SQL query where ALL records in a join match a condition?

Question

I have what seems to be a simple problem, but can not figure out the proper solution via SQL. I'm using postgresql specifically.

Take the following:

SELECT * FROM users INNER JOIN tags ON (tags.user_id = users.id) WHERE tags.name IN ('word1', 'word2')

This does not do what I need. I want to find users whose tags are ONLY included in the list. If the user has a tag that is not in the list, the user should not be included.

'user1' tags: word1, word2, word3
'user2' tags: word1
'user3' tags: word1, word2

Given: word1 and word2. I want to prepare a query that returns 'user2' and 'user3'. 'user1' is excluded because it has a tag that is not in the list.

Hopefully I made this clear. Thanks for your help!

Answer 1

Relying on COUNT(*) = 2 will require that there can be no duplicates of user_id and name in the tags table. If that's the case, I'd go that route. Otherwise, this should work:

SELECT u.* 
FROM users AS u
WHERE u.id NOT IN (
    SELECT DISTINCT user_id FROM tags WHERE name NOT IN ('word1', 'word2')
) AND EXISTS (SELECT user_id FROM tags WHERE user_id = u.id) 

Answer 2

SELECT  user_id
FROM    users
WHERE   id IN
        (
        SELECT  user_id
        FROM    tags
        )
        AND id NOT IN
        (
        SELECT  user_id
        FROM    tags
        WHERE   name NOT IN ('word1', 'word2')
        )

or

SELECT  u.*
FROM    (
        SELECT  DISTINCT user_id
        FROM    tags
        WHERE   name IN ('word1', 'word2')
        ) t
JOIN    users u
ON      u.id = t.user_id
        AND t.user_id NOT IN
        (
        SELECT  user_id
        FROM    tags
        WHERE   name NOT IN ('word1', 'word2')
        )

Answer 3

SELECT distinct users.id
FROM users 
INNER JOIN tags ON (tags.user_id = users.id) 
group by users.id 
having count(*) = 2 
and min(tags.name) = 'word1'
and max(tags.name) = 'word2'

Answer 4

To get all users that don't have a tag that is not in the list, use the query below. Could be that users are returned that have no tag or only one tag matching the words, but I understand that is the desired functionality.

SELECT
  u.*
FROM
  users u
  LEFT JOIN tags t 
    ON t.user_id = u.userid AND
       t.name NOT IN ('word1', 'word2')
WHERE
  t.user_id IS NULL

Answer 5

SELECT u.*
FROM users u
INNER JOIN (
  SELECT user_id FROM tags WHERE name IN ('word1', 'word2')
  EXCEPT
  SELECT user_id FROM tags WHERE name NOT IN ('word1', 'word2')
) s ON u.id = s.user_id