[
https://issues.apache.org/jira/browse/HIVE-20305?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16587022#comment-16587022
]
Teddy Choi commented on HIVE-20305:
-----------------------------------
As [~sershe] wrote, OrcSplit would guard it by using non-LLAP IO. However, it
would be better if we can make LLAP IO work on more cases.
> LlapRecordReader uses OrcInputFormat.getRootColumn(false)
> ---------------------------------------------------------
>
> Key: HIVE-20305
> URL: https://issues.apache.org/jira/browse/HIVE-20305
> Project: Hive
> Issue Type: Bug
> Components: llap, Transactions
> Affects Versions: 3.0.0
> Reporter: Eugene Koifman
> Priority: Major
>
> LlapRecordReader uses OrcInputFormat.getRootColumn(false) so it seems to
> assume that
> if {{AcidUtils.isFullAcidScan(jobConf)}} then the underlying file has acid
> meta columns in it. That is not true, for data added via Load Data, Add
> Partition or converting flat table to full CRUD acid via Alter Table (by
> setting transactional=true tbl property).
> cc [~teddy.choi]
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
