[
https://issues.apache.org/jira/browse/HIVE-4825?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13717813#comment-13717813
]
Edward Capriolo commented on HIVE-4825:
---------------------------------------
Huge effort. I do see what your saying. The win is nice in that something is
either MapWork or ReduceWork and there are some classes that do not need to
redundantly set reduce tasks to 0 when they run on the map side. Even though
this patch touches many files it pretty much touch them all in a small way that
should not be too much trouble for anyone to deal with. I am +1, most of the
changes this would cause would be cosmetic. Im only trying to look out for
things that Navis and Yin have on the queue.
> Separate MapredWork into MapWork and ReduceWork
> -----------------------------------------------
>
> Key: HIVE-4825
> URL: https://issues.apache.org/jira/browse/HIVE-4825
> Project: Hive
> Issue Type: Improvement
> Reporter: Gunther Hagleitner
> Assignee: Gunther Hagleitner
> Priority: Minor
> Attachments: HIVE-4825.1.patch, HIVE-4825.2.code.patch,
> HIVE-4825.2.testfiles.patch, HIVE-4825.3.testfiles.patch, HIVE-4825.4.patch
>
>
> Right now all the information needed to run an MR job is captured in
> MapredWork. This class has aliases, tagging info, table descriptors etc.
> For Tez and MRR it will be useful to break this into map and reduce specific
> pieces. The separation is natural and I think has value in itself, it makes
> the code easier to understand. However, it will also allow us to reuse these
> abstractions in Tez where you'll have a graph of these instead of just 1M and
> 0-1R.
--
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
