- From: Matthew Perry <matthew.perry@oracle.com>
- Date: Tue, 25 Jan 2011 11:53:44 -0500
- To: Gregory Williams <greg@evilfunhouse.com>
- CC: W3C SPARQL Working Group <public-rdf-dawg@w3.org>
It seems strange to me to get any results with an empty dataset. I don't agree that terms from the query should be included in nodes(G). From my understanding, { ?x ?p "o" } will not match if "o" does not appear in the dataset, so I don't think { ?x :p{0} "o" } should match. - Matt On 1/25/2011 11:16 AM, Gregory Williams wrote: > On Jan 25, 2011, at 9:41 AM, Matthew Perry wrote: > >> I have reviewed the property path tests. We need tests for {m,n}, {,n}, ? and () for precedence, and I think we need more tests for combinations of property path constructs. >> >> I have disagreements with some of the answers given in the current tests. >> >> 1) pp15 -- I don't see why we are returning results on an empty dataset. > pp15 returns results because zero-length property paths bind the path endpoints to any subjects or objects in the graph *and* any bound term explicitly in the query. So in pp15, ?X :p{0} "o" will bind ?X="o". Likewise for ?Y and ?Z. This is part of the evaluation semantics for ZeroLengthPath (http://www.w3.org/2009/sparql/docs/query-1.1/rq25.xml#sparqlAlgebraEval), though maybe it could be made clearer in the description of property paths in section 9 (from the evaluation semantics, I take "graph node" in section 9 to include terms not necessarily in the dataset but that are present in the query). > > .greg > > > >
Received on Tuesday, 25 January 2011 16:54:26 UTC