W3C home > Mailing lists > Public > semantic-web@w3.org > January 2013

Re: indexing and ordered searches

From: Martin Voigt <martin.voigt@tu-dresden.de>
Date: Tue, 22 Jan 2013 07:13:13 +0100
Message-ID: <50FE2DF9.9090202@tu-dresden.de>
To: Jeremy J Carroll <jjc@syapse.com>
CC: "semantic-web@w3.org" <semantic-web@w3.org>
Hi,

we are working with Oracle and its SEM_MATCH functionality. It allows 
for indexing different XSD datatypes. More information at

http://docs.oracle.com/cd/E11882_01/appdev.112/e25609/sdo_rdf_concepts.htm#CIHBJEDA

Regards,
Martin

Am 22.01.2013 01:11, schrieb Jeremy J Carroll:
>
> What triple stores offer any optimization for integer or other numeric
> queries, e.g. with the following query
>
> SELECT  ?a
> WHERE { ?a  eg:prop ?i .
>          FILTER ( 10 < ?i && ?i < 15 ) }
>
>
> a naive approach would be to find all eg:prop triples and compute the
> filter for each one (O(N) complexity); if the store only has say
> integers as objects of eg:prop then a Predicate/Object index could
> achieve the same result in O(log(N)) by a numeric binary chop on the
> Object part of the index.
>
> I looked at the TDB code and got the impression that the indices do in
> fact preserve ordering of at least some types, but could not see FILTERs
> being treated in the way indicated here … then I thought I would ask here.
>
> Jeremy
>

-- 
-----------------------------------
Dipl.-Medieninf. Martin Voigt
Wissenschaftlicher Mitarbeiter

Technische Universität Dresden
Fakultät Informatik
Institut für Software- und Multimediatechnik
Lehrstuhl Multimediatechnik
01062 Dresden

Tel.: +49 (351) 463-38037
Fax : +49 (351) 463-38518
Web: http://mmt.inf.tu-dresden.de/voigt
E-Mail: martin.voigt@tu-dresden.de
Skype: m.voigt.1982
Received on Tuesday, 22 January 2013 06:13:36 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 21:45:53 GMT