Hi, all

I am using kylin 1.6.0. I have met three problem:

1. in one of my Metrics, some of the values are NULL, when I tried to
calculate the average of the column, the COUNT function will not filter out
NULL value, which means the average result is biased. One solution I found
is using another column to mark whether the value is NULL or not, but there
are hundreds of columns like this. I don't think adding another hundreds of
mark column as dimensions is a good way. Any suggestions about this
situation?

2. I need to do filter using WHERE clause in some metrics columns, such as
count rows that having value of one field over 100. It seems that I have to
add new columns such as A_FIELD_OVER_100 to achieve this. But what if the
*100* is a variable? User of our system need to filter out result based on
metrics value, should I add metrics into dimensions? Is this requirement an
uncommon case?

3. It seems that querying all-null columns issue is fixed in this issue
<https://issues.apache.org/jira/browse/KYLIN-1527> (Kylin 1527). But I
still got NullPointerError from RawMesureType.valueOf method. I just want
to make sure that Kylin support columns with all null values, right?

Any suggestion is welcome. Thank you.
-- 
TONG, Da / 佟达

Reply via email to