[
https://issues.apache.org/jira/browse/HIVE-11528?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15051602#comment-15051602
]
Sergey Shelukhin commented on HIVE-11528:
-----------------------------------------
Yes, IIRC order by query always ends with a single reducer to sort the data.
Without order by, multiple reducers output the results to multiple files, and
then we read these.
I think the pipeline is the same for both cases, the question is whether the
last stage is a single reducer.
> incrementally read query results when there's no ORDER BY
> ---------------------------------------------------------
>
> Key: HIVE-11528
> URL: https://issues.apache.org/jira/browse/HIVE-11528
> Project: Hive
> Issue Type: Improvement
> Reporter: Sergey Shelukhin
> Assignee: Keisuke Ogiwara
>
> May require HIVE-11527. When there's no ORDER BY and there's more than one
> reducer on the last stage of the query, it should be possible to return data
> to the user as it is produced, instead of waiting for all reducers to finish.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
