CODEWITHSUNDEEP
sqloracle-database
Xiezi
Xiezi

Reputation: 2989

SQL nested sums with multiple joins and group by

I'm not very familiar with SQLs. I'm using oracle. I met a question with over summing fields.

Here are the example tables:

A:
    A_ID
    A_NAME

B:
    B_ID
    A_ID
    B_NAME
    B_QTY
C:
    C_ID
    B_ID
    C_QTY

So the data structure is like A -> *B -> *C

I need to get the total quantities of Bs and Cs grouped by B_NAME and A_ID. For example:

A:
    A_ID    A_NAME
    1       A1
B:
    B_ID    A_ID    B_NAME  B_QTY
    1       1       B1  20
    2       1       B1  5
    3       1       B1  5
    4       1       B2  5
C:
    C_ID    B_ID    C_QTY
    1       1       3
    2       1       4
    4       2       2
    5       2       1
    6       3       1
    7       4       1

The expected result is:

A_ID    A_NAME  B_NAME  B_QTY   C_QTY
1       A1      B1      30      11
1       A1      B2      5       1

The 30 of B_QTY in the 1st line is result of 20 + 5 + 5.

The 11 of C_QTY in the 1st line is result of 3 + 4 + 2 + 1 + 1.

Here is my sql:

select a.A_ID,
    a.A_NAME,
    b.B_NAME
    sum(b.B_QTY),
    sum(c.C_QTY)
from A a left outer join B b on b.A_ID = a.A_ID
left outer join C c on c.B_ID = b.B_ID
group by a.A_ID
order by a.A_ID, b.B_NAME
    where a.XXXX = XXXXX;

So the problem is:

Since the B mapps to multiple Cs, the B_QTY will be summed multiple times. I'm not very familiar with SQL so I don't know if there is any simple way to distict the summing based on some fields (which is B_ID in my example). Thank you!

Upvotes: 4

Views: 2065

Answers (3)

Jess
Jess

Reputation: 25147

I created an SQL fiddle for this problem. The trick is that the B_QTY was appearing in your results more than once. Summing on it was giving an artificially high value. So instead, run a sub select to use the B_NAME only once! Great question! :^D

A.B.Cade's answer is cool, but this solution will work for many databases. I've used this technique before with SQL Server, Oracle, and Informix.

Data/Schema:

create table a (A_ID int, A_NAME char(10));
create table b (B_ID int, A_ID int, B_NAME char(10), B_QTY int);
create table c (C_ID int, B_ID int, C_QTY int);
-- One dude
insert into a values (1,'Xiezi');
-- 2 orders? of 4 and 3
insert into b values (1,1,'B1',20);
insert into b values (2,1,'B1',5);
insert into b values (3,1,'B1',5);
insert into b values (4,1,'B2',5);
-- 2 order with 2 lines each.
insert into c values (1,1,3);
insert into c values (2,1,4);
insert into c values (4,2,2);
insert into c values (5,2,1);
insert into c values (6,3,1);
insert into c values (7,4,1);

SQL (The answer):

select a.A_ID,
    a.A_NAME,
    b.B_NAME,
    (select sum(b2.B_QTY) from b b2 where b2.B_NAME = b.B_NAME)
    as sum_b_qty,
    sum(c.C_QTY)
from a left outer join b on b.A_ID = a.A_ID
left outer join c on c.B_ID = b.B_ID
group by a.A_ID,
    a.A_NAME,
    b.B_NAME
order by a.A_ID
;

Output:

A_ID A_NAME B_NAME  SUM_B_QTY SUM(C.C_QTY)
1    Xiezi  B1      30        11
1    Xiezi  B2      5         1

Upvotes: 1

Gentlezerg
Gentlezerg

Reputation: 286

You can also do like this:

SELECT DISTINCT A_ID,A_NAME,B_NAME,B_SUM,SUM(C_QTY) OVER(PARTITION BY A_NAME,B_NAME) C_SUM
FROM (
SELECT A.A_ID,A_NAME,B_NAME,B_ID,SUM(B_QTY) OVER(PARTITION BY A_NAME,B_NAME) B_SUM
FROM A JOIN B
ON A.A_ID=B.A_ID) T1 
JOIN C
ON T1.B_ID=C.B_ID

enter image description here

Upvotes: 1

A.B.Cade
A.B.Cade

Reputation: 16915

This can also be done like this:

WITH b2 AS
(SELECT b.*, sum(b.b_qty) over (partition BY b.a_id, b.b_name) b_qty_s
             FROM b)
SELECT a.a_id, a.a_name, b2.b_name, b2.b_qty_s, sum(c.c_qty) c_qty_s
FROM a JOIN b2 ON a.a_id = b2.a_id
JOIN c ON b2.b_id = c.b_id
GROUP BY a.a_id,a.a_name, b2.b_name, b2.b_qty_s

Here is a sqlfiddle demo

Upvotes: 2

Related Questions