Its better to normalize your schema do not store relations in form of comma separated list instead create a junction table for this like to maintain a m:m many to many relation between users and filters,create a new table as user_filters
with columns filter id and user id and in each row save one association per user and filter like in your current schema relation for filter 1 with many users (1, '1, 2, 3')
will become like
filter id user id
(1, '1'),
(1, '2'),
(1, '3'),
Sample schema will be like this
CREATE TABLE user_filters
(`fid` int, `u_id` varchar(50))
;
INSERT INTO user_filters
(`fid`, `u_id`)
VALUES
(1, '1'),
(1, '2'),
(1, '3'),
(2, '5'),
(2, '5')
;
CREATE TABLE filters
(`id` int, `title` varchar(50))
;
INSERT INTO filters
(`id`, `title`)
VALUES
(1, 'test'),
(2, 'test 1')
;
CREATE TABLE users
(`id` int, `name` varchar(6))
;
INSERT INTO users
(`id`, `name`)
VALUES
(1, 'Tom'),
(2, 'Tim'),
(3, 'Sue'),
(4, 'Bruce'),
(5, 'Ann'),
(6, 'George')
;
For above schema you can easily query with join as, below query can be optimized using indexes
select u.*
from users u
join user_filters uf on(uf.u_id = u.id)
where uf.fid =1
If you are not able to alter your schema and want to stick with the current one you can query as below but this one cannot be optimized enough as compare to above query
select u.*
from users u
join filters f on(find_in_set(u.id,replace(`u_ids`,' ','')) > 0)
where f.id =1