0

I have performance problem! I searched for Inner Join, tried optimize table but still the same. I have Users, Entries, Tags and tag_user tables.

This code works perfectly :

select * from `entries` 
where (
    select count(*) from `users` where `entries`.`profile_id` = `users`.`id`
) >= 1 
order by `id` desc

And I should add this to complete my query :

select count(*) from `tags` 
inner join `tag_user` on `tags`.`id` = `tag_user`.`tag_id` 
where `tag_user`.`user_id` = `users`.`id` and `tags`.`id` in (?,?)) = 2

It becomes like this :

select * from `entries` 
where (
   select count(*) from `users` 
   where `entries`.`profile_id` = `users`.`id` and (
        select count(*) from `tags` 
        inner join `tag_user` on `tags`.`id` = `tag_user`.`tag_id` 
        where `tag_user`.`user_id` = `users`.`id` and `tags`.`id` in (119,36)) = 2
    ) >= 1 
order by `id` desc

Any solution for better performance ?

0

The first query will be faster this way:

select * from `entries` 
    where EXISTS (
        select * from `users` where `entries`.`profile_id` = `users`.`id`
                 ) 
order by `id` desc

It needs an index (PRIMARY KEY?) on users.id. It is faster because the subquery quits when finds one row, rather than counting all of them.

The query part cannot use that technique. But I think it is much simpler than what you have:

SELECT  e.*
    FROM  
      ( SELECT  user_id
            FROM  `tag_user`
            WHERE  tag_id IN (119, 36)
            GROUP BY  user_id
            HAVING  count(*) = 2 
      ) x
    JOIN  entries AS e  ON e.profile_id = x.user_id 

Notice that users does not seem to be of any use.

entries will need INDEX(profile_id).

I assume tag_user is a many-to-many mapping table? Have you followed all the advice in this? The advice there may provide further performance boost.

If you need to discuss this further, please provide SHOW CREATE TABLE.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.