we don't have such plan yet. As a workaround you can create a hive view
with an additional column for cnt * val, and use the view as fact table in
your cube

On Fri, Jul 1, 2016 at 1:06 PM, Joel Victor <[email protected]> wrote:

> Hi,
>
> I have a fact table with two measure columns one is cnt and other is val
> with count & sum measures defined over both columns. Val has either median
> or average for that day.
> Now if I want to calculate mean & median for a month I would be required
> to do the following, sum(cnt * val) / sum(cnt).
> I run the following query select sum(cnt * val) / sum(cnt) from
> test_table where data_type=23 and event_date between '2016-04-01' AND
> '2016-04-30';
> I get a NullPointerException. Complete error attached to the mail.
> I am guessing this is probably due it not being able to find a realization
> that matches the sum(cnt * val) measure. If not is there something else I
> am doing wrong ?
> Usually when you try to query with a aggregation that is not a part of the
> cube the error is that it can't find any realizations. But this query just
> returns null.
>
> Are there any plans to support these kind of aggregates ?
>
> Thanks,
> -Joel
>



-- 
Regards,

*Bin Mahone | 马洪宾*

Reply via email to