Inefficient SPARQL generation

[luxbe]

When SOQL is translated into SPARQL, duplicate graph patterns can be generated:

SELECT p FROM Person p WHERE p.phone.number = :phoneNumber AND p.phone.brand = :brand

SELECT ?x WHERE {
    ?x a <http://krizik.felk.cvut.cz/ontologies/jopa/types#Person> .
    ?x <http://krizik.felk.cvut.cz/ontologies/jopa/attributes#hasPhone> ?phone .
    ?phone <http://krizik.felk.cvut.cz/ontologies/jopa/attributes#phoneNumber> ?phoneNumber .
    ?x <http://krizik.felk.cvut.cz/ontologies/jopa/attributes#hasPhone> ?phone .
    ?phone <http://krizik.felk.cvut.cz/ontologies/jopa/attributes#brand> ?brand .
}

When multiple nested attributes exits (e.g. p.phone.number and p.phone.brand), they both generate the same triple ?x <http://krizik.felk.cvut.cz/ontologies/jopa/attributes#hasPhone> ?phone .

While this should not have any performance impacts, because the SPARQL Parser should optimize the SPARQL query, it makes debugging the generated SPARQL harder and also leads to more traffic having to be send to the database.