[
https://issues.apache.org/jira/browse/HIVE-5658?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13806011#comment-13806011
]
Gunther Hagleitner commented on HIVE-5658:
------------------------------------------
[~sershe]: Fear not, captain null is here :-P
This is not a bug - it's correct behavior. Count distinct is supposed to count
all non-null unique expressions, which is 0 for the null group in your query.
> count(distinct) produces confusing results for null
> ---------------------------------------------------
>
> Key: HIVE-5658
> URL: https://issues.apache.org/jira/browse/HIVE-5658
> Project: Hive
> Issue Type: Bug
> Reporter: Sergey Shelukhin
>
> {code}
> select csmallint, count(distinct(csmallint)) from alltypesorc group by
> csmallint limit 3;
> {code}
> produces:
> {noformat}
> NULL 0
> -16379 1
> -16373 1
> {noformat}
> There are records in table with NULL values; however the count in this case
> should be 1, not 0, it seems. This is with TopN disabled, so it is unrelated
> to the other bug
--
This message was sent by Atlassian JIRA
(v6.1#6144)
