[
https://issues.apache.org/jira/browse/HIVE-14222?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15375624#comment-15375624
]
Gopal V commented on HIVE-14222:
--------------------------------
Yes, the plan is not deserialized again for a new split.
The same operator tree is used but at the end each split it calls closeOp() and
then calls initializeOp().
This means we can hold onto hashtables and similar bits which were built during
a previous split.
> PTF: Operator initialization does not clean state
> -------------------------------------------------
>
> Key: HIVE-14222
> URL: https://issues.apache.org/jira/browse/HIVE-14222
> Project: Hive
> Issue Type: Bug
> Components: PTF-Windowing
> Affects Versions: 1.2.1, 2.2.0
> Reporter: Gopal V
> Assignee: Wei Zheng
> Attachments: HIVE-14222.1.patch
>
>
> PTFOperator::initializeOp() does not reset currentKeys to null.
> {code}
> if (currentKeys != null && !keysAreEqual) {
> ptfInvocation.finishPartition();
> }
> ....
> if (currentKeys == null) {
> currentKeys = newKeys.copyKey();
> } else {
> currentKeys.copyKey(newKeys);
> }
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
