[
https://issues.apache.org/jira/browse/HIVE-19493?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16471281#comment-16471281
]
Matt McCline commented on HIVE-19493:
-------------------------------------
[~vihangk1] thank you for catching this. I've started writing random data
VectorExpression unit tests (e.g. HIVE-19384 for IF statements). Given this
miss, I after I finish CAST statements I'm going to do date/timestamp area next.
The one method propagateNullsColCol in VectorUDFDateDiffColCol makes me a
little nervous.
I tried to add a comment to the review board but it hung...
{noformat}
output.isNull[0] = false;
{noformat}
is missing is the isRepeated and input.noNulls || !input.isNull[0] block.
The isRepeating block can be forced by inserting single rows in the Q file into
ORC (and Parquet?) and that causes ORC to return a VectorizedRowBlock with
isRepeating set.
I'm not sure how to stimulate the other paths other than making sure there are
more NULLs in the data...
So, some Q file changes and/or UT changes would be great.
> VectorUDFDateDiffColCol copySelected does not handle nulls correctly
> --------------------------------------------------------------------
>
> Key: HIVE-19493
> URL: https://issues.apache.org/jira/browse/HIVE-19493
> Project: Hive
> Issue Type: Bug
> Components: Vectorization
> Reporter: Vihang Karajgaonkar
> Assignee: Vihang Karajgaonkar
> Priority: Major
> Attachments: HIVE-19493.01.patch
>
>
> The {{copySelected}} method in {{VectorUDFDateDiffColCol}} class was missed
> during HIVE-18622
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
