PREFIX ORACLE_SEM_FS_NS timeout via Jena Adapter

Question

Hi,
I am executing SPARQL via Jena Adapter and would like to specify timeout.
I use PREFIX ORACLE_SEM_FS_NS: <http://oracle.com/semtech#timeout=5> , according the the sem dev guide Jena examples, but the timeout value has no effect at all.
For some heavy queries QueryExecution->execSelect() never returns
I found this thread - Re: Query IDs in ORACLE_SEM_FS_NS : 32 or 64 bits? , which discuss the similar subject. I tried to add qid to the prefix, but that did not help. I don't use Joseki endpoint nor OracleQueryMgtServlet to manage(kill) long queries. Maybe I will provide the latter function in the future, but I would like to start with that users specify the timeout and query would interrupt itself if not ready.
So, is there anything else I have to do in Java/Jena code or somewhere else?
Thanks, best regards
Jürgen
After some more testing it turns that the problem with timeout is related to certain type of queries only. Normally the timeout works.
But when LET statement is used in some way, then timeout does not occur and it seems the query simply hangs:
This query works and times out (if timeout value is set short enough):
SELECT DISTINCT
?HR_Zone
(count(?e) AS ?major_count)
WHERE {
?e rdf:type r:ST_Episode .
OPTIONAL
?e r:isClassifiedAs rn:major_st_episode .
OPTIONAL{
?e r:mappedHRZone ?zone .
?zone r:index ?HR_Zone
GROUP BY ?HR_Zone
This query does not finsih (in boundaries of my patients) and does not time out:
SELECT DISTINCT
?HR_Zone
(count(?e) AS ?major_count)
WHERE {
?e rdf:type r:ST_Episode .
OPTIONAL
?e r:isClassifiedAs rn:major_st_episode .
LET( ?emajor := ?e )
OPTIONAL{
?e r:mappedHRZone ?zone .
?zone r:index ?HR_Zone
GROUP BY ?HR_Zone
However, one might think that LET is not supported - but without ARQ functions it works. This query works and times out (with short timeout value)
SELECT DISTINCT
?emajor ?HR_Zone
WHERE {
?e rdf:type r:ST_Episode .
OPTIONAL
?e r:isClassifiedAs rn:major_st_episode .
LET( ?emajor := ?e )
OPTIONAL{
?e r:mappedHRZone ?zone .
?zone r:index ?HR_Zone
Edited by: user8793616 on 2010-sep-29 04:51

Answer 1

Hi Jurgen,
The timeout pragma is most effective when used with SPARQL 1.0 queries, since they are supported natively. Once you start using ARQ features - which are not available in Oracle so they are processed in the mid-tier - timeout becomes less reliable.
Cheers,
Vladimir