[
https://issues.apache.org/jira/browse/HIVE-15520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15819349#comment-15819349
]
Aihua Xu commented on HIVE-15520:
---------------------------------
When the data size gets larger, the performance improvement will be significant.
Different from Lag/Lead, avg() is not allowed from sum(1/avg(val)) over
(partition by k1 order by k2) because avg() is aggregation while lead/Lag is
not.
For sum, there should be such property, otherwise, our streaming support for
many functions including sum would also have issues.
> Improve the sum performance for Range based window
> --------------------------------------------------
>
> Key: HIVE-15520
> URL: https://issues.apache.org/jira/browse/HIVE-15520
> Project: Hive
> Issue Type: Sub-task
> Components: PTF-Windowing
> Reporter: Aihua Xu
> Assignee: Aihua Xu
> Attachments: HIVE-15520.1.patch, HIVE-15520.2.patch,
> HIVE-15520.3.patch, HIVE-15520.4.patch
>
>
> Currently streaming process is not supported for range based windowing. Thus
> sum( x ) over (partition by y order by z) is O(n^2) running time.
> Investigate the possibility of streaming support.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
