How can I produce rows that sum subgroups of rows, without grouping? - sql

Question

Let's say I have a table like this:
ID Income Expenses
-------------------------
1 $1000 $200
1 $2000 $400
2 $500 $200
2 $100 $60
I'd like to add one row per ID that sums the ID's Income and Expenses. For example:
ID Income Expenses
-------------------------
1-SUM $3000 $600
1 $1000 $200
1 $2000 $400
2-SUM $600 $260
2 $500 $200
2 $100 $60
Is this possible with SQL? I read about OVER and PARTITION BY but that'll add a new column, not a new row. Might a UNION be the best way to go here?

Answer 1

Try this
create table yourtable(id int,income int,expenses int)
insert into yourtable values (1,100,200),(1,150,250),(2,200,300),(2,250,350)
with CTE
as
(select ID,sum(income) as income,sum(expenses) as expenses,1 as isSum
from yourtable
group by id
)
select * from (
select *,0 as isSum from yourtable
union all
select * from CTE ) as t
order by id,issum
output

Answer 2

Using GROUP BY:
SQLFiddleDemo
CREATE TABLE tab(id INT , income INT , expenses INT);
INSERT INTO tab(id, income, expenses)
VALUES (1, 1000, 200),(1, 2000, 400),(2, 500, 200),(2, 100, 60)
SELECT id,
[income] = SUM(income),
[expenses] = SUM(expenses),
[grouping_level] = GROUPING_ID(id, income, expenses) /* 0 - actual data , 3 - group based on column income, expenses*/
FROM tab
GROUP BY GROUPING SETS ((id), (id,income, expenses))
Why grouping is useful:
Using GROUPING SETS gives great flexibility for instance when you want total sum use: GROUP BY GROUPING SETS ((id), (id,income, expenses), ())
GROUPING SETS can be combined with ROLLUP/CUBE.
GROUPING_ID indicates grouping level.
Number works like binary. GROUPING(a, b, c) if b i c is used for grouping you get 011B => 3