Query related to the group by rollup,cube

Question

Hello experts,
I am not getting how the below query gets excuted ,
{code}
SELECT department_id,   job_id,   manager_id,   SUM(salary) FROM employess GROUP BY department_id,   ROLLUP(job_id),   CUBE(manager_id)
{code}
can anybody simplified this plz ?
Thanks in advance

Answer 1

Hi,
SShubhangi wrote:
Hello experts,
I am not getting how the below query gets excuted ,
{code}
SELECT department_id,   job_id,   manager_id,   SUM(salary) FROM employess GROUP BY department_id,   ROLLUP(job_id),   CUBE(manager_id)
{code}
can anybody simplified this plz ?
Thanks in advance
Here's how it works.
Since the GROUP BY clause includes department_id (not modified by ROLLUP or CUBE), every row of the result set will be limited to a specific department_id.
Since the GROUP BY clause includes ROLLUP (job_id). some rows of the result set will represent a specific job_id, and some rows will be Super-Aggregate rows, representing all job_ids at the same time.
Since the GROUP BY clause includes CUBE (manager_id). some rows of the result set will represent a specific manage_id, and some rows will be super-aggregates, representing all manager_ids at the same time.  (When there is only 1 expression inside the parentheses, CUBE means the same thing as ROLLUP).
Here's how it can be simplified:
GROUP BY  department_id
,         CUBE (job_id, manager_id)
Here's why:
Since ROLLUP (x) means the same thing as CUBE (x) - when x is a single expression - then what you posted is equivalent to
GROUP BY  department_id
,         CUBE (job_id)
,         CUBE (manager_id)
and
GROUP BY  ...
          CUBE (x)
,         CUBE (y)
is just a longer way of saying
GROUP BY  ...
          CUBE (x, y)
To understand this better, do some experiments yourself.  Try different combinations of ROLLUP and CUBE, and see what results they produce.
DO NOT use the hr.employees table for your experiments; it has far too many groups for anyone to understand. Also, department_id and manager_id can be NULL, so it's hard to tell super-aggregate rows from normal aggregate rows.  Even scott.emp is more complicated than necessary.  I suggest you make your own table, like this:
CREATE TABLE  simp_emp  AS
SELECT  ename
,       deptno
,       job
,       CASE
            WHEN  job IN ('MANAGER', 'PRESIDENT')
            THEN  'NO'
            ELSE  'YES'
        END                            AS unionized
,       EXTRACT (YEAR FROM hiredate)   AS hireyear
,       sal
FROM    scott.emp
In this table, there are only 2 possible values for unionized, 3 values for deptno, and 4 values for hireyear, and none of those colums are ever NULL.