[
https://issues.apache.org/jira/browse/ORC-380?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16518625#comment-16518625
]
Matt McCline commented on ORC-380:
----------------------------------
[~prasanth_j] FYI
> Add isOnlyImplicitConversion boolean function to SchemaEvolution
> ----------------------------------------------------------------
>
> Key: ORC-380
> URL: https://issues.apache.org/jira/browse/ORC-380
> Project: ORC
> Issue Type: Bug
> Components: ORCv2
> Reporter: Matt McCline
> Assignee: Matt McCline
> Priority: Critical
>
> The current hasConversion boolean function returns true if there is *any*
> conversion between the reader and file schemas. Prasanth pointed out this is
> too strict for the check I'm adding in HIVE-19951: "Vectorization: Need to
> disable encoded LLAP I/O for ORC when there is data type conversion (Schema
> Evolution)". He points out that integer type widening and float to double
> ought to be allowed.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
