[
https://issues.apache.org/jira/browse/HIVE-10131?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14484281#comment-14484281
]
Matt McCline commented on HIVE-10131:
-------------------------------------
The readCurrent/advance model was too complicated. Originally the bytes,
offset, and length were available by separate methods. But by moving to
returning a ByteSegmentRef, moving the the first/next model seems to simplify
everything.
> LLAP: BytesBytesMultiHashMap and mapjoin container should reuse refs
> --------------------------------------------------------------------
>
> Key: HIVE-10131
> URL: https://issues.apache.org/jira/browse/HIVE-10131
> Project: Hive
> Issue Type: Sub-task
> Reporter: Sergey Shelukhin
> Assignee: Matt McCline
> Attachments: HIVE-10131.01.patch, HIVE-10131.02.patch,
> HIVE-10131.03.patch, HIVE-10131.04.patch, HIVE-10131.05.patch,
> HIVE-10131.06.patch, HIVE-10131.07.patch, HIVE-10131.08.patch
>
>
> Refs are always allocated and cleared. Should be reused.
> Iterator is another option
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
