[
https://issues.apache.org/jira/browse/HIVE-19995?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16523835#comment-16523835
]
Zoltan Haindrich commented on HIVE-19995:
-----------------------------------------
This is only about basic stats: like rowcount.
It seems OrccRecordUpdater already provides rowCountDelta correctly (and it
can be negative) ; so it will keep track of deletes as well - for free :D
I think this will probably work very reliably in general for acid tables.
> Aggregate row traffic for acid tables
> -------------------------------------
>
> Key: HIVE-19995
> URL: https://issues.apache.org/jira/browse/HIVE-19995
> Project: Hive
> Issue Type: Sub-task
> Components: Statistics, Transactions
> Reporter: Zoltan Haindrich
> Assignee: Zoltan Haindrich
> Priority: Major
>
> for transactional tables we store basic stats in case of explicit
> analyze/rewrite; but doesn't do anything in other cases....which may even
> lead to plans which oom...
> It would be better to aggregate the total row traffic...because that is
> already available; so that operator tree estimations could work with a real
> upper bound of the row numbers.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
