Hi George,
When 4Store returns a set of results that it believes may not be
complete it will return a warning message. As yet I have not
encountered a query where I haven't be able to increase what they
describe as the "soft limit" and get it to return all the results so
there is no longer a warning message.
I think it is inevitable that any triplestore (or any database come to
that) needs to have some way of handling complex or data heavy queries
so that they don't kill the server. It does seem that 4Store's solution
is a fair compromise. I don't currently forsee how it works being a
showstopper but it is something we need to keep and eye on.
Regards
David Newman
On Thu, 2009-11-19 at 11:33 +0000, Jerzy Orlowski wrote:
Hi David
Sorry for not replying but I had 150 emails to read...
The size limit of a query might be a really big blocker in the future.
For example imagine a query that returns all links that we want to run
for making the statistics. There should be a method to do it in SPARQL
George
David R Newman wrote:
Hi All,
I have been working on upgrading myExperiment’s SPARQL endpoint to use
4Store (http://www.4store.org) from JENA. This should make it much
quicker at finding results than it has been in the past. I am
currently running a test instance of the 4Store-backed SPARQL endpoint
at http://rdf.myexperiment.org/sparql4. So if people want to try out
some of their more complex SPARQL queries to see if they run quicker
that would be most appreciated.
I have endeavoured to maintain all the previous functionality in as
similar way as possible but inevitably there are some differences:
- No JSON results format
- 4Store does not have a way of timing out queries that take too long.
They either return only partial results and/or send a warning message
about the complexity. Instead of a timeout option I have provided a
“Resource Usage” option. This is defaulted to 5% which is suitable for
most basic queries but if you get a warning saying that a query could
not return all results then increasing this percentage should help.
In theory the endpoint should be updated each morning much like the
current endpoint but I have only just sorted out the scripts to do
this and I wouldn’t want to guarantee their reliability at the moment.
Anyway, any comments/feedback would be most appreciated.
Regards
David Newman
------------------------------------------------------------------------
Scanned for malicious content. See headers for details.
------------------------------------------------------------------------
_______________________________________________
Myexperiment-discuss mailing list
address@hidden
http://lists.nongnu.org/mailman/listinfo/myexperiment-discuss