So, you hit this bug, https://issues.apache.org/jira/browse/KYLIN-1971

2016-09-21 19:41 GMT+08:00 Sandeep Khurana <sand...@infoworks.io>:

> Found the issue, when fact and dimension have same column names on which
> group by is being done then this problem occurs. I removed similar named
> column from the fact table.
>
> On Tue, Sep 20, 2016 at 2:13 AM, Sandeep Khurana <sand...@infoworks.io>
> wrote:
>
>> btw, we are using Kylin version 1.5.2
>>
>> On Tue, Sep 20, 2016 at 2:03 AM, Sandeep Khurana <sand...@infoworks.io>
>> wrote:
>>
>>> Hello
>>>
>>> I have a query where i do group by on a varchar column. The column
>>> values are long sentences (not just single words). This column is part of a
>>> dimension table.
>>>
>>> When i select just from dimension table with this group by then I do
>>> get ~2000 records   .
>>>
>>> But when I join this dimension with the fact table and run the group by
>>> query then I get just 1 record as Kylin somehow assumes the VARCHAR column
>>> values as NULL. There is not even a single row which has value of this
>>> VARCHAR field as null.
>>>
>>> Same query I copy paste and run on the Hive tables, I do get more than
>>> thousand records.
>>>
>>> Strange thing is when I change the aggregate column to another VARCHAR
>>> column (city_name) whose values are just one word and run on kylin SQL
>>> editor then I do get proper records .
>>>
>>> 2 questions
>>>
>>> - Any idea why such behaviour ? Especially when Hive gives proper
>>> records whereas kylin returns just one record which has value of this big
>>> varchar field as NULL.
>>>
>>> - Is there any work around?
>>>
>>>
>>
>>
>
>
> --
> Architect
> Infoworks.io
> http://Infoworks.io
>



-- 
With Warm regards

Yiming Liu (刘一鸣)

Reply via email to