SPARQL queries equivalence issue?

Question

(especially @JeenBroekstra)

Using the following graphs, I've got this SPARQL query that currently works on a Sesame repository:

PREFIX rdfs:<http://www.w3.org/2000/01/rdf-schema#>

SELECT DISTINCT *
FROM <http://spinrdf.org/sp>
FROM <http://spinrdf.org/spl>
FROM <http://spinrdf.org/spin>
FROM <http://topbraid.org/examples/kennedysSPIN>
FROM <http://topbraid.org/examples/kennedys>

FROM NAMED <http://topbraid.org/examples/kennedys>
WHERE {
    GRAPH <http://topbraid.org/examples/kennedys> {
        ?s ?p ?o .
    } .
    OPTIONAL {
        FILTER (sameTerm(rdf:type, ?p)) .
        ?o rdfs:subClassOf+ ?supC .
    } .
    OPTIONAL {
        FILTER (sameTerm(rdf:type, ?p) && NOT EXISTS { ?s a ?newSupC . }) .
        ?o rdfs:subClassOf+ ?newSupC .
    } .
    FILTER (bound(?newSupC)) .
}

Since both last optional patterns share some filter I figured I could merge them and I resulted to this query :

PREFIX rdfs:<http://www.w3.org/2000/01/rdf-schema#>

SELECT DISTINCT *
FROM <http://spinrdf.org/sp>
FROM <http://spinrdf.org/spl>
FROM <http://spinrdf.org/spin>
FROM <http://topbraid.org/examples/kennedysSPIN>
FROM <http://topbraid.org/examples/kennedys>

FROM NAMED <http://topbraid.org/examples/kennedys>
WHERE {
    GRAPH <http://topbraid.org/examples/kennedys> {
        ?s ?p ?o .
    } .
    OPTIONAL {
        FILTER (sameTerm(rdf:type, ?p)) .
        ?o rdfs:subClassOf+ ?supC .
        OPTIONAL {
            FILTER ( NOT EXISTS { ?s a ?supC . } ) .
            BIND (?supC as ?newSupC)
        } .
    } .
    FILTER (bound(?newSupC)) .
}

but when executing it, I get an empty resultset.

Thought to ask here before filing a bug request : could you tell me where the query is wrong or is this a bug in Sesame ?

Thanks in advance, Max.

EDIT : I of course put the FILTER (bound(?newSupC)) just to expose the failing pattern. On the global scape, I still need all results even with variable ?newSupC unbound.

EDIT : Here are more details about configuration I use: Except for named graph http://topbraid.org/examples/kennedys, I updated the graphs with RDFS entailment. The Sesame repository is a simple Native store without inference support.

Answer 1

After thought I was sticking too much of trying to merge the two OPTIONAL pattern where the solution was much more direct :

PREFIX rdfs:<http://www.w3.org/2000/01/rdf-schema#>

SELECT DISTINCT *
FROM <http://spinrdf.org/sp>
FROM <http://spinrdf.org/spl>
FROM <http://spinrdf.org/spin>
FROM <http://topbraid.org/examples/kennedysSPIN>
FROM <http://topbraid.org/examples/kennedys>

FROM NAMED <http://topbraid.org/examples/kennedys>
WHERE {
    GRAPH <http://topbraid.org/examples/kennedys> {
        ?s ?p ?o .
    } .
    OPTIONAL {
        FILTER (sameTerm(rdf:type, ?p)) .
        ?o rdfs:subClassOf  ?supC .
    } .
    BIND (IF(NOT EXISTS { ?s a ?supC . }, ?supC, ?__unbound__) as ?newSupC) . 
}

Anyway thanks for you time and advice, guys. Max.