[
https://issues.apache.org/jira/browse/FLINK-5315?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16563634#comment-16563634
]
Fabian Hueske commented on FLINK-5315:
--------------------------------------
I like {{myUdagg.distinct('a)}} and {{multiParamAgg.distinct('a, 'b)}} as
proposed by [~hequn8128].
I'm not sure about {{'a.count.distinct}} though, because of the point raised by
[~walterddr].
The query
{code}
table.groupBy('a).select('b.count.distinct)
{code}
could either mean
{code}
SELECT COUNT(DISTINCT b) FROM table GROUP BY a
{code}
or
{code}
SELECT DISTINCT cnt FROM (SELECT COUNT(b) AS cnt FROM table GROUP BY a)
{code}
On the other hand, the option {{'a.distinct.count}} would be inconsistent with
the approach for UDAGGs where distinct is applied on the agg function and not
the parameters.
> Support distinct aggregations in table api
> ------------------------------------------
>
> Key: FLINK-5315
> URL: https://issues.apache.org/jira/browse/FLINK-5315
> Project: Flink
> Issue Type: Sub-task
> Components: Table API & SQL
> Reporter: Kurt Young
> Assignee: Rong Rong
> Priority: Major
>
> Such as
> {code}
> t.select("count(distinct a), sum(b)")
> {code}
> or
> {code}
> t.select('a.count.distinct), 'b.sum)
> {code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
