[
https://issues.apache.org/jira/browse/HIVE-28488?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17877924#comment-17877924
]
Sungwoo Park commented on HIVE-28488:
-------------------------------------
On 10TB TPC-DS benchmark (tested with Hive 4 on MR3),
query 49, before: 26.1s, after: 25.3s
query 75, before: 224.2s, after: 204.8s
> Merge adjacent union distinct
> -----------------------------
>
> Key: HIVE-28488
> URL: https://issues.apache.org/jira/browse/HIVE-28488
> Project: Hive
> Issue Type: Improvement
> Reporter: Seonggon Namgung
> Assignee: Seonggon Namgung
> Priority: Major
> Attachments: 1.MergeAdjacentUnionDistinct.pptx
>
>
> Current Hive compiles
> "SELECT * FROM TBL1 UNION SELECT * FROM TBL2 UNION SELECT * FROM TBL3"
> to
> {code:java}
> TS - GBY - RS
> TS - GBY - RS - GBY - RS
> TS - GBY - RS - GBY {code}
> This can be optimized as follows:
> {code:java}
> TS - GBY - RS
> TS - GBY - RS
> TS - GBY - RS - GBY {code}
> Please check out the attached slides for detailed explanation and feel free
> to ask any questions or share suggestions. Also, it would be glad if one can
> share about better location of this optimization (e.g. SemanticAnalyzer,
> Calcite, etc.).
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
