Re: [TEST] simple SPARQL condition and SNAF condition

Francois Bry wrote:
> Dear All,
> 
> The issue Jos has raised with his suggestion to include builtin
> metapredicates like findall in the RIF condition language is, in my
> opinion, that of grouping (and implicitely aggregation) in both SPARQL
> and RIF.
> 
> These are by no means sdimple issues, whatver examples might let think.
> These are difficult semantics and evaluation issues, and also (often
> neglected) language design issues. They are in my opinion as difficult
> as nonmonotonic negation.

yes, aggregates and grouping need some form of closing the world or 
scoping... for a good article defining the semantics of aggregates in LP 
style rules possibly involving even unstratified negation as failure and 
disjunctive rules, see:

Wolfgang Faber, Nicola Leone, Gerald Pfeifer: Recursive Aggregates in 
Disjunctive Logic Programs: Semantics and Complexity. JELIA 2004: 200-212

(Remark for LP enthusiasts:  This paper has a very elegant reformulation 
of the Gelfond-Lifschitz reduct for defining the stable model semantics 
generalized to programs with aggregates! ;-))

This is, to my knowledge, the most general semantics for aggregates in 
rules so far.

axel


> I feel a bit concerned that the RIF might include metapredicates like
> findall in its condition language without these issues being properly
> worked out - in RIF and SPARQL.
> 
> Regards,
> 
> Francois
> 
> 
>>>>unifies ?ANSWER with a list that contains all the
>>>>instantiations of ?SELECT satisfying the ?WHERE
>>>>clause in the ?SCOPE of all asserted n3 formulae
>>>>and their log:conclusion""";
>>>>  a rdf:Property;
>>>>  rdfs:domain log:Formula;
>>>>  rdfs:range rdf:List.
>>>>
>>>>I mean useful for 2 reasons
>>>>1/ it is a way to have a simple SPARQL condition
>>>>2/ it is a way to have a SNAF condition when used as
>>>>   ?SCOPE e:findall (?SELECT ?WHERE rdf:nil).
>>>>
>>>>There is a test case result at
>>>>http://eulersharp.sourceforge.net/2006/02swap/medicE.n3
>>>>      
>>
>>  
> 
> 
> 
> 
> 


-- 
Dr. Axel Polleres
email: axel@polleres.net  url: http://www.polleres.net/

Received on Tuesday, 19 September 2006 14:50:30 UTC