[
https://issues.apache.org/jira/browse/CRUNCH-22?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13419811#comment-13419811
]
Matthias Friedrich commented on CRUNCH-22:
------------------------------------------
Does anybody know the reason for Hadoop's object pooling? MR applications are
typically I/O-bound so garbage collector overhead shouldn't matter much. I'm
just curious, I've seen lots of bugs because of this.
> Join functions break when non-mapped types are used as a join key
> -----------------------------------------------------------------
>
> Key: CRUNCH-22
> URL: https://issues.apache.org/jira/browse/CRUNCH-22
> Project: Crunch
> Issue Type: Bug
> Reporter: Gabriel Reid
> Assignee: Gabriel Reid
> Fix For: 0.3.0
>
> Attachments: CRUNCH-22.patch
>
>
> When non-mapped types (i.e. custom Writables or Avro records) are used as a
> join key in all join functions, the join does not operate correctly (and
> becomes closer to producing a Cartesian product) due to object re-use in
> Hadoop.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
