Speed up slow 3 table MYSQL query - php

Question

I'm querying 3 tables in an eCommerce site.
orders Table:
id
order_number
name
etc...
order_lines table:
id
order_number
sku
quantity
etc.
products Table
id
sku
title
ship_by (INT)
etc.
order_number links orders table to order_lines table. SKU links order_lines table to products table.
Notice the ship_by column in the products table, this denotes which supplier will ship the item. The query needs to pull orders for a particular supplier. Orders may contain items sold by different suppliers.
This is the query I managed to cobble together:
SELECT
orders.`order_number` as `orderId`,
orders.`shipping` as `fPostageCost`,
FROM_UNIXTIME(orders.`time`) as `dReceievedDate`,
(
CASE orders.`dob`
WHEN 'COURIER 3 DAY' THEN 'UK_SellersStandardRate'
WHEN 'COURIER 24 HOUR' THEN 'UK_OtherCourier24'
ELSE orders.`dob`
END
)
...(plus a number more)
FROM orders
INNER JOIN order_lines
ON orders.order_number = order_lines.order_number
WHERE
(
(SELECT COUNT(order_lines.sku)
FROM order_lines, products
WHERE order_lines.sku = products.sku
AND products.ship_by = 1
AND order_lines.order_number = orders.order_number) > 0
)
AND ( orders.`printed` = 'N' OR orders.`printed` IS NULL )
AND orders.status = 'Awaiting Despatch'
AND ( orders.payment_status = 'Success' OR orders.payment_status = 'Paypal Paid' OR orders.payment_status = 'Manual Payment' )
GROUP BY orders.`order_number`
ORDER BY orders.order_number ASC
It is taking about 7 seconds to execute the query.
If I remove the line 'AND order_lines.order_number = orders.order_number' from the second SELECT query it is executed almost instantly, but without this line it doesn't work as I need it to.
Aside from adding the 'ship_by' column into the order_lines table (which I'd rather not do since I'd have to change a lot of php code), is there any way of modifying this query to speed it up?
The query is being run from outside of PHP so it has to be a pure mysql query.
Here is the EXPLAIN on the query:
id select_type table type possible_keys key key_len ref rows Extra
1 PRIMARY order_lines ALL NULL NULL NULL NULL 9627 Using temporary; Using filesort
1 PRIMARY orders eq_ref order_idx order_idx 17 .order_lines.order_number 1 Using where
2 DEPENDENT SUBQUERY order_lines ALL NULL NULL NULL NULL 9627 Using where
2 DEPENDENT SUBQUERY products ref sku_2,sku,sku_3 sku_2 63 order_lines.prod_code 11 Using where
Thanks

Answer 1

Your order_lines table needs an index on order_number at the very least. We would need to see the schema to best choose one. Perhaps a composite index would pick up speed in other areas.
But in choosing index changes, it must be carefully weighed against other queries in your system, and the impact on insert and update speeds.
The goal shouldn't be to make 10% of your app fast, at the expense of the 90.
To show very useful information, publish show create table tableName for relevant tablenames. (versus describing it free-hand, like my table has this and that). We need to see the schema for orders, order_lines, products.
The manual page on Create Index

Answer 2

Idk how large the table is on your inline view but if it is large then
it may help to use a union instead of having to query the view for every record.
Something like.
SELECT L1*
FROM
(SELECT
orders.`order_number` as `orderId`,
0 AS SKU_CNT,
orders.`shipping` as `fPostageCost`,
FROM_UNIXTIME(orders.`time`) as `dReceievedDate`,
(
CASE orders.`dob`
WHEN 'COURIER 3 DAY' THEN 'UK_SellersStandardRate'
WHEN 'COURIER 24 HOUR' THEN 'UK_OtherCourier24'
ELSE orders.`dob`
END
)
...(plus a number more)
FROM orders
INNER JOIN order_lines
ON orders.order_number = order_lines.order_number
WHERE ( orders.`printed` = 'N' OR orders.`printed` IS NULL )
AND orders.status = 'Awaiting Despatch'
AND ( orders.payment_status = 'Success' OR orders.payment_status = 'Paypal Paid' OR orders.payment_status = 'Manual Payment' )
UNION ALL
Select
0 AS ORDERID,
COUNT(OL.SKU) AS SKU_CNT,
0 as `fPostageCost`,
NULL as `dReceievedDate`,
...
FROM ORDER_LINES OL,
PRODUCTS PR
WHERE order_lines.sku = products.sku
AND products.ship_by = 1
AND order_lines.order_number = orders.order_number
GROUP BY ORDERID, ORDER_LINES_CNT, fPostageCost`, dReceievedDate`,
)L1
WHERE L1.SKU_CNT > 0
GROUP BY L1.`order_number`
ORDER BY L1.order_number ASC

Answer 3

You can replace the subquery with exists:
WHERE EXISTS (SELECT 1
FROM order_lines ol2 JOIN
products p2
ON ol2.sku = p.sku AND
p2.ship_by = 1
ol.order_number = o.order_number
) AND
. . .
But, your snippet of the query is not using order_lines in the outer query. I suspect you can just get rid of it with the aggregation:
FROM orders o
WHERE EXISTS (SELECT 1
FROM order_lines ol2 JOIN
products p2
ON ol2.sku = p2.sku AND
p2.ship_by = 1
ol.order_number = o.order_number
) AND
. . .
This will at least return all the orders information -- and the lack of aggregation may simplify other parts of the query. If some versions of the query are running fast, then indexes are probably set up reasonable.