Optimising MySql Query with LEFT JOINS - mysql

Question

I am trying to get a list of customer who haven't ordered for 6months or more. I have 4 tables which I have used in the query
accounts (account_id)
stores (store_id, account_id)
customers (store_id, customer_id)
orders (order_id, customer_id, store_id)
The customer and orders table are very big, 3M and 26M rows respectively, so using left joins in my query make the query time extremely long. I believe I have index my tables correctly
here is my query i have used
SELECT cus.customer_id, MAX(o.order_date), cus.store_id, s.account_id, store_name
FROM customers cus
LEFT JOIN stores s ON s.store_id=cus.store_id
LEFT JOIN orders o ON o.customer_id=cus.customer_id AND o.store_id=cus.store_id
WHERE account_id=26 AND
(SELECT order_id
FROM orders o
WHERE o.customer_id=cus.customer_id
AND o.store_id=cus.store_id
AND o.order_date < CURRENT_DATE() - INTERVAL 6 MONTH
ORDER BY order_id DESC LIMIT 0,1) IS NOT NULL
GROUP BY cus.customer_id, cus.client_id;
I need to get the last order date and this is the reason why I have joined the orders table, however since the customers can have multiple orders it is returning multiple rows of the customer and that is why I have used the group by clause.
If anyone can assist me with my query.

Answer 1

Start with this:
SELECT customer_id, MAX(order_date) AS last_order_date
FROM orders
GROUP BY customer_id
HAVING last_order_date < NOW() - INTERVAL 6 MONTH;
Assuming that gives you the relevant customer_ids, then move on to
SELECT ...
FROM ( that-select-as-a-subquery ) AS old
JOIN other-tables-as-needed ON USING(customer_id)
If necessary, JOIN back to orders to get more info. Do not try to get other columns in that subquery. (That's a "groupwise max" problem.)

Answer 2

Your strategy of using an ordered and limited subquery on your orders table is probably responsible for your poor performance.
This subquery will generate a virtual table showing the date of the most recent order for each distinct customer. (I guess a distinct customer is distinguished by the pair customer_id, store_id).
SELECT MAX(order_date) recent_order_date,
customer_id, store_id
FROM orders
GROUP BY customer_id, store_id
Then, you can use that subquery as if it were a table in your query.
SELECT cus.customer_id, summary.recent_order_date,
cus.store_id, s.account_id, store_name
FROM customers cus
JOIN stores s ON s.store_id=cus.store_id
JOIN (
SELECT MAX(order_date) recent_order_date,
customer_id, store_id
FROM orders
GROUP BY customer_id, store_id
) summary ON summary.customer_id = cus.customer_id
AND summary.store_id = s.store_id
WHERE summary.recent_order_date < CURRENT_DATE - INTERVAL 6 MONTH
AND store.account_id = 26
This approach moves the GROUP BY to an inner query, and eliminates the wasteful ORDER BY ... LIMIT query pattern. The inner query doesn't have to be remade for every row in the outer query.
I don't understand why you used LEFT JOIN operations in your query.
And, by the way, most people, when they're new to SQL, don't have great intuition about which indexes are useful and which aren't. So, when asking for help, it's always good to show your indexes. In the meantime, read this:
http://use-the-index-luke.com/