Property path patterns in SPARQL*

[hartig]

I have added the definition of a SPARQL* property path pattern into the draft just for the sake of having such a definition. We need to think about whether it is useful to add this to SPARQL*, in which case we need to define the semantics of such SPARQL* property path patterns.

In fact, no matter what we decide, even for standard property path patterns, the semantics may have to be extended to use them over RDF* graphs.

[hartig]

I have looked through this in detail now and I would say that we actually do not need to do anything here---at least, if we make two assumptions (which we may make explicit in the spec):

1. We assume SA mode.
2. We assume that every mention of the words "RDF term" and "term" in Sec.18.4 of the SPARQL 1.1 spec is meant to refer to our notion of an RDF* term.

The first assumption is necessary for the following reason.

In SA mode, every embedded triple t that is not also contained directly in the given RDF* graph G (i.e., t ∉ G) is not asserted and, thus, should be ignored during path pattern matching. Ignoring such an embedded triple t happens trivially with the current definitions of property path pattern evaluation in Sec.18.4 of the SPARQL 1.1 spec.

In contrast, an embedded triple t' that is also contained directly in the given RDF* graph G (i.e., t' ∈ G) is asserted and, thus, must not be ignored during path pattern matching. However, due to the fact that t' ∈ G, the triple is considered by the current definitions of property path pattern evaluation, as desired.

The second assumption is necessary to be able to consider nested RDF* triples (i.e., triples that have an embedded triple as their subject or their object) during path pattern matching. Then, a path that matches a property path expression may start from an embedded triple or may end in an embedded triple (or both). It is even possible that intermediate nodes along such a path are embedded triples.

As an example, consider the RDF* graph in the following snippet of Turtle*.

<<:s1 :p1 :o1>>  :q1  <<:s2 :p2 :o2>> .
<<:s2 :p2 :o2>>  :q2  :a .

...and consider the following SPARQL* query.

SELECT ?x ?y WHERE {
   ?x  :q1/:q2*  ?y 
}

The result of this query over the RDF* graph consists of two solution mappings:

• μ1 = { ?x -> t1, ?y -> t2 } and
• μ2 = { ?x -> t1, ?y -> :a },

where t1 is the triple (:s1,:p1,:o1) and t2 is the triple (:s2,:p2,:o2).

[afs]

See also #65.

[pchampin]

I think this issue can now be closed. @hartig @afs do you concur?

[afs]

Concur

[hartig]

Concur