[
https://issues.apache.org/jira/browse/IMPALA-10400?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17252964#comment-17252964
]
Tim Armstrong commented on IMPALA-10400:
----------------------------------------
There's no real right answer here. I don't think we want to revert the change
cause it would just cause more problems going forward. It's an option to add a
compatibility flag I guess, but I don't know how much complexity that would add.
For the two-argument versions, having the type depend on the value of the
second argument like some of the examples I think is a bad idea and leads to
problems like IMPALA-7613 that we have for decimal.
> Floor/Ceil/Trunc should return BIGINT instead of DOUBLE
> -------------------------------------------------------
>
> Key: IMPALA-10400
> URL: https://issues.apache.org/jira/browse/IMPALA-10400
> Project: IMPALA
> Issue Type: Improvement
> Reporter: Sheng Wang
> Priority: Major
>
> Floor/Ceil/Trunc returns BIGINT in version 2.12.0, but change to DOUBLE in
> IMPALA-6230. I think these functions should use BIGINT as return value type
> instead of DOUBLE. We compared severals systems:
> *mysql*：
> floor(num): BIGINT
> ceil(num): BIGINT
> truncate(num, d): d>0 DOUBLE,d=0 BIGINT,d<0 0
> *hive 2.1.1*：
> floor(num): BIGINT
> ceil(num): BIGINT
> trunc(date, format): truncate date
> *clickhouse*：
> floor(num): BIGINT
> ceil(num): BIGINT
> truncate(num, [d])，d>0 DOUBLE,d=0 BIGINT,d<0 0
> *greenplum*：
> floor(num): BIGINT
> ceil(num): BIGINT
> truncate(num, [d])，d>0 DOUBLE,d=0 BIGINT,d<0 0
> *spark*：
> floor(num): BIGINT
> ceil(num): BIGINT
> trunc(date, format): truncate date
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-all-unsubscr...@impala.apache.org
For additional commands, e-mail: issues-all-h...@impala.apache.org
