[
https://issues.apache.org/jira/browse/HIVE-8988?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14244535#comment-14244535
]
Julian Hyde commented on HIVE-8988:
-----------------------------------
If I wrote {{select f(deptno, gender), count(*) from emp group by deptno,
gender having f(deptno, gender) > 100}} and f was a "regular" function, then
Calcite could push it down. And you'd probably want it to.
So we'd have to flag that "grouping_id" is not a regular function.
In my approach you apply the $Foo function to the indicator columns, which do
not exist until the Aggregate creates them, hence the function cannot be pushed
down. (The call to $Foo could still be pulled up.)
> Support advanced aggregation in Hive to Calcite path
> -----------------------------------------------------
>
> Key: HIVE-8988
> URL: https://issues.apache.org/jira/browse/HIVE-8988
> Project: Hive
> Issue Type: Improvement
> Affects Versions: 0.15.0
> Reporter: Jesus Camacho Rodriguez
> Assignee: Jesus Camacho Rodriguez
> Labels: grouping, logical, optiq
> Fix For: 0.15.0
>
> Attachments: HIVE-8988.01.patch, HIVE-8988.patch
>
>
> CLEAR LIBRARY CACHE
> To close the gap between Hive and Calcite, we need to support the translation
> of GroupingSets into Calcite; currently this is not implemented.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
