Uptimize select product details each customer purchased from sales tables - mysql

Question

I have 4 tables, Customers, Products, Sales & Sale_Items. I pull data from it using the query below.
SELECT (
SELECT c.name
FROM Customers c
WHERE s.customer_id=c.id
) customer
,(
Select group_concat(description)
FROM (
SELECT si.id
,si.sale_id
,concat("x", si.Qty, " ", p.name, " ",(si.total)) description
FROM Sale_Items si
LEFT JOIN Products p ON p.id = si.product_id
) p
where s.id = Sale_ID
GROUP BY Sale_ID
) detail,
s.total
FROM Sales s
The query produces the result, but it becomes slow with just 2000 records (takes 114 seconds to finish)
Customer Product Total
--------------------------------------
James x1 ItemA 10.00 75.00
x3 ItemB 15.00
x1 ItemC 20.00
Mark x2 ItemA 10.00 50.00
x2 ItemB 15.00
Bisi x1 ItemC 20.00 30.00
x2 ItemA 10.00
how can i make this faster?
An attempt has been made here
https://www.db-fiddle.com/f/pkL2HtsT659EXgRSevFSAm/4

Answer 1

If we want to stick with correlated subqueries, we can eliminate the inline view p.
That's going to get materialized for every row retrieved from Sales. The predicate in the WHERE clause in the outer query doesn't get "pushed" down into the view. So the materialized view (or "derived table" in the MySQL parlance) is going to be a full set, and from that, we're going to pick out just a few rows. And we're going to repeat that for each row from Sales.
Unwinding that derived table should give us some performance benefit. This would be reasonable approach for a small number of rows returned from Sales, with suitable indexes defined. That is, if we were limiting the number of rows examined by the outer query with a WHERE clause. With a large number of rows, those correlated subqueries are going to drag down performance.
SELECT ( SELECT c.name
FROM Customers c
WHERE c.id = s.customer_id
) AS customer
, ( SELECT GROUP_CONCAT(CONCAT('x',si.Qty,' ',p.name,' ',si.total) ORDER BY p.name SEPARATOR '\r\n')
FROM Sale_Items si
LEFT
JOIN Products p
ON p.id = si.product_id
WHERE si.sale_id = s.id
) AS detail
, s.total
FROM Sales s
WHERE ...
ORDER
BY ...
If the query is returning all rows from Sales and we are doing the whole bloomin' set, then I'd tend to avoid the correlated subqueries. (That's because the subqueries gets executed for each and every row returned by the outer query. Those subqueries are going to eat our lunch, in terms of performance, with a large number of rows returned.)
Assuming id is unique in customers, we're usually much better off with a join operation.
SELECT c.name AS customer
, d.detail
, s.total
FROM Sales s
LEFT
JOIN Customers c
ON c.id = s.customer_id
LEFT
JOIN ( SELECT si.sale_id
, GROUP_CONCAT(CONCAT('x',si.Qty,' ',p.name,' ',si.total) ORDER BY p.name SEPARATOR '\r\n') AS detail
FROM Sale_Items si
LEFT
JOIN Products p
ON p.id = si.product_id
GROUP
BY si.sale_id
) d
ON d.sale_id = s.id
ORDER
BY ...
The inline view d is going to be expensive with large sets; but at least we're only doing that query one time, materializing the results into a "derived table". Then the outer query can run, and retrieve rows from the derived table.