Question

I am trying to create a sequel query for emails of customers without accounts that have only placed orders on over 3 months. If i do the query

SELECT checkouts.email 
FROM orders, checkouts 
WHERE orders.id = checkouts.order_id AND orders.user_id is NULL 
AND orders.created_at < CURDATE() - INTERVAL 3 MONTH

I get all emails from orders without accounts that were placed within the last 3 months, BUT it doesn’t account for orders placed within the last 3 months without an account with the same emails from the sql query email list. Therefore, i want to use a MINUS operator to subtract all emails from recent orders. When i try this, i get an error:

SELECT checkouts.email 
FROM orders, checkouts 
WHERE orders.id = checkouts.order_id AND orders.user_id is NULL 
AND orders.created_at < CURDATE() - INTERVAL 3 MONTH 
MINUS 
SELECT checkouts.email 
FROM orders, checkouts 
WHERE orders.id = checkouts.order_id AND orders.user_id is NULL 
AND orders.created_at > CURDATE() - INTERVAL 3 MONTH

If i run this same query but with a UNION instead of a MINUS, it works, and essentially gives me all emails from orders without an account, regardless of the date.

Why does UNION work and not MINUS? How do i fix this query so i can get MINUS to work?

Answer 1

Try this construct instead (there may be better execution plans with other ways of saying it):

SELECT checkouts.email 
FROM orders, checkouts 
WHERE orders.id = checkouts.order_id AND orders.user_id is NULL 
AND orders.created_at < CURDATE() - INTERVAL 3 MONTH 
AND checkouts.email NOT IN (
    SELECT checkouts.email 
    FROM orders, checkouts 
    WHERE orders.id = checkouts.order_id AND orders.user_id is NULL 
    AND orders.created_at > CURDATE() - INTERVAL 3 MONTH
)

You may also want to add handling for checkouts.email IS NULL, since that WILL NOT be NOT IN (if that phrasing makes any sense), because NULL is never = or <> to anything.

SELECT checkouts.email 
FROM orders, checkouts 
WHERE orders.id = checkouts.order_id AND orders.user_id is NULL 
AND orders.created_at < CURDATE() - INTERVAL 3 MONTH 
AND (
    checkouts.email IS NULL
    OR
    checkouts.email NOT IN (
        SELECT checkouts.email 
        FROM orders, checkouts 
        WHERE orders.id = checkouts.order_id AND orders.user_id is NULL 
        AND orders.created_at > CURDATE() - INTERVAL 3 MONTH
    )
)

You might also consider more simply:

SELECT checkouts.email, MAX(orders.created_at)
FROM orders, checkouts 
WHERE orders.id = checkouts.order_id AND orders.user_id is NULL
GROUP BY checkouts.email 
HAVING MAX(orders.created_at) < CURDATE() - INTERVAL 3 MONTH