Please excuse the (possible) triviality of the question. I want to confirm that my understanding of the SPARQL specification is correct.
I imagine joining in SPARQL based on blank RDF nodes works just fine, i.e., when the join variable(s) of two graph patterns is mapped to a blank node in the solution, the join still works as expected.
Assume the dataset:
<http://www.example.com/s1> <http://www.example.com/p> _:a .
_:a <http://www.example.com/p> <http://www.example.com/s2> .
And the query:
SELECT * WHERE{
{SELECT ?x1 ?z1 WHERE{
?x1 ?y1 ?z1
}}
{SELECT ?z1 ?z2 WHERE{
?z1 ?y2 ?z2
}}
}
The query should return the following result (tried it with GraphDB):
?x1 | ?z1 | ?z2
<http://www.example.com/s1> | _:a | <http://www.example.com/s2>
That means that the solution form the first subquery:
?x1 | ?z1
<http://www.example.com/s1> | _:a
And the solution form the second subquery:
?z1 | ?z2
_:a | <http://www.example.com/s2>
Are joined correctly based on the mapping of the variable ?z1 (_:a).
