Question

I have a payments table that has a user_id and an amount. Users have a minimum withdrawal of $5.00, and they can have any number of payments to them. Their total balance isn’t stored anywhere (to keep it simple and avoid synchronization issues), instead it’s found by doing a query like:

SELECT SUM(amount) 
  FROM payment 
 WHERE user_id = 5;

I want to find out the grand total of all balances that meet the withdrawal requirement. What I wish I could do is something like:

  SELECT SUM(SUM(amount)) 
    FROM payment 
   WHERE sum(amount) >= 5.00
GROUP BY user_id;

Unfortunately I get the error

ERROR: aggregates not allowed in WHERE clause

How can I write this query? I’m using postgres 8.4.

Part 2:

I’m not sure if this is even possible, but this payments table also includes payments made by the user: the payment table includes amount, from_user_id, and to_user_id. It’s a single-entry accounting system. It would be great if I could also subtract the payments the user has made. The pseudo-sql is:

SELECT (sum of all amounts) 
 WHERE (sum(amount where to_user_id = 5) - sum(amount where from_user_id = 5)) > 5.00

Answer 1

You need to use the HAVING clause

Never heard anyone try and do nested aggregation. I suggest you do the inner bit separately, either with a with statment or using something like

select sum(foo) from (select sum(amount) as foo ...having ...)

I believe you want something like this for your second part, though I might be second-guessing.

with a as(
  select sum(amount) as credit, from_user from payment group by from_user
), b as (
  select sum(amount) as debit, to_user from payment group by to_user
)
-- glossing over any null values with coalesce:
select to_user, coalesce(credit,0) - coalesce(debit,0) -- coalesce users perhaps?
from a outer join b on from_user = to_user
where coalesce(credit,0) - coalesce(debit,0) > limit -- or vice-versa