[
https://issues.apache.org/jira/browse/OPENJPA-2549?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14231547#comment-14231547
]
Rick Curtis commented on OPENJPA-2549:
--------------------------------------
If setting sendStringParametersAsUnicode=false as a jdbc property makes your
problem go away, I don't think this is an OpenJPA problem... (but I may be
wrong?) I suspect the reason you are seeing different performance when using
only JDBC to execute the query is because you are going down a slightly
different path than the OpenJPA runtime.
> As soon as OpenJPA comes into the picture it seems to pass ntext() conversion
> down to the driver
I don't believe that OpenJPA is calling ntext()?
If you get the time / care enough, please do a bit more debugging to see what
is different between your JDBC test and what the OpenJPA runtime is doing. Let
us know if you have any questions / need some pointers.
> Slow query with full index scan when using OpenJPA query with string parameter
> ------------------------------------------------------------------------------
>
> Key: OPENJPA-2549
> URL: https://issues.apache.org/jira/browse/OPENJPA-2549
> Project: OpenJPA
> Issue Type: Bug
> Reporter: brian yoder
>
> I am not sure if this is an OpenJPA issue or not but here goes:
> This happens with MS SQL JDBC driver type 4
> -- This takes 7 seconds or more - full index scan
> Query q = em.createNativeQuery("select top 100 email_msg_id from im_email_msg
> where email_id like ? and firm_id = ?");
>
>
> -- This takes 10 milliseconds - index seek
> c.prepareCall("select top 100 email_msg_id from im_email_msg where email_id
> like ? and firm_id = ?");
> email ID parameter is a string, firm ID is integer. I have a work arround on
> this issue by setting the “sendStringParametersAsUnicode=false” JDBC
> parameter, but I think the issue may actually be OpenJPA.
> The reason I think this is an OpenJPA issue is that when I do a native call
> using the JDBC Connection API it works fine. As soon as OpenJPA comes into
> the picture it seems to pass ntext() conversion down to the driver, which is
> causing full index scan, since MS SQL Server is actually converting all
> values to UNICODE. The data in my column is text, not ntext, therefore no
> ntext should be used. But for some reason I think OpenJPA is calling the
> incorrect setString function, or somehow using unicode as the string paramer.
> This link may be helpful:
> http://stackoverflow.com/questions/8948872/jpa-hibernate-native-query-for-prepared-statement-slow
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
