[
https://issues.apache.org/jira/browse/HIVE-10752?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14565401#comment-14565401
]
Mithun Radhakrishnan commented on HIVE-10752:
---------------------------------------------
bq. Given that HIVE-5193 broke some functionality and it was just for columnar
table performance improvement, in addition that patch provided in HIVE-10720
did still not solve the issue.
While I agree that HIVE-5193 did introduce a bug, I can't yet agree that we
should revert it. [~viraj] is currently testing whether the one-liner posted in
HIVE-10720 doesn't resolve the issue. (My understanding was that this does.)
I'll let him confirm shortly.
In the meantime, please consider that the fix
({{ColumnProjectionUtils.setReadColumnIDs(job.getConfiguration(), null);}}) is
only applied when {{requiredFieldsInfo == null}}, which is shorthand for Pig
requiring all columns. So the deserialization is not done in all cases. It's
only for when all fields are required. There isn't any loss of performance in
this case.
Am I missing something?
> Revert HIVE-5193
> ----------------
>
> Key: HIVE-10752
> URL: https://issues.apache.org/jira/browse/HIVE-10752
> Project: Hive
> Issue Type: Sub-task
> Components: HCatalog
> Affects Versions: 1.2.0
> Reporter: Aihua Xu
> Assignee: Aihua Xu
> Attachments: HIVE-10752.patch
>
>
> Revert HIVE-5193 since it causes pig+hcatalog not working.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
