CODEWITHSUNDEEP
sqlteradata
C.C
C.C

Reputation: 179

Teradata SQL count on Top n subgroups

I am using TD v15. I have a table as below - each row is a single record, I want to perform Count in the following way:

In Question Column: I have 4 'A', 5 'B', 3 'C' and 2 'D'. Select top 2 from them, which are A & B. Group the rest Questions as 'OtherQ' - Put them in Result Question Column.

In Change Column, I have 2 'AA', 3 'AB', 2'AC', 2 'AD', 4 'AE' and 2 'AG', select top 2, which are AE & AB, group the rest Change as 'Other' - Put them in Result Change Column.

Then, count according...

Question    Result Change  
    A         Pass   AG          
    A         Pass   AE           
    A         Pass   AA           
    A         Pass   AB       
    B         Pass   AC      
    B         Pass   AG        
    B         Pass   AB         
    B         Pass   AE       
    B         Pass   AD
    B         Pass   AA
    C         Pass   AB
    C         Pass   AC
    C         Pass   AD
    D         Pass   AE
    D         Pass   AE
    A         Fail   Null
    A         Fail   Null
    C         Fail   Null
    E         Fail   Null
    B         Fail   Null

This is the desired result, it counts on top 2 questions (A&B) and OtherQ with Top 2 changes (AE&AB) and other Changes, also, it counts Pass&Fail for A&B and OtherQ.

The sum of Count is 20, this should match the 20 individual row in the table above.

 Question    Result Change    Count
    A         Pass   AE         1
    A         Pass   AB         1
    A         Pass   Other      2
    B         Pass   AE         1
    B         Pass   AB         1
    B         Pass   Other      4   
  OtherQ      Pass   AE         2
  OtherQ      Pass   AB         1
  OtherQ      Pass   Other      2
    A         Fail   Null       2
    B         Fail   Null       1
  OtherQ      Fail   Null       2

Could you please kindly help? It's very large data table, needs the code to be efficient. Many thank for your time and help in advance.

Upvotes: 0

Views: 522

Answers (1)

Gordon Linoff
Gordon Linoff

Reputation: 1270773

I would suggest using aggregations and subqueries:

select coalesce(tq.question, 'Other') as question
       (case when t.change is null then null
             else coalesce(tch.change, 'Other') 
        end) as change,
       count(*)
from t left join
     (select question, count(*) as cnt,
             row_number() over (order by count(*) desc) as seqnum
      from t
      group by question
     ) tq
     on tq.question = t.question and tq.seqnum <= 2 left join
     (select change, count(*) as cnt,
             row_number() over (order by count(*) desc) as seqnum
      from t
      group by change
     ) tch
     on tch.change = t.change and tch.seqnum <= 2
group by coalesce(tq.question, 'Other'),
         coalesce(tch.change, 'Other');

Upvotes: 2

Related Questions