Divergence between object property's range and restrictions using it

[mereolog]

In general, if:

1. ?property rdfs:range ?range.
2. ?restriction owl:onProperty ?property.

then the restricting class (owl:onClass) in ?restriction should be a subclass of ?range.

Currently, IDMP Ontology has one such example: ManufacturingAuthorization is restricted by property hasAuthorizedParty and class Manufacturer, which is not a subclass of hasAuthorizedParty's range: AuthorizedParty. Then if ABC is an instance of ManufacturingAuthorization, then the restriction implies that there is XYZ that belongs both to Manufacturer and to hasAuthorizedParty 's range: AuthorizedParty. Logically, this is possible, but it may also indicate that Manufacturer should be a subclass of AuthorizedParty.

The following query should find all such cases:

SELECT DISTINCT ?class ?property ?restrictingClass ?range
WHERE
{
    ?property rdfs:range ?range.
    ?restriction rdf:type owl:Restriction.
    ?restriction owl:onProperty ?property.
    ?class rdfs:subClassOf ?restriction.
    {
        {
            ?restriction owl:someValuesFrom ?restrictingClass.
        }
        UNION
        {
            {
                {?restriction owl:minQualifiedCardinality ?cardinality.}
                UNION
                {?restriction owl:minCardinality ?cardinality.}
            }
            FILTER (?cardinality > 0)
        }
    }
    FILTER (ISIRI(?range) && ISIRI(?restrictingClass))
    
    FILTER NOT EXISTS
    {
        ?restrictingClass rdfs:subClassOf* ?range.
    }
}

This should probably be added to hygiene tests.

Note that in principle we could drop FILTER (?cardinality > 0) if we want to check also the min 0 restrictions.

[ElisaKendall]

@merelog I'm not sure that I want to make all manufacturers a subclass of authorized party. They are not authorized parties until they get authorization from a regulatory agency for something. So I don't know that I want this to be anything other than a warning, but there could be others over time. One option might be to create a subclass of manufacturer that is both an authorized party and manufacturer, but again they are only authorized in the context of having manufactured a specific drug that is authorized in a given jurisdiction. I think this part of the model will evolve in the next phase, so perhaps "stay tuned"?