- From: Steve Harris <steve.harris@garlik.com>
- Date: Mon, 23 Oct 2006 13:43:54 +0100
- To: public-rdf-dawg-comments@w3.org
[sorry, replying from web interface, so no quote or in-reply-to header]
I don't think it's a real problem. My RDF store (JXT) handles
anything up to 10 bn triples and can deal with regex queries. Just
because it's in the spec doesn't mean you're required to use it in
queries.
Other query forms, such as
SELECT *
WEHRE { ?v :p ?w . ?x :q ?y }
or excessive use of UNION are far more expensive on the whole, and
have to be handled somehow.
Many regex queries can be reduced to (case insensitive) prefix
substring matches, or simpler classes of pattern matching, and those
that can't can be refused by the engine should the query be too
onerous. For example by default my engine refuses to fully execute
things like:
SELECT *
WHERE { ?x ?y ?z . FILTER(regex(str(?x), ".*foo", "i") }
as it has no indexes to bring to bear on the problem, and
individually processing 10 billion objects is very time consuming.
- Steve
Received on Monday, 23 October 2006 12:44:54 UTC