W3C home > Mailing lists > Public > public-rdf-dawg-comments@w3.org > May 2005

comments on ORDER BY / LIMIT / OFFSET

From: Libby Miller <libby@asemantics.com>
Date: Tue, 3 May 2005 04:08:46 -0700 (PDT)
To: public-rdf-dawg-comments@w3.org
Message-ID: <20050503035238.O30380@skutsje.san.webweaving.org>


re:

http://www.w3.org/TR/2005/WD-rdf-sparql-query-20050419/

In some of our experiments we're finding ORDER BY and LIMIT to be very
useful indeed, for things like: get me the most recent images (or news
items, or calendar entries etc) available. RDF Calendar applications in
particular usually need some sorting. Together ORDER BY and LIMIT are very
useful for RSS-reader-like applications, especially when processing them
using javascript, which gets overwhelmed by data relatively quickly.

In practically the all scenarios we encountered the sorting of literals and
URIs was sufficient, and full sorting of sub-graphs and connected
bNodes has been less relevant at the moment. We strongly feel that
ORDER BY clause should simply cover human-readable content such as
literals, and perhaps also URIs. Adding simple sorting to our native DB
layer of xsd:double, xsd:integer and xsd:dateTime (after normalization to
UTC) has been pretty straightforward by using well-known database
techniques such as b-trees, while full graph-branches sorting could be
tedious due to the unordered nature of a graph.

We have also noticed that having a LIMIT clause without a corresponding
(optional) OFFSET clause it would be a major limitation to the query
language; even though cursors are drawn as out of scope for this DAWG
phase, we strongly feel the need to have at least the simplest
offset/limit cursors for all our applications.

best wishes,

Libby

--
Libby Miller, @semantics S.R.L.
http://www.asemantics.com
Received on Tuesday, 3 May 2005 11:08:58 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 8 January 2008 14:14:48 GMT