[
https://issues.apache.org/jira/browse/HIVE-15338?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15713852#comment-15713852
]
Matt McCline commented on HIVE-15338:
-------------------------------------
[~jdere] [~sershe] I'm see a difference in the vectorized vs non-vectorized Q
file output for the updated vectorized_date_funcs.q in the patch.
In the VectorUDFDateDiffScalarCol class, when the scalar parameter is a
TIMESTAMP we call:
{code}
date.setTime(timestampValue.getTime());
baseDate = DateWritable.dateToDays(date);
{code}
but GenericUDFDateDiff executes apparently different logic (it uses a
TimestampConverter object). I'll have to dig in a figure out what is
happening...
> Wrong result from vectorized DATEDIFF with scalar parameter of type DATE
> ------------------------------------------------------------------------
>
> Key: HIVE-15338
> URL: https://issues.apache.org/jira/browse/HIVE-15338
> Project: Hive
> Issue Type: Bug
> Components: Hive
> Reporter: Matt McCline
> Assignee: Matt McCline
> Priority: Critical
> Attachments: HIVE-15338.01.patch
>
>
> Vectorization in vectorized DATEDIFF when parameter is (date '12-01-16',
> column). Jason noted that ('12-01-16', column) treats the first parameter as
> a string and is a different case.
> Current Q file test vectorized_date_funcs.q DOES NOT test the date scalar
> type case.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
