You are viewing a plain text version of this content. The canonical link for it is here.

Posted to user@flink.apache.org by xiatao123 <ta...@udacity.com> on 2018/02/14 23:15:05 UTC

Support COUNT(DISTINCT 'field') Query Yet?

SELECT TUMBLE_START(event_timestamp, INTERVAL '1' HOUR), COUNT(DISTINCT
session), COUNT(DISTINCT user_id), SUM(duration), SUM(num_interactions) FROM
unified_events GROUP BY TUMBLE(event_timestamp, INTERVAL '1' HOUR)

I have the above statement my flink query running on Flink 1.3.2, but got
the error message Caused by: org.apache.flink.table.api.TableException:
Cannot generate a valid execution plan for the given query

Is the feature supported yet? if Yes, in which version of flink? If no, any
timeline to support it? 

Thanks,
Tao



--
Sent from: http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/

Re: Support COUNT(DISTINCT 'field') Query Yet?

Posted by Fabian Hueske <fh...@gmail.com>.

Hi Tao,

DISTINCT aggregates in group windows are not supported yet.
There's currently a discussion on the dev mailing list about this feature
[1].

Since we are only a few days before the feature freeze of Flink 1.5.0, it
might be included in 1.6.0, about 4-5 months from now.

As a workaround, you can implement a custom user-defined aggregation
function [2] that performs distinct counts.

Cheers,
Fabian

[1]
https://lists.apache.org/thread.html/4bdf66ce56ad5561461c8947203ccefa3fe2e4a1e48e29236f1bb8fc@%3Cdev.flink.apache.org%3E
[2]
https://ci.apache.org/projects/flink/flink-docs-release-1.4/dev/table/udfs.html#aggregation-functions

2018-02-15 0:15 GMT+01:00 xiatao123 <ta...@udacity.com>:

> SELECT TUMBLE_START(event_timestamp, INTERVAL '1' HOUR), COUNT(DISTINCT
> session), COUNT(DISTINCT user_id), SUM(duration), SUM(num_interactions)
> FROM
> unified_events GROUP BY TUMBLE(event_timestamp, INTERVAL '1' HOUR)
>
> I have the above statement my flink query running on Flink 1.3.2, but got
> the error message Caused by: org.apache.flink.table.api.TableException:
> Cannot generate a valid execution plan for the given query
>
> Is the feature supported yet? if Yes, in which version of flink? If no, any
> timeline to support it?
>
> Thanks,
> Tao
>
>
>
> --
> Sent from: http://apache-flink-user-mailing-list-archive.2336050.
> n4.nabble.com/
>