Limit results on an GROUP_CONCAT() or INNER JOIN

Question

I've perused extensively the other threads talking about limits on group_concat() and inner joins but haven't found my answer, so I guess I'll go ahead and ask it:

I'm developing an existing photo community site. I want to retrieve members who have their birthday on a given day (today) and then retrieve each member's 5 most highly rated photos. But I also only want the 10 "most favorite" birthday members (ie with the highest favorite count). Here's what I have:

SELECT users.user_id, users.user_name, 
       GROUP_CONCAT(CONVERT(photos.photo_id,char(32)) 
                    ORDER BY photos.average_rate) as photo_ids
FROM users 
INNER JOIN photos ON photos.user_id=users.user_id
WHERE users.day_of_birth = DATE_FORMAT('2012-04-17', '%m-%d') 
  AND users.photo_count>0 
GROUP BY users.user_id
ORDER BY users.favorite_count DESC, users.photo_count DESC LIMIT 0,10

This does what I want, EXCEPT that I cannot limit the amount of photo_id s to 5. This is a problem since the output will be sent as JSON to the app, and some members have uploaded upwards of 20,000 photos already, leading to an unacceptably long output string. The only "solution" that seems to work for me is setting the sever variable group_concat_max_len to something reasonable that will hold at least 5 ids, but this is very hacky and unreliable. Is there any way to return exactly 5 photo_id s per user with a single query? Or will I need to do a loop in my PHP?

I don't necessarily need the photo_ids in a comma-separated value, I can also ditch the group_concat() approach entirely and do an inner join if that is more feasible. But even there I'm not aware of a way to limit the results to 5.

Answer 1

These advanced ones are what makes me love MySQL :)

SELECT user_id, user_name, 
    GROUP_CONCAT(CONVERT(photo_id, char(32)) ORDER BY photos.average_rate) as photo_ids
FROM (  SELECT user_id, user_name, photo_id, favorite_count, photo_count, 
            (case when @user_id = user_id then @rownum := @rownum + 1 else CONCAT(@rownum := 1, @user_id := user_id) end) AS dummy_val
        FROM (  SELECT users.user_id, users.user_name, users.favorite_count, users.photo_count, photos.photo_id
                FROM users 
                INNER JOIN photos
                ON photos.user_id=users.user_id
                WHERE users.day_of_birth = DATE_FORMAT('2012-04-17', '%m-%d') 
                    AND users.photo_count > 0 
                ORDER BY users.id ASC, photos.average_rate ASC
             ) AS h, 
             (  @rownum := NULL, 
                @user_id := NULL
             ) AS vars
        HAVING rownum <= 5) AS h2
GROUP BY user_id
ORDER BY favorite_count DESC, photo_count DESC LIMIT 0, 10

Basicly I get all rows, and throw away all photos which are 6 or higher in calculated rownum.

Answer 2

SELECT u.user_id
     , u.user_name 
     , GROUP_CONCAT(p.photo_id ORDER BY p.average_rate) AS photo_ids
FROM 
    ( SELECT user_id
           , user_name
           , favorite_count 
           , photo_count
      FROM users
      WHERE day_of_birth = DATE_FORMAT('2012-04-17', '%m-%d') 
        AND photo_count > 0 
      ORDER BY favorite_count DESC
             , photo_count DESC 
      LIMIT 10
    ) AS u
  INNER JOIN
    photos AS p
      ON  p.user_id = u10.user_id
      AND p.average_rate >= 
          ( SELECT pp.average_rate 
            FROM photos AS pp
            WHERE pp.user_id = u10.user_id
            ORDER BY pp.average_rate DESC
            LIMIT 1 OFFSET 4
          ) 
GROUP BY u.user_id
ORDER BY u.favorite_count DESC
       , u.photo_count DESC