MySQL sum with group by gives wrong results - mysql

Question

Here's the SQLFiddle with schema and data.
I'm trying to sum 2 columns, one at parent level and the other at child level.
The current query I'm using gives me the right sum amount on child level, but doubles up the amount on parent level, due to another 1-many relationship involved on the child level.
Ugh... that's a terrible explanation - here's the English version:
Joe the salesman is involved in 2 sales.
For the 1st sale, he get's 2 sets of commissions, based on 2 different commission types. I'm trying to show Joe's total sale value, alongside the total value of his applicable splits. The split value total is fine, but sale value get's doubled up because I'm obviously, grouping/joining incorrectly (see the last example below).
This is fine:
select sp.person_name, pr.description,
sum(spl.split) as SplitValue
from sale s, product pr, sales_person sp, sales_split spl
where s.product_id = pr.id
and s.id = spl.sale_id
and sp.id = spl.sales_person_id
group by sp.id;
person_name | description | SplitValue
----------- ----------- | ----------
Joe | Widget 1 | 50
Sam | Widget 1 | 10
This is also yields the correct split and sale values, but now 3 rows are displayed for Joe (i.e 2nd row is a duplicate of the 1st one) - I only want to display Joe's "Widget 1" sale once, so not correct:
select sp.person_name, pr.description,
sum(s.sale_value) as SaleValue, sum(spl.split) as SplitValue
from sale s, product pr, sales_person sp, sales_split spl, sales_split_agreement ssa
where s.id = spl.sale_id
and s.product_id = pr.id
and sp.id = spl.sales_person_id
and sp.id = ssa.sales_person_id
and spl.sales_person_id = ssa.sales_person_id
and ssa.id = spl.sales_split_agreement_id
group by sp.id, spl.id;
person_name | description | SplitValue | SaleValue
----------- ----------- ---------- ---------
Joe | Widget 1 | 10 | 20
Joe | Widget 1 | 10 | 20
Joe | Widget 2 | 30 | 30
Sam | Widget 1 | 10 | 20
Now the duplicated row is gone, but Joe's SaleValue is incorrect - it should be 50, not 70:
select sp.person_name, pr.description,
sum(spl.split) as SplitValue, sum(s.sale_value) as SaleValue
from sale s, product pr, sales_person sp, sales_split spl, sales_split_agreement ssa
where s.id = spl.sale_id
and s.product_id = pr.id
and sp.id = spl.sales_person_id
and sp.id = ssa.sales_person_id
and spl.sales_person_id = ssa.sales_person_id
and ssa.id = spl.sales_split_agreement_id
group by sp.id;
person_name | description | SplitValue | SaleValue
----------- ----------- --------- ----------
Joe | Widget 1 | 50 | 70
Sam | Widget 1 | 10 | 20
I.e. I'm after the query that will yield this result (i.e. Joe's correct SaleValue of 50):
person_name | description | SplitValue | SaleValue
----------- ----------- --------- ----------
Joe | Widget 1 | 50 | 50
Sam | Widget 1 | 10 | 20
Any help will be greatly appreciated!
UPDATE 1:
For clarity - here's the schema and test data from the fiddle:
CREATE TABLE product
(`id` int, `description` varchar(12))
;
INSERT INTO product
(`id`, `description`)
VALUES
(1, 'Widget 1'),
(2, 'Widget 2')
;
CREATE TABLE sales_person
(`id` int, `person_name` varchar(7))
;
INSERT INTO sales_person
(`id`, `person_name`)
VALUES
(1, 'Joe'),
(2, 'Sam')
;
CREATE TABLE sale
(`id` int, `product_id` int, `sale_value` int)
;
INSERT INTO sale
(`id`, `product_id`, `sale_value`)
VALUES
(1, 1, 20.00),
(2, 2, 30.00)
;
CREATE TABLE split_type
(`id` int, `description` varchar(6))
;
INSERT INTO split_type
(`id`, `description`)
VALUES
(1, 'Type 1'),
(2, 'Type 2')
;
CREATE TABLE sales_split_agreement
(`id` int, `sales_person_id` int, `split_type_id` int, `percentage` int)
;
INSERT INTO sales_split_agreement
(`id`, `sales_person_id`, `split_type_id`, `percentage`)
VALUES
(1, 1, 1, 50),
(2, 1, 2, 50),
(3, 2, 1, 50),
(4, 1, 1, 100)
;
CREATE TABLE sales_split
(`id` int, `sale_id` int, `sales_split_agreement_id` int, `sales_person_id` int, `split` int )
;
INSERT INTO sales_split
(`id`, `sale_id`, `sales_split_agreement_id`, `sales_person_id`, `split`)
VALUES
(1, 1, 1, 1, 10),
(2, 1, 2, 1, 10),
(3, 1, 3, 2, 10),
(4, 2, 4, 1, 30)
;

