Sum of 2 aggregated rows

Question

Hi,
I have the following data as a result of my SQL. I have used GROUP BY CUBE
Category Status Count
A S 5
A F 10
A R 5
A Tot 20
B S 15
B F 10
B R 5
B Tot 30
I need to get the new row ST which is Sum of S and F status records for each Category. For Example.
Category Status Count
A S 5
A F 10
A        ST     15
A R 5
A Tot 20
B S 15
B F 10
B         ST     25
B R 5
B Tot 30
Can someone please help me solve this issue.
Thank you very much.
Edited by: 959026 on Sep 13, 2012 12:09 PM

Answer 1

Hi,
Interesting problem!
959026 wrote:
Hi,
I have the following data as a result of my SQL. I have used GROUP BY CUBEAre you sure? If you used GROUP BY CUBE, wouldn't there be super-aggregate rows where category was NULL? If you're using CUBE, then I suspect you're doing some additional manipulation after you get the CUBE results. Without seeing exactly what, I can't tell how to incorporate it into the solution.
Category Status Count
A S 5
A F 10
A R 5
A Tot 20
B S 15
B F 10
B R 5
B Tot 30
I need to get the new row ST which is Sum of S and F status records for each Category. For Example.
Category Status Count
A S 5
A F 10
A        ST     15
A R 5
A Tot 20
B S 15
B F 10
B         ST     25
B R 5
B Tot 30
Can someone please help me solve this issue.It would have been very helpful if you'd posted CREATE TABLE and INSERT statements for your sample data, and your existing GROUP BY query.
Since I don't have your table, I'll use scott.emp to illustrate.
Here's a query that's similar to yours. The 4 values of job in scott.emp are kind of like the 3 values of status in your table.
SELECT       deptno
,       job
,       COUNT (*)     AS cnt
FROM       scott.emp
WHERE       job     IN ( 'ANALYST', 'CLERK', 'MANAGER', 'PRESIDENT')
GROUP BY  CUBE (deptno, job)
ORDER BY        deptno, job
;Output:
`   DEPTNO JOB              CNT
        10 CLERK              1
        10 MANAGER            1
        10 PRESIDENT          1
        10                    3
        20 ANALYST            2
        20 CLERK              2
        20 MANAGER            1
        20                    5
        30 CLERK              1
        30 MANAGER            1
        30                    2
           ANALYST            2
           CLERK              4
           MANAGER            3
           PRESIDENT          1
                             10Say we want to get a sub-total of just two jobs, MANAGER and PRESIDENT, within each department, like this:
`   DEPTNO JOB                          CNT
        10 MANAGER                        1
        10 PRESIDENT                      1
        10   Management Subtotal          2
        10 CLERK                          1
        10     Total                      3
        20 MANAGER                        1
        20   Management Subtotal          1
        20 ANALYST                        2
        20 CLERK                          2
        20     Total                      5
        30 MANAGER                        1
        30   Management Subtotal          1
        30 CLERK                          1
        30     Total                      2One way to do that is to derive a new column, job category, which can have either of 2 values: ' Management Subtotal' or 'Others'. We can then include that new column in the GROUP BY clause. That would give us a sub-total for all the rows in the Management group, but it would also produce a sub-total row for the Others group, which we don't want to see. The query below uses a HAVING clause to eliminate the super-aggregate rows for 'Others'.
WITH     got_job_grp      AS
     SELECT     deptno, job
     ,     CASE
              WHEN  job  IN ('MANAGER', 'PRESIDENT')
              THEN  '  Management Subtotal'
              ELSE  'Others'
          END     AS job_grp
     FROM       scott.emp
     WHERE     job     IN ( 'ANALYST', 'CLERK', 'MANAGER', 'PRESIDENT')
SELECT       deptno
,       COALESCE (job, job_grp, '    Total')     AS job
,       COUNT (*)                            AS cnt
FROM       got_job_grp   g
GROUP BY  GROUPING SETS ( (deptno, job_grp, job)
                          , (deptno, job_grp     )
               , (deptno)
HAVING       COALESCE (job, job_grp, '?')     != 'Others'
ORDER BY  deptno
,            job_grp     -- add DESC NULLS LAST, if neessary
,       g.job
;I used GROUPING SETS instead of CUBE, because we don't want all possible rollups; we only want a few, easily defined rollups.
Notice that the management group is listed first in the output for each department. ORDER BY ... job_grp does that, but only because the label I used for the management group (' Management Subtotal') happens to come before the label for the other group ('Others'). If you want to use other labels, which are more meaningful to you, then you may need to ORDER BY job_grp in DESCending order, in which case you would also need to specify NULLS LAST, to get the output in the right order.