RE: attempting to tackle ACTION-514

>  Regarding the algorithm in section 3.1 it comes from section 18.2.2.5 in the main query documtent, 
> there is just a new case which is SERVICE.
 
Would it be a problem to put the whole algorithm and just mark the new part with bold face?
 
> These are the variables which will be returned as a Select vars is done
 
So, why can't it then be jsut treated (on the level of the definition) as 'SELECT *' (then you don't have to explain all that or do I miss something?
 
Axel
 
-- 
Dr. Axel Polleres
axel.polleres@deri.org    http://www.polleres.net/

________________________________

From: Carlos Buil Aranda [mailto:cbuil@fi.upm.es]
Sent: Mon 26/09/2011 21:23
To: Polleres, Axel
Cc: Gregory Williams; SPARQL Working Group
Subject: Re: attempting to tackle ACTION-514


Hi Axel and sorry for this late reply. 

during all the emails between Greg and me I've been trying to fix the problems that Greg saw in the document. His main concern was in the SERVICE VAR section, in which the algorithm was evaluated using a bottom up approach. Using that approach it was not possible to assure that a variable was bound or not. So me marked the section as informative, not providing the definition section. Apart from that, the other main issue remaining was the one aboutthe IRI that you answered before.

Regarding the algorithm in section 3.1 it comes from section 18.2.2.5 in the main query documtent, there is just a new case which is SERVICE.

Regarding what "the set of variables in-scope in pattern P" means, that means the variables that are in pattern P which is evaluated inside the SERVICE. These are the variables which will be returned as a Select vars is done.

I uploaded the changes you suggested to the CVS.

Carlos


2011/9/20 Axel Polleres <axel.polleres@deri.org>


	Carlos, Greg,
	
	
	I had another look on Greg's comments on Fed-Query as per the thread starting at
	       http://lists.w3.org/Archives/Public/public-rdf-dawg/2011JulSep/0007.html
	going upto:
	
	       http://lists.w3.org/Archives/Public/public-rdf-dawg/2011JulSep/0035.html
	and
	       http://lists.w3.org/Archives/Public/public-rdf-dawg/2011JulSep/0165.html
	
	Admittedly, I am a bit confused now (probably just because I haven't tackled that earlier and seem to
	have difficulties recunstructing the "history" here)... when I look at section 3.1 in the current incarnation of
	  http://www.w3.org/2009/sparql/docs/fed/service
	it seems, some of the things referred to in the mail thread seem to be lost... Am I missing something? Looking at the right version?
	
	Particularly, as for http://lists.w3.org/Archives/Public/public-rdf-dawg/2011JulSep/0035.html, I am not sure anymore whether all that is referred to
	in this email is still in the current draft... Can you summarize/explain again in one mail what is the exact remaining issue? For me,
	the "algorithm" in section 3.1 is in its current state quite confusing and not really clear.
	
	As for Section 3.2 and the comment in http://lists.w3.org/Archives/Public/public-rdf-dawg/2011JulSep/0165.html
	
	> "The evaluation of Service is defined in terms of the SPARQL Results [RESULTS] returned
	> by a SPARQL Protocol [SPROT] execution of the nested graph pattern"
	>
	> The "RESULTS" link is to the XML Results document. Can it be defined also in terms of the (new) JSON Results document?
	
	Well, I think this is anyways a bit tricky.
	eval() is the evaluation of an algebra expression with respect to an active graph... so, it is supposed to return a
	multiset of solution mappings, whereas this sentence seems to suggest that we are talking about an XML results document here. Also I find the
	 if IRI
	in the definition a bit strange
	... I suggest to let go of the pseudo-algorithm for "Definition: Evaluation of a Service Pattern" in Section 3.2
	and just write it in more natural language:
	
	-------------------------------
	The evaluation of Service is defined in terms of the multiset of solutions corresponding to the result returned by a SPARQL Protocol [SPROT]
	execution of the nested graph pattern against a SPARQL endpoint:
	
	 Definition: Evaluation of a Service Pattern
	
	 Let
	    - iri be an IRI,
	    - vars be the set of variables in-scope in pattern P,
	    - ?0 the solution set with one empty solution, and
	    - SilentOp be a boolean variable to indicate that SERVICE execution should ignore errors when true.
	
	 Then:
	   eval(D(G), Service(IRI,P,SilentOp)) = Invocation( iri, vars, P, SilentOp )
	
	where: Invocation(IRI, Vars, P, SilentOp) is
	    * the multiset of solution mappings corresponding to the results of executing query
	     SELECT Vars WHERE P against the service endpoint with IRI iri, in case of a successful service invocation according to the SPARQL protocol, and otherwise
	    * ?0, in case SilentOp is true, and otherwise
	    * err
	-------------------------------
	
	What is still not quite clear to me is what "the set of variables in-scope in pattern P" exactly means, can you clarify?
	
	Axel

Received on Tuesday, 27 September 2011 07:19:05 UTC