dc/terms/> .
<http://example.org/instances/ada> a ex:Person ; dct:title "Ada" .
"""
report = vocab_check(ONTOLOGY, data)
assert report["conforms"] is True
assert "http://purl.org/dc/terms/" not in report["checked_namespaces"]
def test_standard_vocabularies_are_never_policed():
data = """
@prefix ex: <http://example.org/onto#> .
@prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#> .
<http://example.org/instances/ada> a ex:Person ; rdfs:label "Ada" .
"""
report = vocab_check(ONTOLOGY, data)
assert report["conforms"] is True
def test_no_vocabulary_never_returns_a_pass():
"""The footgun this check exists to kill: a green light from an empty ontology."""
report = vocab_check("", CLEAN_DATA)
assert report["conforms"] is False
assert "nothing was checked" in report["warning"]
assert report["ontology_terms"] == 0
def test_extra_namespaces_can_be_policed_explicitly():
report = vocab_check(
"", INVENTED_TERM, extra_namespaces=["http://example.org/onto#"]
)
assert report["conforms"] is False
assert "http://example.org/onto#hasProteinName" in report["undeclared_terms"]
def test_a_term_carrying_only_a_domain_axiom_counts_as_declared():
ontology = """
@prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#> .
@prefix ex: <http://example.org/onto#> .
ex:hasName rdfs:domain ex:Person .
"""
data = """
@prefix ex: <http://example.org/onto#> .
<http://example.org/instances/ada> ex:hasName "Ada" .
"""
report = vocab_check(ontology, data)
assert report["conforms"] is True