[
https://issues.apache.org/jira/browse/MAPREDUCE-5478?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Chris Nauroth updated MAPREDUCE-5478:
-------------------------------------
Hadoop Flags: Reviewed
+1 for the patch. I verified by running {{TestTeraSort}}. It does look like
the subclass is unnecessary at this point. The only difference will be that
{{FileSplit#toString}} doesn't print the locations, but that's just a cosmetic
thing.
> TeraInputFormat unnecessarily defines its own FileSplit subclass
> ----------------------------------------------------------------
>
> Key: MAPREDUCE-5478
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-5478
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Components: examples
> Affects Versions: 2.1.0-beta
> Reporter: Sandy Ryza
> Assignee: Sandy Ryza
> Priority: Minor
> Attachments: MAPREDUCE-5478.patch
>
>
> TeraInputFormat defines its own TeraFileSplit subclass of FileSplit that adds
> a locations field, which is already included in FileSplit.
> This is causing MR2 TeraSort to fail on MR1, which, for a System.arraycopy,
> requires splits to be of the FileSplit class. While nobody is promising that
> everything that runs on MR2 should run on MR1, fixing this would be easy and
> make it possible to compare MR2 TeraSort performance between MR1 and MR2.
> We should just get rid of TeraFileSplit and use FileSplit.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
