Re: scoping set for the graph selector variable of a GRAPH pattern

Fred Zemke wrote:
> In Perez et.al. "Semantics of SPARQL" Definition 3.9
> item (3) second dashed bullet, I think it is saying that the
> value of a graph selector variable in a GRAPH pattern
> does not need to be in the scoping set; rather, it needs to
> be among the names of named graphs.  I agree with this position.

I agree as well.

Generally, IRIs can be placed into the results quite safely - it's blank nodes 
that need careful handling in the presence of entailment.

	Andy

> 
> For example
> 
> SELECT ?g
> FROM NAMED <g1>
> FROM NAMED <g2>
> SELECT { GRAPH ?g { 'Tom' <loves> 'Mary' } }
> 
> the query asks which graphs know that Tom loves Mary.
> I do not believe there is any requirement for the value of
> ?g to be in the scoping set, that is, the RDF terms of the
> default graph, which in this particular query will default to
> some system default graph.  I see no reason to demand that
> ?g must be mapped to some term in the default RDF graph.
> I see no reason to expect that the default RDF graph is
> even aware of the two named graphs <g1> and <g2>.
> 
> Of course if ?g was also constrained by appearing in some basic
> graph pattern, then the constraints on pattern matching the BGP
> would also come into play and ?g would need to match a term
> of the graph in which the BGP is evaluated.
> 
> Fred
>

Received on Tuesday, 14 November 2006 18:23:24 UTC