Re: SPARQL transitive closure discussion

> of SPARQL OPTIONAL, ASK, and CONSTRUCT 
should read as SPARQL SELECT, ASK, and CONSTRUCT

----- Original Message -----
From: Satya Sahoo <sahoo.2@wright.edu>
Date: Sunday, March 21, 2010 4:19 pm
Subject: Re: SPARQL transitive closure discussion
To: James Cheney <jcheney@inf.ed.ac.uk>
Cc: "<public-xg-prov@w3.org>" <public-xg-prov@w3.org>


> Hi James,
> > Does it make sense to have a query language that can take a directed graph and calculate the   >reachability relation on that graph?  This is conceptually no problem, 
> Yes
>  
> >..and this is what transitive closure in some flavors of SQL allows you to do.
> Indeed some SW applications use workarounds for this, a post listing some examples of workaround: http://www.thefigtrees.net/lee/sw/sparql-faq#transitiv8. Among the suggested solutions, we have results comparing use of SPARQL OPTIONAL, ASK, and CONSTRUCT functions for computing transitive closure over 308 million RDF triples dataset in [1] - expectedly ASK performs the best.
>  
> >  I see no reason why RDF would have to be changed to allow this in a query language such as >SPARQL... Does it make sense to have a query language that returns graphs whose edges have 
> >(implicit) semantics "this edge is transitive", when the underlying graph model does not have such >edges.
> The two points are related and Pat's post highlights this - since the graph model does not identify a predicate as being transitive, do we indicate in a query pattern when to perform transitive closure (maybe add specialized notations */& etc.)? Does the use of specialized notations in SPARQL affect the "intended" interpretation of a RDF graph? Maybe Pat or someone from the SPARQL WG can describe all the reasons they considered before deciding against transitive closure in SPARQL. 
>  
> A technical report discussing generic SPARQL property paths: http://www.w3.org/TR/sparql11-property-paths/.  
>  
> Best,
> Satya
>  
> [1] http://knoesis.wright.edu/library/download/ProvenanceManagement_sahoo09.pdf
> 
> ----- Original Message -----
> From: James Cheney <jcheney@inf.ed.ac.uk>
> Date: Sunday, March 21, 2010 8:02 am
> Subject: Re: SPARQL transitive closure discussion
> To: Satya Sahoo <sahoo.2@wright.edu>
> Cc: "<public-xg-prov@w3.org>" <public-xg-prov@w3.org>
> 

> > Hi, 


> > I'm having trouble understanding the issue under discussion in the linked thread.
> > 

> > I thought the point raised on Friday was that SPARQL lacks the ability to say "follow an edge zero or more times", which is needed for provenance queries.
> > 

> > As I understand it (i.e., not very well), SPARQL allows queries based on graph patterns, which are close to what database people call conjunctive queries (with some extensions, such as optional steps).  Such queries cannot perform transitive closure queries such as reachability in a graph.
> > 

> > Does it make sense to have a query language that can take a directed graph and calculate the reachability relation on that graph?  This is conceptually no problem, and this is what transitive closure in some flavors of SQL allows you to do.  I see no reason why RDF would have to be changed to allow this in a query language such as SPARQL.  And I think this is what would be needed for querying provenance.
> > 

> > However, the discussion in the thread you linked to seems to be about something else:
> > 

> > Does it make sense to have a query language that returns graphs whose edges have (implicit) semantics "this edge is transitive", when the underlying graph model does not have such edges?  I think this is what Pat Hayes was arguing against in the thread; while I agree this is potentially confusing, it doesn't seem relevant to the kind of transitive queries needed for provenance.
> > 

> > But this is based on < 15 minutes of thinking.  So maybe I'm missing something.
> > 

> > --James
> > 

> > 

> > On Mar 19, 2010, at 4:11 PM, Satya Sahoo wrote:
> > 

> > Hi,
> > The link to discussion thread on SPARQL transitive closure: http://lists.w3.org/Archives/Public/public-swbp-wg/2005Dec/0036.html
> >  
> > Thanks.
> >  
> > Satya
> >  
> > Kno.e.sis Center
> > http://knoesis.wright.edu/researchers/satya
> > 


> > The University of Edinburgh is a charitable body, registered in
> > Scotland, with registration number SC005336.