Question

I have this database schema. I’m calling the stuff I’m tagging objects just for the sake of making it simple:

Users

• user_id
• login

User to Object table (many to many)

• user_id
• object_id

Object table

• object_id
• object_stuff

Object to tag with user_id table (many to many with extra column)

• user_id
• object_id
• tag_id

Tag table

• tag_id
• tag_name

I have been using a query like this in order to get the objects of a particular user as well as the objects’ tags (if the user tagged any). If you think you can optimize this, feel free to let me know.

@user_id = 'whatever user_id I want';
SELECT 
    o.object_id AS object_id,
    o.object_stuff AS object_stuff,
    IF (tags.tag_name IS NOT NULL, GROUP_CONCAT(tags.tag_name SEPARATOR '|'), 'N/A') AS tags
FROM  object_table AS o
LEFT OUTER JOIN  obj_to_users_table AS o2u ON o.object_id = o2u.object_id
LEFT OUTER JOIN  users AS u ON u.user_id = o2u.user_id 
LEFT OUTER JOIN  obj_to_tag_users AS o2tu ON o.object_id = o2tu.object_id AND u.user_id = o2tu.user_id
LEFT OUTER JOIN  tags AS t ON t.tag_id = o2t.tag_id 
WHERE u.user_id = @user_id GROUP BY o.object_id

Which would yield results like this:

object_id     object_stuff        tags
1             stuff1...           tag1|tag2|tag3
2             stuff2...           tag4|tag6|tag1
3             stuff3...           tag7|tag2|tag5

My problem is that I would like to, for example, search for objects that are tagged as ‘tag2’ and I should get:

object_id     object_stuff        tags
1             stuff1...           tag1|tag2|tag3
3             stuff3...           tag4|tag2|tag1

But instead, I lose the other tags that accompany the object:

object_id     object_stuff        tags
1             stuff1...           tag2
3             stuff3...           tag2

How do I modify my WHERE clause so that I can get the objects tagged as ‘tag2’ but also keeping the other tags the object has in the result set?

... WHERE u.user_id = @user_id AND t.tag_name LIKE '%tag2%' ...

Answer 1

First, calculate the concatenated results.

Next, perform your search against those concatenated results to preserve the groupings of the tags:

SELECT * 
FROM
(
    SELECT 
        o.object_id AS object_id,
        o.object_stuff AS object_stuff,
        IF (tags.tag_name IS NOT NULL, 
            GROUP_CONCAT(tags.tag_name SEPARATOR '|'), 'N/A') AS tags
    FROM  object_table AS o
    LEFT OUTER JOIN  obj_to_users_table AS o2u 
        ON o.object_id = o2u.object_id
    LEFT OUTER JOIN  users AS u 
        ON u.user_id = o2u.user_id 
    LEFT OUTER JOIN  obj_to_tag_users AS o2tu 
        ON o.object_id = o2tu.object_id AND u.user_id = o2tu.user_id
    LEFT OUTER JOIN  tags AS t ON t.tag_id = o2t.tag_id 
    WHERE u.user_id = @user_id 
    GROUP BY o.object_id
) sub
WHERE sub.tags LIKE '%tag2%'

EDIT:

---

Here’s an example that uses an EXISTS. I also switched the LEFT JOINs to INNER JOINs, since searching for a specific tag should eliminate any objects without tags anyways.

While this query looks more complicated, it should have a better chance of using any available indexes.

I had also mentioned doing the same type of query using an additional set of INNER JOINs, but that one would look even more complex since you’d still have to use a sub-query to use DISTINCT to get rid any possible duplicates introduced by the additional JOINs… so I’d suggest using EXISTS instead…

SELECT 
    o.object_id AS object_id,
    o.object_stuff AS object_stuff,
    GROUP_CONCAT(t.tag_name SEPARATOR '|') AS tags
FROM  object_table AS o
INNER JOIN  obj_to_users_table AS o2u 
    ON o.object_id = o2u.object_id
INNER JOIN  users AS u 
    ON u.user_id = o2u.user_id 
INNER JOIN  obj_to_tag_users AS o2tu 
    ON o.object_id = o2tu.object_id AND u.user_id = o2tu.user_id
INNER JOIN tags AS t 
    ON t.tag_id = o2t.tag_id
WHERE 
    u.user_id = @user_id 
    AND EXISTS (
        SELECT 1
        FROM tags AS targetTag
        WHERE 
            targetTag.tag_id = o2t.tag_id
            AND targetTag.tag_name = 'tag2'
    )
GROUP BY o.object_id