Vladimir Prus created KYLIN-2848:
------------------------------------
Summary: Query of non-aggregated columns returns bogus results
Key: KYLIN-2848
URL: https://issues.apache.org/jira/browse/KYLIN-2848
Project: Kylin
Issue Type: Bug
Reporter: Vladimir Prus
I have a cube called fact_secondary_event, with a dimension called has_text.
That dimension is not used as measure at all. I execute a query such as:
{code}
select dim_date_id, count(*), sum(has_text)
from mart.star_secondary_event
where has_text = 1
group by dim_date_id
{code}
Expected result: an error saying that there's no aggregation for `sum(has_text)`
Actual result: rows such as
{code}
2017-08-31,18598,1
{code}
In other words, for aggregation that cannot be computed, I get back '1'. I
don't think that behaviour of returning random values is helpful, and it's not
even easy to debug after the data travels through a few levels of backend and
UI code. Could an error be omitted in such cases?
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
