[
https://issues.apache.org/jira/browse/HIVE-14214?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15386665#comment-15386665
]
Prasanth Jayachandran commented on HIVE-14214:
----------------------------------------------
Left some comments in RB.
I think many of the complications in the patch can be avoided if we just
provide a Reader api that returns true if there is conversion.
Reader.hasConversion(). Behind the scenes we should do all the magic of
determining if the conversion is required based on reader schema, file schema
and included. In OrcInputFormat the only place we need to disable PPD is in ETL
strategy, which creates ORC reader. If this reader returns hasConversion() then
we should disable PPD. Similarly for task side.
> ORC Schema Evolution and Predicate Push Down do not work together (no rows
> returned)
> ------------------------------------------------------------------------------------
>
> Key: HIVE-14214
> URL: https://issues.apache.org/jira/browse/HIVE-14214
> Project: Hive
> Issue Type: Bug
> Components: Hive
> Reporter: Matt McCline
> Assignee: Matt McCline
> Priority: Critical
> Attachments: HIVE-14214.01.patch, HIVE-14214.02.patch,
> HIVE-14214.03.patch, HIVE-14214.04.patch, HIVE-14214.05.patch,
> HIVE-14214.WIP.patch
>
>
> In Schema Evolution, the reader schema is different than the file schema
> which is used to evaluate predicate push down.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