Answer 1

I think you were on to the right track, but I decided to restart and approach from the beginning. Getting the SplitValue for each person does not require all those tables. In fact, all you need are sales_split and sales_person, like this:
SELECT sp.person_name, SUM(ss.split) AS SplitValue
FROM sales_person sp
JOIN sales_split ss ON sp.id = ss.sales_person_id
GROUP BY sp.id;
Similarly, you can get the total sale value for each person with a join between sale, sales_split, and sales_person:
SELECT sp.person_name, SUM(s.sale_value) AS SaleValue
FROM sale s
JOIN sales_split ss ON ss.sale_id = s.id
JOIN sales_person sp ON sp.id = ss.sales_person_id
GROUP BY sp.id;
At this point, I realize you have an error in your expected results (for this data set). Joe does in fact have a sale value of 70, because sale id 1 (value 20), 2 (value 20), and 4 (value 30) add up to 70. However, I still think this query will help you out more than the one you have.
At this point, you can get the values for each sales_person_id by joining those two subqueries to the sales_person table. I took out the join to sales_person in the subqueries, as it became irrelevant now. It even makes the subqueries a little cleaner:
SELECT sp.person_name, COALESCE(t1.SplitValue, 0) AS SplitValue, COALESCE(t2.SaleValue, 0) AS SaleValue
FROM sales_person sp
LEFT JOIN(
SELECT ss.sales_person_id, SUM(ss.split) AS SplitValue
FROM sales_split ss
GROUP BY ss.sales_person_id) t1 ON t1.sales_person_id = sp.id
LEFT JOIN(
SELECT ss.sales_person_id, SUM(s.sale_value) AS SaleValue
FROM sale s
JOIN sales_split ss ON ss.sale_id = s.id
GROUP BY ss.sales_person_id) t2 ON t2.sales_person_id = sp.id;
Here is an SQL Fiddle example.
EDIT: I understand now why Joe's actual sale price is 50, because he split twice on sale id 1. To work around this, I first got a list of distinct sales for each sales_person like this:
SELECT DISTINCT sale_id, sales_person_id
FROM sales_split;
This way, there is only one row for sales_person_id = 1 and sale_id = 1. Then, it was easy enough to join that to the sale table and get the proper sales value for each sales_person:
SELECT t.sales_person_id, SUM(s.sale_value) AS SaleValue
FROM(
SELECT DISTINCT sale_id, sales_person_id
FROM sales_split) t
JOIN sale s ON s.id = t.sale_id
GROUP BY t.sales_person_id;
The rest of my answer above still fits. I wrote one query to get SplitValue, and one query to get SaleValue, and I joined them together. So, all I have to do now is replace the subquery I just gave you, with the incorrect subquery from further up:
SELECT sp.person_name, COALESCE(t1.SplitValue, 0) AS SplitValue, COALESCE(t2.SaleValue, 0) AS SaleValue
FROM sales_person sp
LEFT JOIN(
SELECT ss.sales_person_id, SUM(ss.split) AS SplitValue
FROM sales_split ss
GROUP BY ss.sales_person_id) t1 ON t1.sales_person_id = sp.id
LEFT JOIN(
SELECT t.sales_person_id, SUM(s.sale_value) AS SaleValue
FROM(
SELECT DISTINCT sale_id, sales_person_id
FROM sales_split) t
JOIN sale s ON s.id = t.sale_id
GROUP BY t.sales_person_id) t2 ON t2.sales_person_id = sp.id;
Here is the updated SQL Fiddle.
You mentioned in the comments that you shortened your data for brevity, which is fine. I am leaving my joins as they are, and I trust that it gives you enough direction that you can adjust them accordingly to match your proper structure.