[
https://issues.apache.org/jira/browse/HIVE-19794?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16504479#comment-16504479
]
Prasanth Jayachandran commented on HIVE-19794:
----------------------------------------------
Added a config to force single split for order by.
Tez split grouping is horrendously complex, there are so many variables to tune
to get single split. Added some new tests but still there may be cases where
tez split grouper does not honor the split count request based on locality
(haven't tested this yet, also the unit tests are single locality).
> Disable removing order by from subquery in GenericUDTFGetSplits
> ---------------------------------------------------------------
>
> Key: HIVE-19794
> URL: https://issues.apache.org/jira/browse/HIVE-19794
> Project: Hive
> Issue Type: Bug
> Affects Versions: 3.1.0, 4.0.0
> Reporter: Prasanth Jayachandran
> Assignee: Prasanth Jayachandran
> Priority: Major
> Attachments: HIVE-19794.1.patch, HIVE-19794.2.patch
>
>
> spark-llap always wraps query under a subquery, until that is removed from
> spark-llap
> hive compiler is going to remove inner order by in GenericUDTFGetSplits.
> disable that optimization until then.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
