[
https://issues.apache.org/jira/browse/DERBY-3004?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Knut Anders Hatlen updated DERBY-3004:
--------------------------------------
Attachment: d3004.stat
d3004.diff
Attaching a fix for this issue. The patch removes EmbedResultSet.finalize() and
adds EmbedStatement.finalize(). The new finalizer checks whether EmbedStatement
contains a reference to a ResultSet, and if it does and the ResultSet has a
single-use activation, the activation is marked as unused. If the ResultSet is
no longer referenced by the EmbedStatement, it means that it has been closed
already and no clean-up action is needed.
Derbyall and suites.All ran cleanly (except a failure in lang/compressTable.sql
which is also seen in the tinderbox).
> Get rid of the finalizer in EmbedResultSet
> ------------------------------------------
>
> Key: DERBY-3004
> URL: https://issues.apache.org/jira/browse/DERBY-3004
> Project: Derby
> Issue Type: Improvement
> Components: JDBC, Performance
> Affects Versions: 10.3.1.4
> Reporter: Knut Anders Hatlen
> Assignee: Knut Anders Hatlen
> Priority: Minor
> Attachments: d3004.diff, d3004.stat
>
>
> Finalizers make object allocation and garbage collection more expensive and
> should be avoided for object types that are frequently allocated.
> EmbedResultSet has a finalize() method and is frequently allocated for many
> types of load. The following comment was posted on DERBY-1876:
> By commenting out the finalize() method in EmbedResultSet, the time to run
> the test on my laptop is reduced from ~3 seconds to ~1.3 seconds
> (OpenSolaris, Sun Java SE 6). Since the finalizer is a no-op if the
> activation isn't single-use, I think we should investigate whether there are
> ways to avoid this cost for ResultSets that don't need the finalizer.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
