Using 'AND' in a many-to-many relationship - sql

Question

I have a Users table and a Groups table. Users can be in multiple groups via a 'UserInGroup' table and Groups can have a 'GroupTypeId'.
[User]
--------------
Id | Name
1 | Bob
2 | James
[UserInGroup]
-----------------
UserId | GroupId
1 1
1 2
[Group]
Id | Name | TypeId
------------------------
1 | Directors | 1
2 | IT | 1
3 | London | 2
I want to create a query to return for example users that are in both 'Directors' AND 'London' (rather than 'Directors' OR 'London'). However, I only want to AND groups of a different 'Type', I want to OR groups of the same type. I could do with having a separate table per group type but I can't as they are created dynamically.
Ideally I want to be able to query users who are in 'Directors' OR 'IT' AND 'London'.
What is the most efficient way of doing this?

Answer 1

This problem is commonly known as Relational Division.
SELECT a.Name
FROM [user] a
INNER JOIN UserInGroup b
ON a.ID = b.UserID
INNER JOIN [Group] c
ON b.groupID = c.TypeId
WHERE c.Name IN ('Directors','London')
GROUP BY a.Name
HAVING COUNT(*) = 2
SQLFiddle Demo
SQL of Relational Division
But if a UNIQUE constraint was not enforce on GROUP for every USER, DISTINCT keywords is needed to filter out unique groups:
SELECT a.Name
FROM [user] a
INNER JOIN UserInGroup b
ON a.ID = b.UserID
INNER JOIN [Group] c
ON b.groupID = c.TypeId
WHERE c.Name IN ('Directors','London')
GROUP BY a.Name
HAVING COUNT(DISTINCT c.Name) = 2
OUTPUT from both queries
╔══════╗
║ NAME ║
╠══════╣
║ Bob ║
╚══════╝

Answer 2

I arrived at the following solution (with help from J W and this article):
SELECT
u.Name UserName
FROM [User] u
INNER JOIN [UserInGroup] uig
ON uig.UserId = u.Id
INNER JOIN [Group] g
ON g.Id = uig.GroupId
WHERE
g.Id IN (1,2,3) -- these are the passed in groupids
GROUP BY
u.Name
having count(distinct g.TypeId)
= (select count(distinct g1.TypeId)
from [group] g1 where g1.Id IN (1,2,3))
This allows me to group the relational division by a discriminator field. An alternative would be this:
SELECT a.Name
FROM [User] a
INNER JOIN
(
SELECT b.UserID
FROM UserInGroup b
INNER JOIN [Group] c
ON b.groupID = c.Id
WHERE c.Name IN ('Directors','IT')
GROUP BY b.UserID
HAVING COUNT(DISTINCT c.Name) >= 1
) b ON a.ID = b.UserID
INNER JOIN
(
SELECT DISTINCT b.UserID
FROM UserInGroup b
INNER JOIN [Group] c
ON b.groupID = c.Id
WHERE c.Name = 'London'
) c ON a.ID = c.UserID
With an extra join for each GroupTypeId. Execution plans look similar, so I went with the first option.