Re: CONSTRUCT operator

Pat Hayes wrote:
>> Moreover, I think that we should adopt a different notation for the
>> variable substitution in graph templates, since it looks the same as
>> the one for BGP but it behaves differently. In particular wrt
>> variables not present in the domain of a pattern solution.
> 
> 
> There once was a tweak in the definition of substitution mapping which I
> think handles this, where S(v)=v if v is outside the domain of S, so S
> maps V to (RDF_T union V). This makes all S's total on templates and
> maps templates to templates (graphs being the special case of templates
> with no variables). Would this handle the issue that you are referring
> to? (If not, I'd like to see an example, because in that case Im not
> following you.)

It wont take care of the problem, since by the definition in 10.3.2:


"""
Definition: Graph Template

A graph template is a set of triple patterns.

If T = { t_j | j = 1,2 ... m } is a graph template and S is a solution
then S(t_j) is a set of one RDF triple if all variables in t_j are in
the domain of S. S(t_j) is the empty set otherwise.

Write S(T) for the union of S(t_j).
"""

triples with variables without assignment "disappear"; while mapping the
variables into themselves would generate an RDF graph with variables.

--sergio

Received on Thursday, 26 January 2006 13:42:17 UTC