You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@tez.apache.org by "Siddharth Seth (JIRA)" <ji...@apache.org> on 2014/03/04 21:08:22 UTC
[jira] [Updated] (TEZ-910) Add a non-sorted non-grouped K-V Input
[ https://issues.apache.org/jira/browse/TEZ-910?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Siddharth Seth updated TEZ-910:
-------------------------------
Attachment: TEZ-910.1.wip.txt
WIP patch.
[~bikassaha], [~hitesh] - do you see any utility in separating the Input that is used in case of Braodcast (ShuffledUnorderedKVInput) and a similar unordered input used in case of Dynamic partitioning by Hive (Instead of getting physical output 0 only, which is the broadcast case, each task gets a specific physical output index) ?
The main utility I see is some minor error checking in case of braodcast - where we can try checking whether the input is set incorrectly based on the srcIndex. Other than that these two Inputs are the same (ShuffledUnorderedKVInput can work in either case).
> Add a non-sorted non-grouped K-V Input
> --------------------------------------
>
> Key: TEZ-910
> URL: https://issues.apache.org/jira/browse/TEZ-910
> Project: Apache Tez
> Issue Type: Improvement
> Reporter: Siddharth Seth
> Assignee: Siddharth Seth
> Attachments: TEZ-910.1.wip.txt
>
>
> This should not be linked to Broadcast in any manner. The current one (linked to Braodcast) restricts srcIndices as a form of error checking.
> Required by Hive Bucketed Map Joins.
--
This message was sent by Atlassian JIRA
(v6.2#6252)