3 Where clauses one no entry gives me NULL - mysql

Question

I have three tables, payout, earnings, user. I try to get first the userid which I use to get earnings and payouts information about the user. With this I also count together with DISTINCT and substract it to get the CURRENT earnings. The problem is if any of the table has no entry example the payouts table so I get NULL instead of number. How can I solve this. I tried IFNULL but did not work
This is how I do
SELECT SUM(DISTINCT user_earnings.amount) -
SUM(DISTINCT user_payouts.payout_amount) as total_earnings
FROM user, user_earnings, user_payouts
WHERE user.id = 103
and user_earnings.user_id = user.id
and user_payouts.user_id = user.id
EDIT: I made the current tables
user_earnings ( the problem is just only with the user_id 103
id user_id amount
1 102 250
2 102 1000
3 101 5000
4 102 352
18 102 375
19 102 442
20 103 338 <-----
user_payouts
id user_id payout_amount
1 102 500
2 102 100
3 101 1000
user
id payout_address
102 ***
103 ***
As you see payouts have has no entry about user_id 103 because he never did a payout. Thats why I get null (I think)

Answer 1

Try this:
SELECT u.id,
(SUM(DISTINCT ue.amount) -
COALESCE(SUM(DISTINCT up.payout_amount),0)) as total_earnings
FROM
user u
left outer join user_earnings ue on u.id = ue.user_id
left outer join user_payouts up on u.id = up.user_id
WHERE
u.id = 103;
Sorry, I didn't put enough effort in to the above. While it will work for a small sample, the above breaks down with a larger set as you are filtering out any instance where a single user got two payouts of the same amount.
I think this is your real solution:
SELECT
u.id,
(ue.earnings - up.payouts) as total_earnings
FROM
user u
inner join
(select distinct
u1.id,
SUM(COALESCE(ue.amount,0)) as earnings
from
user u1
left join user_earnings ue on u1.id = ue.user_id group by u1.id) ue on u.id = ue.id
inner join
(select distinct
u1.id,
SUM(COALESCE(up.payout_amount,0)) as payouts
from
user u1
left join user_payouts up on u1.id = up.user_id group by u1.id) up on u.id = up.id
WHERE
u.id = 103;
Just for fun, here's another option, much shorter:
SELECT
u.id,
((select COALESCE(SUM(ue.amount),0) from user_earnings ue where ue.user_id = u.id) -
(select COALESCE(SUM(up.payout_amount),0) from user_payouts up where up.user_id = u.id)) as total_earnings
FROM
user u
WHERE
u.id = 103;

Answer 2

Note1:
I would definitely NOT use Distinct in this manner... If the user in the future has another earned_amount for 338, the second 338 would not get added to the total.
Note2:
See my usage of COALESCE instead of IFNULL.
Note3:
I updated your join syntax. Please confirm I kept your join as desired.
EDIT: Updated answer and fiddle to account for duplicate earned_amounts per user as well as duplicate payout_amounts per user
View Fiddle to notice duplicates of both entered for user 103 (400 + 400 - 100 - 100))
CLICK Here for SQLFiddle
SELECT (
COALESCE(ue.earned_amount,0) - COALESCE(up.payout_amount,0)
) AS total_earnings
FROM user u
JOIN (SELECT user_id, SUM(earned_amount) AS earned_amount
FROM Earned
GROUP BY user_id) ue
ON ue.user_id = u.id
JOIN (SELECT user_id, SUM(payout_amount) AS payout_amount
FROM Payouts
GROUP BY user_id) up
ON up.user_id = u.id
WHERE u.id = 103
Performance Note:
If your Earned or Payout tables are expected to be large a few ideas for performance improvement will be.
limit the size of the returned result sets of the nested queries by adding a Where clause to each so they only return results for user_id 103.
Add indexes to both Earned and Payout for the column user_id.
OR first create these joins to a temp table with indexes, then join the temp table to this query

Answer 3

You have 3 problems -- DISTINCT, JOIN followed by GROUP BY, and failure to deal with NULL.
SELECT u.id,
(
SELECT ( IFNULL(SUM(amount), 0)
FROM user_earnings WHERE user_id = u.id ) -
SELECT ( IFNULL(SUM(payout_amount), 0 )
FROM user_payouts WHERE user_id = u.id )
) as total_earnings
FROM user u
WHERE u.id = 103;
should solve all of them.
DISTINCT was already discussed -- there could be duplicate value for a given user.
JOIN followed by GROUP BY -- I call this "inflate-deflate syndrone". The JOIN happens first, thereby inflating the number of rows, then the GROUP BY tries to compensate for it. Usually this is a performance problem; in your case it mangles the data, too.
NULL was discussed, but I thing this formulation is 'correct'.
This is a 'rare' case where subqueries is better than LEFT JOIN. It avoids the over-calculation of SUM. ghenghy's first solution is probably just as good for a single u.id; mine should work for multiple ids.