[
https://issues.apache.org/jira/browse/HIVE-21385?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16785855#comment-16785855
]
Jesus Camacho Rodriguez commented on HIVE-21385:
------------------------------------------------
I would say it depends. For instance, Aggregate may not increase data size, but
it may not reduce it much, and we would prevent reading in parallel using JDBC.
The advantage of keeping those operators in the Hive plan is that we can read
in parallel and compute aggregation/sort/union using Hive parallel operators
too.
For the time being, I believe disabling pushdown completely (already there) and
disabling operators that may prevent reading in parallel gives the user enough
flexibility.
> Allow disabling pushdown of non-splittable computation to JDBC sources
> ----------------------------------------------------------------------
>
> Key: HIVE-21385
> URL: https://issues.apache.org/jira/browse/HIVE-21385
> Project: Hive
> Issue Type: Improvement
> Components: CBO, StorageHandler
> Reporter: Jesus Camacho Rodriguez
> Assignee: Jesus Camacho Rodriguez
> Priority: Major
> Attachments: HIVE-21385.01.patch, HIVE-21385.patch
>
>
> Until pushdown is cost-based decision, we will be able to enable / disable
> pushdown of operators that prevent reading results from the JDBC connection
> in parallel.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
