Re: [Myexperiment-discuss] MyExperiment SPARQL endpoint returns incorrect results

[Peter Ansell]

Hi David,

I can filter the results using a simple rule for labels and titles, as
I never expect those to be blank nodes in practice. It will likely
only be an issue if rdf:type _:LL came back, which I think is unlikely
with MyExperiment data all having rdf:type declarations that I have
seen, so it isn't critical. If you could test this query on the new
version when you have it working it would be great.

Thanks,

Peter

On 22 March 2011 13:02, David R Newman <address@hidden> wrote:
> Hi Peter,
>
> I have been waiting on a colleague of mine to finish his reasoner for 4Store 
> [1] before upgrading.  He is currently upgrading the reasoner so it supports 
> Raptor2 (latest versions of 4Store only support Raptor2 rather than Raptor).  
> Although upgrading doesn't take that long, all being well, I like to make 
> sure any version upgrades I do work on a test server before deploying them on 
> the live server.  For that reason, version upgrades can be a bit sporadic, 
> dependent on me finding time out from the project I am employed on.  If this 
> is a critical feature you require, I will see if I can set aside some time to 
> look into doing an upgrade.  When I do, I will probably upgrade to a checkout 
> from GIT, as I am aware of some significant performance improvements added 
> since 1.1.2.
>
> Regards
>
> David Newman
>
> [1] http://4sreasoner.ecs.soton.ac.uk/
>
> -----Original Message-----
> From: address@hidden [mailto:address@hidden On Behalf Of Peter Ansell
> Sent: 22 March 2011 02:33
> To: address@hidden
> Subject: [Myexperiment-discuss] MyExperiment SPARQL endpoint returns 
> incorrect results
>
> Hi,
>
> The SPARQL endpoint at http://rdf.myexperiment.org/sparql isn'
> working properly right now for queries with optional unbound items.
>
> The query below should only return bindings for ?linkedRecordType
> ?label, ?title, or ?skoslabel when they actually exist.
>
> CONSTRUCT {
> ?s ?p <http://www.myexperiment.org/workflows/72> .
> ?s a  ?linkedRecordType .
> ?s  <http://www.w3.org/2000/01/rdf-schema#label> ?label .
> ?s  <http://purl.org/dc/elements/1.1/title> ?title .
> ?s  <http://www.w3.org/2004/02/skos/core#prefLabel> ?skoslabel .
> } WHERE {
> ?s ?p <http://www.myexperiment.org/workflows/72> .
> OPTIONAL { ?s a ?linkedRecordType . }
> OPTIONAL { ?s  <http://www.w3.org/2000/01/rdf-schema#label> ?label . }
> OPTIONAL { ?s <http://purl.org/dc/elements/1.1/title> ?title . }
> OPTIONAL { ?s <http://www.w3.org/2004/02/skos/core#prefLabel> ?skoslabel . }
> } LIMIT 500 OFFSET 0
>
> Right now the Garlik 4Store SPARQL engine is returning bindings that
> don't actually exist as blank nodes with an rdf:nodeID="LL". This is
> worse than it seems as it is inconsistent. It implies that all of the
> resources that will be bound to ?s have labels/titles/prefLabel's, and
> that they all point to the same blank node, although there is no other
> information available about the blank node they reference.
>
> You may want to try upgrading to the latest version of 4Store as you
> are one version behind right now, 1.1.2 is the latest [1].
>
> Cheers,
>
> Peter
>
> [1] http://4store.org/download/
>
> _______________________________________________
> Myexperiment-discuss mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/myexperiment-discuss
>
>
> _______________________________________________
> Myexperiment-discuss mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/myexperiment-discuss
>