Breaking Project

SPARQL Basic Graph Pattern Matching Andy writes, "Example data:

  :a :p 1 .
  :a :p 2 .

Example query pattern:

 { ?x :p [] }

How many answers? Blank nodes are existential variables in RDF; named variables (the regular ?x ones) are universal variables. Queries don't return the binding of an existential; queries can return the binding of a named variable and the bound value of a named variables can be passed to other parts of the query (FILTERs, OPTIONALs etc) via the algebra.

In the absence of a DISTINCT in the query, are the solutions to this pattern:

• 1 : ?x = :a
• 2 : ?x = :a , ?x = :a
• Either 1 or 2

"
I would say there is a fourth option here, see this example, which would be:

• { { ?x:subject = :a, p1:predicate1 = :p, o1:object1 = 1 } , { ?x:subject = :a, p1:predicate1 = :p, o1:object = 2 } }

The DAWG test case, rdfSemantics-bNode-type-var, redefines the meaning of project.

If you keep track of where the variable were bound you can still count the distinct instances without redefining project.