Question

I have this Points table:

===========================================
FIELD           :      TYPE
===========================================
id                 int(11)
user_id            int(11)
value              int(11)
type               enum('add','subtract')
timestamp          int(11)
===========================================

Basically this table stores records of user points.

Now my question is: How do I get the top 10 users with the highest amount of total points?

Note: I have been told to change my schema — drop the ‘type’ field and put positive and negative values for the ‘value’ field. I just want to know if I could achieve the desired query using the current schema.

Thanks in advance

Answer 1

I would try something like this:

SELECT
  user_id,
  SUM( IF(`type`='add',+1,-1) * `value`) AS user_points
FROM Points
GROUP BY user_id
ORDER BY SUM( IF(`type`='add',+1,-1) * `value`) DESC
LIMIT 10

Seems to work: http://sqlfiddle.com/#!2/dd207/1

Alternative, maybe faster, maybe not, the trick is moving out the SUM(IF()) part from the inner query:

http://sqlfiddle.com/#!2/dd207/9

SELECT
  user_id,
  SUM( IF(`type`='add',+1,-1) * `valuesum`)
FROM (
      SELECT
        user_id,
        `type`,
        SUM(`value`) AS `valuesum`
      FROM     Points
      GROUP BY user_id, `type`
     ) AS p
GROUP BY p.user_id
ORDER BY SUM( IF(`type`='add',+1,-1) * `valuesum`) DESC
LIMIT 10

Probably both solution have to do a full table scan to summarize everything, so this situation might be a good candidate for some wise denormalization (collecting sums per users in another table, for example).