Oracle SQL grouping elements - sql

Question

I know this question is kind of trivial but I have difficulties writing the SQL for the example shown below.
As shown in the first table, is the result that i generated and they are ok for analytics,
REGION SUBREGION SUM
------ --------- ------
CORP CORP1 5
CORP CORP2 10
CORP CORP3 5
SB SB1 10
SB SB2 10
MID null 10
LARGE null 20
but for summary report i need to display result as shown in the second table. Any clues?
REGION SUM
------ ----
CORP 20
CORP1 5
CORP2 10
CORP3 5
SB 20
SB1 10
SB2 10
MID 10
LARGE 20

Answer 1

Simply change your existing GROUP BY to GROUPING SET:
SELECT
Coalesce(subregion, region) AS region,
Sum(column)
FROM mytable
GROUP BY GROUPING SETS(region, subregion)
HAVING Coalesce(subregion, region) IS NOT NULL

Answer 2

sounds like you need to aggregate the same table by different fields, but get both back as if it were one field. The solution that comes to mind is a UNION
select sum() as sum, REGION as sf from table group by REGION
union ALL
select sum() as sum, SUB_REGION as sf from table group by SUB_REGION;
Hope that helps
based on Dan's question below, I add, if you don't want to agg the vals, just take out the sum and group by and do the straight union
select REGION as sf from table
union ALL
select SUB_REGION as sf from table;
EDIT:
One more thought, perhaps when you do the query in the first place, you may want to look into the concept of ROLLUPs as an additional clause on your group bu and agg function, might help solve this in one shot.

Answer 3

Do the same group by query you're doing, but try using group by ROLLUP:
Something like (untested):
select region, subregion, sum(some_column) as sum
from some_table
group by rollup(region, subregion)
order by region, subregion;

Answer 4

Try using OLAP functions rollup and grouping like this:
select
nvl(subregion, region) region, sum("sum")
from t
group by region, rollup(subregion)
having case when count(*) = 1 then 0 else 1 end = grouping(subregion);
In the above,
having case when count(*) = 1 then 0 else 1 end = grouping(subregion);
The above excludes the rolluped row if there is only one row for that region so that there are no duplicates.
Also, avoid using reserved keywords such as sum or count in your identifiers.
Demo:
SQL> with t(REGION ,SUBREGION ,s) as (
2 select 'CORP' , 'CORP1' , 5 from dual union all
3 select 'CORP' , 'CORP2' , 10 from dual union all
4 select 'CORP' , 'CORP3' , 5 from dual union all
5 select 'SB' ,'SB1' ,10 from dual union all
6 select 'SB' ,'SB2' ,10 from dual union all
7 select 'MID' , null , 10 from dual union all
8 select 'LARGE' , null , 20 from dual
9 )
10 select
11 nvl(subregion, region) region, sum(s)
12 from t
13 group by region, rollup(subregion)
14 having case when count(*) = 1 then 0 else 1 end = grouping(subregion);
REGIO SUM(S)
----- ----------
SB1 10
SB2 10
SB 20
MID 10
CORP1 5
CORP2 10
CORP3 5
CORP 20
LARGE 20
9 rows selected.
SQL>