Concatenated groupings question

Question

Hello,
Regarding to concatenated groupings, if i have: GROUP BY grouping sets(a,b), grouping sets(c,d), the result is a cross product of groupings from each grouping set. So, we have the following groupings: (a,c), (a,d), (b,c), (b,d).
But if we have GROUP BY department_id, ROLLUP(job_id), CUBE(manager_id), which are the groupings in this case?
Also, on GROUP BY department_id, ROLLUP(job_id, <other column>), CUBE(manager_id) ?
The last 2 examples are a bit confusing for me.
Thanks!

Answer 1

Hi,
Roger22 wrote:
Ok, so:
SQL> select department_id, job_id, manager_id, sum(salary), grouping(department_id)
2  from employees e
3  group by rollup(department_id, (job_id, manager_id))
4  ;
DEPARTMENT_ID JOB_ID     MANAGER_ID SUM(SALARY) GROUPING(DEPARTMENT_ID)
SA_REP            149        7000                       0
7000                       0
10 AD_ASST           101        4400                       0
10                              4400                       0
23 rows selected
What table are you using? When I run that query with my hr.employees table, I get 46 rows, not 23. If your employees table is a subset of hr.employees, then post the code you used to cfreate it.
I don't understant the first two rows(those with salary 7000). There is no grouping of (job_id, manager_id) in this example , because it's treated as a unit,Sorry, I don't understand what you mean here.
There is a grouping of (job_id, manager_id) in this example. The GROUP BY clause contains "(job_id, manager_id)", so there is a separate row for each distinct combination of job_id and manager_id.
it's a composite column and ROLLUP does not roll up in each direction possible (so, with all combinations in GROUP BY clause), but rolls up from the highest level to the outermost level (from right to left). So, the groupings are:
department_id, job_id, manager_id
department_id
and () -- which is the grand total calculated
Please clarify in this case.. hope i was more clearThat's right.
What exactly don't you understand about the beginning of the output?
I really think it would help if you displayed GROUPING for all of the GROUP BY columns, not just one of them, like this:
SELECT        department_id
,       job_id
,       manager_id
,       SUM (salary)               AS sum_salary
,       GROUPING (department_id)     AS g_d
,       GROUPING (job_id)          AS g_j
,       GROUPING (manager_id)          As g_m
FROM       hr.employees     e
GROUP BY  ROLLUP ( department_id
           , (job_id, manager_id)
;Output (abridged):
DEPARTMENT_ID JOB_ID     MANAGER_ID SUM_SALARY G_D G_J G_M
              SA_REP            149       7000   0   0   0
                                          7000   0   1   1
           10 AD_ASST           101       4400   0   0   0
           10                             4400   0   1   1
                                        691400   1   1   1As you said, ROLLUP is treating job_id and manager_id as a uint.
On each row where an individual job_id is used, an individual manager_id is used also.
Whenever all l job_ids are combined, all manager_ids are combined also.
GROUPING (job_id) = GROUPING (mnaager_id) on every row.
It's unclear, but you seem to have a question about how the 2nd row above
DEPARTMENT_ID JOB_ID     MANAGER_ID SUM_SALARY G_D G_J G_M
                                          7000   0   1   1is being produced. Look at the GROUPING output. GROUPING (department_id) = 1, meaning the row represents only one department_id (GROUP BY treats NULL as a value), but GROUPING (job_id) = 1 and GROUPING (manager_id) = 1, meaning that this row is a super-aggregate, combining rows regardless of their job_id or manager_id. The 2nd row functions exactly the same as any other row with the same values (0, 1, 1) in the GROUPING columns. For eample, the 4th row above:
DEPARTMENT_ID JOB_ID     MANAGER_ID SUM_SALARY G_D G_J G_M
           10                             4400   0   1   1Do you uderstand how this row is formed? It is a summary of all the rows in the table where department_id=10 (As it happens, there is only one row with department_id=10 in the table, but GROUP BY produces a summary anyway.)
In exactly the same way, the 2nd row of output is a summary of all the rows in the table where department_id is NULL. (As it happens, there is only one row with department_id is NULL in the table, but GROUP BY produces a summary anyway.)