Michael Haas wrote:
> Hello,
>
> as pointed out earlier, I'm having some issues with the new SPARQL
> endpoint. I'm currently using DBpedia to generate dictionaries for a
> task in an information extraction class I'm taking.
>
> For this task, I need a list of entities, e.g. actors. Consider the
> following query:
>
> SELECT ?name WHERE { ?a
> <http://www.w3.org/1999/02/22-rdf-syntax-ns#type>
> <http://dbpedia.org/ontology/Actor> . ?a
> <http://www.w3.org/2000/01/rdf-schema#label> ?name }
>
> With DBpedia 3.2, this would work just fine. With the current release,
> the query will time out after a while, giving me a partial result list.
> This actually is a feature called "Anytime" queries. [0]
>
>
> I wonder if enabling the Anytime feature is a good idea - not because I
> can't get my list of actors, but because it's broken, undocumented and
> proprietary:
>
> Kingsley Idehen wrote:
>
>> There is a bit of a doc mess up right now. New docs are in progress etc..
>>
>
>
>> Re. SPARQL protocol, this should come through HTTP response, but weare
>> still working on this part.
>>
>
> That's not exactly in the "SPARQL Protocol for RDF" recommendation.
>
>
> There is no way right now to let a SPARQL-compliant client know there
> are more results. AFAIK, it is also impossible to set these timeouts
> using the SPARQL Protocol. I don't think proprietary protocol extensions
> are the right thing for an Open project.
>
>
> Additionally, handing out different result sets for the same query
> depending on what kind of data is cached and how far subordinate clauses
> from *previous* queries have been evaluated (see [0]) sounds broken. In
> fact, I don't believe the SPARQL W3C recommendation allows that (section
> 12.5, "Evaluation Semantics").
>
Micheal,
The only issue here is that it isn't documented properly.
And the real question is simply this: how do you answer queries at Web
Scale?
Do you seriously expect a SPARQL endpoint to deliver complete results
for every conceivable query (good, bad, complex) in predictable time,
really?
If you prefer the old behavior, no problem, we switch off this feature,
but don't jump to conclusions re. its purpose.
btw - I could simply have opted for the basic single server edition of
Virtuoso, and in retrospect, maybe that's what should be out there.
Do remember, we are offering a service at our cost to the community,
please do remember this.
Kingsley
>
>
> I do acknowledge that handling web-scale data sets presents a problem,
> but I'd rather see a query language which can do proper chunking of
> results instead of breaking SPARQL.
>
>
> Anyways - I tried to work around this issue by using the LIMIT and
> OFFSET solution sequence modifiers. The W3C recommendation states:
> "Using LIMIT and OFFSET to select different subsets of the query
> solutions will not be useful unless the order is made predictable by
> using ORDER BY." - so throw in an ORDER BY as well. This will break
> after some iterations:
>
> 22023 Error SR353: Sorted TOP clause specifies more then 10100 rows to
> sort. Only 10000 are allowed. Either decrease the offset and/or row
> count or use a scrollable cursor
>
> SPARQL query:
> define sql:signal-void-variables 1 define input:default-graph-uri
> <http://dbpedia.org> SELECT ?name WHERE { ?a
> <http://www.w3.org/1999/02/22-rdf-syntax-ns#type>
> <http://dbpedia.org/ontology/Actor> . ?a
> <http://www.w3.org/2000/01/rdf-schema#label> ?name } ORDER BY ?name
> LIMIT 100 OFFSET 10000
>
>
> Any ideas?
>
>
> Regards,
>
> Michael
>
>
>
>
>
> [0] http://www.openlinksw.com/weblog/oerling/?id=1494
>
>
> ------------------------------------------------------------------------------
> _______________________________________________
> Dbpedia-discussion mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/dbpedia-discussion
>
>
--
Regards,
Kingsley Idehen Weblog: http://www.openlinksw.com/blog/~kidehen
President & CEO
OpenLink Software Web: http://www.openlinksw.com
------------------------------------------------------------------------------
_______________________________________________
Dbpedia-discussion mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/dbpedia-discussion
