Question

Can someone please explain the discrepancy among the following queries:

Query 1 (returns 87 results):

SELECT userId, COUNT(userId) as usercount
FROM `cpnc_PaymentOrder` 
WHERE  created >= UNIX_TIMESTAMP('2011-03-01')
GROUP BY userId
HAVING usercount > 1

Query 2 (returns 177 results):

SELECT userId, COUNT(userId) as usercount
FROM `cpnc_PaymentOrder` 
WHERE  created >= UNIX_TIMESTAMP('2011-02-01')
GROUP BY userId
HAVING usercount > 1

Query 3 (returns 55 results):

SELECT userId, COUNT(userId) as usercount
FROM `cpnc_PaymentOrder` 
WHERE  created >= UNIX_TIMESTAMP('2011-02-01')
AND created < UNIX_TIMESTAMP('2011-03-01')
GROUP BY userId
HAVING usercount > 1

Now I would think that the number of results from Query 2 minus the number of results from Query 1 would equal the number of results from Query 3. But this is not the case. Can someone please explain why?

Thanks, Jonah

EDIT:

for clarification, the query i want to write is:

SELECT userId
FROM `cpnc_PaymentOrder` 
WHERE  created >= UNIX_TIMESTAMP('2011-02-01')
AND created < UNIX_TIMESTAMP('2011-03-01')
AND userId "appears in at least one other record from before '2011-03-01'"

Answer 1

Because you’re comparing number of groups.

Let’s see this on a small example:

user_id | created
      1 | 2011-02-02
      1 | 2011-02-02
      1 | 2011-02-02
      2 | 2011-02-02
      2 | 2011-02-02
      1 | 2011-03-02
      1 | 2011-03-02
      2 | 2011-03-02
      2 | 2011-03-02

First query for this dataset will return 2 rows, second will return also 2 rows, 3rd will return also 2 rows.

Just remove your GROUP BY and see the difference (without GROUP BY the math will match, of course)