I'm using kylin recently, and I don't know how to design cube for such case, e.g. there is a table: t1, and a field: f1, f1 could have many values separated by commas, e.g. A,B,C. If want to group by single value of f1, it seems you should have a separate table t2, have these columns: id(FK), f1, e.g.
id f1 1 A 1 B 1 C 2 A 2 B 3 C 3 D and t2 is not a normal lookup table, because one row of t1 could have many rows of t2(1:N) , So how to design cube could support group by f1 as single value and in the same time single value f1 could combine with other dimensions e.g. date, f1, date f2, f1?
