db-derby-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Kristian Waagan <Kristian.Waa...@Sun.COM>
Subject Re: Upgrading Issues
Date Thu, 05 Nov 2009 09:52:16 GMT
Allgeyer, Jessica wrote:
> Kristian
> Because of performance issues, we are not able to simply skip the replace call and use
a 'delete from table' execute. We have tried that before and the delays are too significant.

I see.
What about using a minimal dummy file for import, then deleting the 
single row after the import?
(I know, awkward, and I don't know your performance requirements)

In any case, I logged the bug as DERBY-4435 [1].

> Do you know if this issue is being addressed now and have a possible ETA? 

I'm not aware of anyone looking at the issue at the moment. Watch the 
Jira issue for updates.
Note that there are no planned 10.5 maintenance releases at the moment. 
You may want to consider rolling your own build from the 10.5 branch 
(after the bug has been addressed).
The plans for the next feature release can be found on the wiki [2].


[1] https://issues.apache.org/jira/browse/DERBY-4435
[2] http://wiki.apache.org/db-derby/DerbyTenSixOneRelease

> Thanks
> Jessica Allgeyer
> (513) 763 - 8523
> Any fool can make things bigger, more complex, and more violent. It takes a touch of
genius - and a lot of courage - to move in the opposite direction. -albert einstein
> -----Original Message-----
> From: Kristian.Waagan@Sun.COM [mailto:Kristian.Waagan@Sun.COM] 
> Sent: Wednesday, November 04, 2009 12:27 PM
> To: Derby Discussion
> Cc: Richard.Hillegas@Sun.COM; Combs, Jeremy; McGrath, Phil; Ritchie, Erin; Lewis, Chris;
Allgeyer, Jessica
> Subject: Re: Upgrading Issues
> Allgeyer, Jessica wrote:
>> Kristian/Rick
>> I've included 2 attachments. One is the full stack trace for the NullPointerException.
The other is the file the call is trying to use. The file is zero bytes and does not contain
any characters. Our platform is Windows and we are using Java to run the call.  Below is the
full import call. 
>> 'C:\bea\projects\claims-dev\unzippedclaims\CHACRGE.rcsv','|','``',null
>> ,1)
> A quick (and not that thorough) debugging session seems to have led me to the problem,
which is related to the addition of collation support.
> The NPE happens here (from line 2348) in InsertResultSet.emptyIndexes:
>             newIndexCongloms[index] =
>                 tc.createAndLoadConglomerate(
>                     "BTREE",
>                     indexRows[index].getRowArray(),
>                     null, //default column sort order
>                     collation[index],
>                     properties,
>                     TransactionController.IS_DEFAULT,
>                     rowSources[index],
>                     (long[]) null);
> The problem is that the variable 'collation' hasn't been initialized and is null.
> Adding "int[][]             collation = new int[numIndexes][];" to the 
> top of the method makes the repro pass, but I haven't studied the code or run the regression
> Not sure there is much you can do about this, except for not using replacement in the
import or simply avoiding import of empty files.
> Maybe detecting empty files and executing a "delete from TABLE" would do?
> (until the bug has been fixed)
> Regards,

View raw message