Mailing-List: contact dev-help@hive.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@hive.apache.org
Date: Sat, 26 Oct 2013 02:24:30 +0000 (UTC)
From: "Sergey Shelukhin (JIRA)" <jira@apache.org>
To: hive-dev@hadoop.apache.org
Message-ID: <JIRA.12673083.1381344121386.6510.1382754270888@arcas>
In-Reply-To: <JIRA.12673083.1381344121386@arcas>
References: <JIRA.12673083.1381344121386@arcas>
Subject: [jira] [Commented] (HIVE-5503) TopN optimization in
 VectorReduceSink
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit


    [ https://issues.apache.org/jira/browse/HIVE-5503?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13805966#comment-13805966 ] 

Sergey Shelukhin commented on HIVE-5503:
----------------------------------------

I have basically moved a lot of logic around TopN from ReduceSink to TopNHash, so TopNHash itself can handle flushing, disabling etc.
Also, the different behavior for different cases is now covered by a small private interface, not by entire class, so it doesn't have to be visible outside.
Finally, after that, vectorized batch processing has been added. It adds all keys to heap; then forwards or stores values for the keys that survived in the heap, same way as for single case.

> TopN optimization in VectorReduceSink
> -------------------------------------
>
>                 Key: HIVE-5503
>                 URL: https://issues.apache.org/jira/browse/HIVE-5503
>             Project: Hive
>          Issue Type: Sub-task
>            Reporter: Jitendra Nath Pandey
>            Assignee: Sergey Shelukhin
>         Attachments: HIVE-5503.patch
>
>
> We need to add TopN optimization to VectorReduceSink as well, it would be great if ReduceSink and VectorReduceSink share this code.


--
This message was sent by Atlassian JIRA
(v6.1#6144)