named individual flag not making any difference in expression-to-sparql conversion

[alkidbaci]

For the example class expression:

ce = OWLObjectAllValuesFrom(property=OWLObjectProperty(IRI('http://www.benchmark.org/family#', 'hasParent')),filler=OWLObjectComplementOf(OWLClass(IRI('http://www.benchmark.org/family#', 'Father'))))

the method owl_expression_to_sparql(ce, named_individuals=False) will output the following query:

SELECT
 DISTINCT ?x WHERE { 
?x ?s_1 ?s_2 . 
FILTER NOT EXISTS { 
?x <http://www.benchmark.org/family#hasParent> ?s_3 . 
FILTER NOT EXISTS { 
?s_3 ?s_4 ?s_5 . 
FILTER NOT EXISTS { 
?s_3 a <http://www.benchmark.org/family#Father> . 
 }
 }
 }
 }

Whereas the method owl_expression_to_sparql(ce, named_individuals=True) will output the following query:

SELECT
 DISTINCT ?x WHERE { 
?x ?s_1 ?s_2 . 
FILTER NOT EXISTS { 
?x <http://www.benchmark.org/family#hasParent> ?s_3 . 
FILTER NOT EXISTS { 
?s_3 a <http://www.w3.org/2002/07/owl#NamedIndividual> . 
FILTER NOT EXISTS { 
?s_3 a <http://www.benchmark.org/family#Father> . 
 }
 }
 }
 }

Notice that the second query wont filter out the subjects that are not named individuals for the ?x variable but for ?s_3, hence a trivial statement.
Instead we should have a rule as follows:

SELECT
 DISTINCT ?x WHERE { 
?x ?s_1 ?s_2 . 
?x a <http://www.w3.org/2002/07/owl#NamedIndividual> . 
....

[alkidbaci]

@nkaralis would appreciate your input on this

[nkaralis]

The idea was to make the evaluation of complements faster for some systems.

?s_3 a <http://www.w3.org/2002/07/owl#NamedIndividual> . this should be evaluated faster than ?s_3 ?s_4 ?s_5 .
In some cases, it might also affect the results.

It seems that I forgot to add it to the most top level and to the forAllDeMorgan case.