[
https://issues.apache.org/jira/browse/HIVE-22262?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17065249#comment-17065249
]
Vineet Garg commented on HIVE-22262:
------------------------------------
[~jcamachorodriguez] My thinking was that while introducing aggregate on top of
view we keep this count beside adding necessary aggregations like sum(fk3).
Since this count doesn' have any aggregate column it should be logically
equivalent.
What do you think?
> Aggregate pushdown through join may generate additional rewriting
> opportunities
> -------------------------------------------------------------------------------
>
> Key: HIVE-22262
> URL: https://issues.apache.org/jira/browse/HIVE-22262
> Project: Hive
> Issue Type: Sub-task
> Components: CBO, Materialized views
> Affects Versions: 3.1.2
> Reporter: Steve Carlin
> Assignee: Vineet Garg
> Priority: Major
> Attachments: eager-v2.sql
>
>
> In this case, there is a function used in the query and materialized view,
> but the aggregate is not being pushed down. Script is attached.
> Example query and materialized view:
> create materialized view av1 stored as orc as select fk1, fk2, fk3,
> to_date(fk4), sum(1) from fact group by 1, 2, 3, 4;
> explain cbo select pk1, dim2.fk4, sum(1), count(c1)
> from fact, dim2
> where to_date(fact.fk4) = dim2.fk4
> group by 1, 2
> order by 1, 2;
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
