db-derby-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Daniel John Debrunner <...@apache.org>
Subject Re: [jira] Commented: (DERBY-1315) largeCodeGen fails with embedded and framework DerbyNetClient
Date Tue, 08 Aug 2006 20:54:52 GMT
A B (JIRA) wrote:
>     [ http://issues.apache.org/jira/browse/DERBY-1315?page=comments#action_12426705 ]

>             
> A B commented on DERBY-1315:
> ----------------------------
 hate to say it, but it looks like this regression is caused by the
changes for DERBY-805.   Before the DERBY-805 changes Derby was
generating plans for subqueries that did not agree with the optimizer's
choices.  Phase 1 of the DERBY-805 patches made it so that Derby
generates the plans chosen by the optimizer, as one might expect.  In
order to do that, I added extra state to the optimizable nodes to keep
track of which query plan went with which level of query.
> 
> My guess is that this extra state is causing the optimizer to increase its memory usage--and
thus for very large queries, we end up running out of memory  


Any pointers to the type of extra state kept around. Is it query tree
nodes, collections, something else?

Dan.



Mime
View raw message