[
https://issues.apache.org/jira/browse/HIVE-9512?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14301717#comment-14301717
]
Jesus Camacho Rodriguez commented on HIVE-9512:
-----------------------------------------------
Thanks! I agree with you, it could be generalized; this is a fix for that
specific case.
Just for giving a bit more info on what was going on with UDFs with the former
patch in case it helps.
E.g. consider udf_inline.q. You get a plan TS[0]-UDFT[1]-Limit[2]-FS[3].
The stats were lost after UDFT was applied because they are stored for input
column col_0, and they propagate to the Limit operator still referring to col_0.
However, the UDF generates tuples with schema (col_1, col_2).
As the logic for calculating the stats for the limit operator was looking for
stats for columns col_1, col_2, and there is a mismatch schema-stats, it was
removing the stats.
> HIVE-9327 causing regression in stats annotation
> ------------------------------------------------
>
> Key: HIVE-9512
> URL: https://issues.apache.org/jira/browse/HIVE-9512
> Project: Hive
> Issue Type: Bug
> Affects Versions: 1.2.0
> Reporter: Prasanth Jayachandran
> Assignee: Jesus Camacho Rodriguez
> Fix For: 1.2.0
>
> Attachments: HIVE-9512.01.patch, HIVE-9512.02.patch, HIVE-9512.patch
>
>
> HIVE-9327 causes regression to statistics annotation test case. Regression
> can be seen here
> https://github.com/apache/hive/blob/trunk/ql/src/test/results/clientpositive/annotate_stats_select.q.out#L1065
> The expected data size is 194 but 0 is returned.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
