W3C home > Mailing lists > Public > public-rdf-dawg@w3.org > April to June 2010

Re: [TF-PP] Zero-length paths (ACTION-251)

From: Andy Seaborne <andy.seaborne@talis.com>
Date: Tue, 08 Jun 2010 18:23:07 +0100
Message-ID: <4C0E7C7B.2080609@talis.com>
To: Gregory Williams <greg@evilfunhouse.com>
CC: Birte Glimm <birte.glimm@comlab.ox.ac.uk>, SPARQL Working Group WG <public-rdf-dawg@w3.org>


On 08/06/2010 4:51 PM, Gregory Williams wrote:
> On Jun 8, 2010, at 11:45 AM, Birte Glimm wrote:
>
>> Thus, a (mathematically) natural interpretation would be to return
>> only subjects and objects because they are the elements of the
>> relations (as also Andy suggested below).  It still does what Greg
>> wants I believe. E.g. lets say G contains:
>> ex:a ex:mylabel "l1".
>> ex:mylabel rdfs:subPropertyOf rdfs:label
>>
>> Greg's query is:
>> SELECT ?label WHERE {
>>        ?p rdfs:subPropertyOf* rdfs:label .
>>        ?s ?p ?label .
>> }
>
> My issue with this example was specifically so that it would handle not only the case you show (with a subproperty of rdfs:label), but also ones where there is no schema information at all. If the data only contained:
>
> <s>  rdfs:label "foo" .
>
> then I believe your approach wouldn't return any bindings for ?label, even though { ?label = "foo" } would be the intuitive result of the query, as far as I'm concerned.
>

An idea: We could define ?x :p{0} ?x as matching subjects+objects, and 
?x :p{0} :uri to match URI :uri regardless (and all the subjects+objects 
but that's not gong to change the matches).  Explicit mention adds to 
the set of possibilities.

	Andy
Received on Tuesday, 8 June 2010 17:23:39 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 16:15:42 GMT