[
https://issues.apache.org/jira/browse/HIVE-5872?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13845715#comment-13845715
]
Xuefu Zhang commented on HIVE-5872:
-----------------------------------
Thanks for the review, Prasad.
{quote}
In general, it looks like we need more exception logic for handling decimals in
UDAF (HIVE-5872, HIVE-5866). It might be useful to add a note in the dev guide
for future work ..
{quote}
I assume you are referring the following code snippet:
{code}
if (t == null) {
return warnedOnceNullMapKey;
}
{code}
I agree with your assessment. Currently Hive emits null as the only error
handling option. Thus, null check is (or is missed) everywhere in the code, not
specific to decimal. For a long run, I agree we need to have a better exception
handling, especially when we introduce different error handling (HIVE-5438).
> Make UDAFs such as GenericUDAFSum report accurate precision/scale for decimal
> types
> -----------------------------------------------------------------------------------
>
> Key: HIVE-5872
> URL: https://issues.apache.org/jira/browse/HIVE-5872
> Project: Hive
> Issue Type: Improvement
> Components: Types, UDF
> Affects Versions: 0.12.0
> Reporter: Xuefu Zhang
> Assignee: Xuefu Zhang
> Fix For: 0.13.0
>
> Attachments: HIVE-5872.1.patch, HIVE-5872.2.patch, HIVE-5872.3.patch,
> HIVE-5872.4.patch, HIVE-5872.patch
>
>
> Currently UDAFs are still reporting system default precision/scale (38, 18)
> for decimal results. Not only this is coarse, but also this can cause
> problems in subsequent operators such as division, where the result is
> dependent on the precision/scale of the input, which can go out of bound
> (38,38). Thus, these UDAFs should correctly report the precision/scale of the
> result.
--
This message was sent by Atlassian JIRA
(v6.1.4#6159)
