Count rows with non-NULL in two columns

Question

I am trying to compare two sets in SQL:

create table my_counts
as select count(*) as total, 
          count(c.cookie) as first,
          count(l.cookie) as second, 
          count(l.cookie and c.cookie) as common
from (select distinct s.cookie from ...) c
full outer join
     (select distinct s.cookie from ...) l
on c.cookie = l.cookie;
select * from my_counts;

However, the above fails with this error:

FAILED: ClassCastException org.apache.hadoop.hive.serde2.objectinspector.primitive.WritableStringObjectInspector cannot be cast to org.apache.hadoop.hive.serde2.objectinspector.primitive.BooleanObjectInspector

Of course (assuming the SQL statement above does what I think it should do), I should have total + common = first + second, so the common column is not strictly necessary, but I still wonder if there is a way to count the rows where both columns are non-NULL.

Answer 1

AND - is a boolean operator. It seems like cookie is a string, not boolean. try to replace count(l.cookie and c.cookie) with this: count(case when l.cookie is not null and c.cookie is not null then 1 else NULL end) as common