- From: Matthew Perry <matthew.perry@oracle.com>
- Date: Fri, 10 Feb 2012 09:30:00 -0500
- To: public-rdf-dawg@w3.org
Hi Axel, I would be ok with either option but have preference for option 2. To me, this seems less awkward to use and a little easier to implement. For example SELECT ?d WHERE { :a :p1{2,3} ?b . { SELECT DISTINCT ?b ?c WHERE { ?b :p2* ?c } } ?c :p3{1,2} ?d } is a bit awkward compared to SELECT ?d WHERE { :a :p1{2,3}/DISTINCT(:p2*)/:p3{1,2} ?d } Thanks, Matt On 2/9/2012 4:10 PM, Axel Polleres wrote: > (in completion of ACTION-587) > > Dear all, as discussed in the last Telco, we have several options on how to proceed with addressing comment JP-4 [1]. > If possible, I would like to get consensus on how to proceed here in the next Telco. > > In the previous Telco [2], we seemed to have consensus that we do not aim to switch the default behaviour from counting semantics to > distinct paths. > > Now two possibilities to proceed were discussed: > > Option 1... keep everything as it is in the grammar, and explain which DISTINCT path subqueries can be optimized: > As outlined in my email below, it might not be entirely trivial to argue in response to the comment that this > would be equivalent to the JP-4 proposed semantics, I am not 100% sure whether/how to define a rewriting to > wrap all path expressions into DISTINCT subqueries, such that it would be equivalent to their semantics > (e.g. regarding bnode [] shortcuts). > > Option 2 ... add DISTINCT around paths: It seems that sticking to our intended semantics and allowing - orthogonally to their > ALLPATHS keyword proposal the keyword DISTINCT( ) around path expressions switching to existential paths semantics would be > equivalent to the JP-4 existential paths semantics as outlined in Section 7.1 of their paper, and thus optimizable. > > Unlike someone sees a 3rd alternative, I would like to propose to decide between those two options next time > and proceed, discussion prior to the call on email would be appreciated. > > Option 2 might be easier to implement, but also requires us to go for another LC round, as it would change the grammar. > I think, in case we skip PR and manage to republish very soon, we would still manage to stay within time limits, but I would > also like to know the team contacts' opinion on that. > > best, > Axel
Received on Friday, 10 February 2012 14:30:25 UTC