Re: A reason for dropping seperate AND clauses

Phil Dawes wrote:
> Hi Andy
> 
> Seaborne, Andy writes:
>  > Phil Dawes wrote:
>  > > 
>  > > I don't understand this - isn't ?label bound to the set of all URIs
>  > > and literals that match the regex '*foo*'?
>  > 
>  > Not in any implementation I know of - there is no access to the set of all URIs 
>  > and literals.  RDF is only triples.
>  > 
>  > Is this query valid to you?
>  > 
>  > select ?label
>  > where (?label LIKE "%foo%")
>  > 
> 
> In veudas - yes. 
> This is triples - i.e. the set of virtual triples that match:
>   (?label LIKE "%foo%").
> e.g. ("foobah" LIKE "%foo%") ("foot" LIKE "%foo%").
> 
> It's not RDF though, since ?label can match literals.  Actually
> 'LIKE' has domain 'literal' in the veudas query engine, so you'd only
> get literal results to this query.
> 
> Literal subjects work fine within veudas internals, not really by
> design, but simply because the store layout doesnt restrict this.
> 
> But thanks - this does explain to me why sparql needs the AND clauses.
> 
> Out of interest, if RDF did move to allow literals as subjects, would
> this remove the necessity for AND?

Somewhat.  Not all constraints are binary predicates.  And it can be 
inconvenient to write "?x + 3 < 2 * ?y" out in triples.

	Andy

> 
> Thanks again,
> 
> Phil
> 
> 
> 

Received on Friday, 28 January 2005 14:59:09 UTC