Mysql match exact set of result in a single table - mysql

Question

I am having trouble with generating result set. This is what my 'user_roles' table looks like,
id user_id role_id
1 1 1
2 1 2
3 2 1
4 3 1
5 3 2
6 3 3
... ... ...
I want this result where user has exact both roles i:e 1 and 2, I do not want those user having roles other than 1,2.
id user_id role_id
1 1 1
2 1 2
... ... ...
I have tried so far,
SELECT
*
FROM
`user_roles`
WHERE `role_id` IN (1,2)
HAVING COUNT(id) = 2
But, it returns null.

Answer 1

Why your query doesn't work
HAVING applies after GROUP BY and your query doesn't have one. When the query contains HAVING or GROUP BY aggregate functions but it doesn't contain the GROUP BY clause, a single group containing all the selected rows is created.
Before applying HAVING, your query selects the rows having id in 1..5 (i.e. 5 rows). A single group is created from them, COUNT(id) returns 5 and the HAVING condition doesn't match. That's why the query doesn't return anything.
In order to correctly count the number of roles of each user it needs to group the records by user_id:
SELECT `user_id`
FROM `user_roles`
WHERE `role_id` IN (1, 2)
GROUP BY `user_id`
HAVING COUNT(`id`) = 2
This way, the WHERE clause selects the user having the roles 1 or 2 (but it ignores other roles), the GROUP BY clause allows the function COUNT(id) to count the number of selected roles for each user and the HAVING clause keeps only those users having both roles (1 and 2). The SELECT clause is not allowed to contain * because for the columns that are not in the GROUP BY clause, MySQL is free to pick any value it finds in the corresponding column and it may return different results on different executions of the query.
However, the query above doesn't return the values you want. It completely ignore the roles that are not 1 or 2 and it will return the user having user_id = 3.
A query that works
This query returns the users having only the roles 1 and 2 is:
SELECT `user_id`
FROM `user_roles`
GROUP BY `user_id`
HAVING COUNT(`role_id`) = 2 AND GROUP_CONCAT(`role_id`) = '1,2'
The condition COUNT(role_id) = 2 is not needed. In theory it should improve the execution speed (because counting works faster that string concatenation) but in real life it might have no impact whatsoever. The MySQL engine knows better.
Update
#martin-schneider asks in a comment:
is the order of GROUP_CONCAT(role_id) deterministic? or could it be that the result is '2,1'?
It's a very good question that has the answer in the documentation of function GROUP_CONCAT():
To sort values in the result, use the ORDER BY clause. To sort in reverse order, add the DESC (descending) keyword to the name of the column you are sorting by in the ORDER BY clause. The default is ascending order; this may be specified explicitly using the ASC keyword.
The complete query is:
SELECT `user_id`
FROM `user_roles`
GROUP BY `user_id`
HAVING COUNT(`role_id`) = 2
AND GROUP_CONCAT(`role_id` ORDER BY `role_id` ASC SEPARATOR ',') = '1,2'
I omitted ORDER BY and SEPARATOR because their default values (sort ascending by the values that are concatenated and use comma as separator) are good for our needs in this query.
Important to notice
There is a limit for the length of the value computed by the GROUP_CONCAT() function. Its result is truncated to the value stored in the system variable group_concat_max_len whose default value is 1024.
This value can be increased using the SET MySQL statement before running the query:
SET group_concat_max_len = 1000000
However, for this particular query the default limit of 1024 characters is more than enough.

Answer 2

You could aggregate by user_id and use HAVING:
SELECT *
FROM `user_roles`
WHERE `user_id` IN (SELECT user_id
FROM `user_roles`
GROUP BY user_id
HAVING SUM(role_id IN (1,2)) = 2
AND SUM(role_id NOT IN (1,2)) = 0);
LiveDemo*
*SQLFiddle does not respond so SQL Server equivalent
Note:
I assumed that user_id, role_id are unique and not null.