- From: Andy Seaborne <andy.seaborne@epimorphics.com>
- Date: Mon, 23 Aug 2010 12:02:13 +0100
- To: Axel Polleres <axel.polleres@deri.org>
- CC: SPARQL Working Group <public-rdf-dawg@w3.org>
On 23/08/10 11:51, Axel Polleres wrote: >> > But: observation: agg06.rq >> > >> > ------------------------------------------ >> > PREFIX :<http://www.example.org> >> > >> > SELECT ?P (COUNT(*) AS ?C) >> > WHERE { ?S ?P ?O } >> > HAVING (COUNT(*)> 0 ) >> > ------------------------------------------ >> > uses non-group key variable ?P in the SELECT line. >> > > thanks, that wad wrong, should've been without the ?P in the select. > >> > In ARQ, this results in, effectively, a new ?P (it's a different scope - >> > the use in the group is hidden because only variable from the group key >> > are in-scope and the mention of ?P in the SELECT line is a new, >> > different ?P and is hence unbound). > Wouldn't we rather want this to be a syntax error (don't remember in detail, but I think we discussed this, need to check) Could be (IIRC we'd suggested this as a warning if an impl chooses) - in ARQ ?P can be resued (e.g. select expression: SELECT (SAMPLE(?P) as ?P) (COUNT(*) AS ?C) {...} If we wish to make it an error, then we really ought to specify the condition to check (e.g. give an algorithm). Leaving it as-is is fine as well. Andy
Received on Monday, 23 August 2010 11:02:50 UTC