- From: <jos.deroo@agfa.com>
- Date: Tue, 5 Jul 2005 22:56:26 +0200
- To: kendall@monkeyfist.com
- Cc: public-rdf-dawg@w3.org
Hi, Kendall
When I said in the tcon that we used bnodes in queries to query billions
of triples I basically meant ( items ) patterns used in queries
and the billions is when we do "24 hours a day - 7 days a week" tests :)
A typical scenario is select some relevant sets of triples out of
relational databases using some running code like
http://cvs.sourceforge.net/viewcvs.py/eulersharp/2004/01swap/src/euler/Codd.java?rev=1.7
which generically maps a database-table-collumn to an RDF-property
such as <http://host.domain/database/table#collumn>. HTTP GET of
http://host.domain/database.url-encoded-SQL-SELECT-query then gives
triple set
[ table1:collumn1 value01; table2:collumn2 value02; ... ].
[ table1:collumn1 value11; table2:collumn2 value12; ... ].
...
Out of those we make e.g. OWL constructs like :C1 owl:oneOf (:v1 :v2)
or triples like in
http://eulersharp.sourceforge.net/2003/03swap/pi-data.n3
Those triples are then further queried with ( items ) patterns
using N3 filter rules {WHERE-triples} => {CONSTRUCT-triples} like in
http://eulersharp.sourceforge.net/2003/03swap/pi-query.n3
and as you can see there are those ( items )'s in the {WHERE-triples}
--
Jos De Roo, AGFA http://www.agfa.com/w3c/jdroo/
Received on Tuesday, 5 July 2005 20:56:48 UTC