Question

This query returns forum questions/answers and their nested comments (similar to the StackOverflow paradigm).

    SELECT forum_qa.*,
           user_profiles.*,
           c.*,
           n.pid,
           v.*,
           Ifnull(n.ans_count, 0) AS ans_count
    FROM   forum_qa
           JOIN user_profiles
             ON user_id = forum_qa_author_id
           LEFT JOIN (SELECT *
                      FROM   votes) AS v
             ON forum_qa_id = v.forum_qa_id_fk
           LEFT JOIN (SELECT forum_cm_id,
                             forum_cm_author_id,
                             forum_qa_id_fk,
                             forum_cm_text,
                             forum_cm_timestamp,
                             forum_cm_flag,
                             first_name  AS forum_cm_first_name,
                             last_name   AS forum_cm_last_name,
                             facebook_id AS forum_cm_fb_id,
                             picture     AS forum_cm_picture,
                             moderator   AS forum_cm_moderator
                      FROM   forum_cm
                             JOIN user_profiles
                               ON user_id = forum_cm_author_id) AS c
             ON forum_qa_id = c.forum_qa_id_fk
           LEFT JOIN (SELECT forum_qa_parent_id AS pid,
                             COUNT(*)           AS ans_count
                      FROM   forum_qa
                      WHERE  forum_qa_parent_id IS NOT NULL
                      GROUP  BY forum_qa_parent_id) AS n
             ON forum_qa_id = n.pid
    WHERE  forum_qa_id LIKE "%"
           AND forum_qa_parent_id IS NULL
    ORDER  BY forum_qa_timestamp DESC
            LIMIT  0,3

I am trying to paginate the results and am hitting the following problem:

By placing LIMIT at the end of the query, I end up limiting the number of total rows, not the number of questions/answers.

For ex, LIMIT 0,3 in the last line gives me (qa: question/answer; cm: comment):

forum_qa_id   qa_text   forum_cm_id    cm_text
1             asd
2             wer       4              this is a comment
2             wer       5              this is another comment

instead of

forum_qa_id   qa_text   forum_cm_id    cm_text
1             asd
2             wer       4              this is a comment
2             wer       5              this is another comment
3             zxc       
3             zxc       7              yet another comment

Any suggestions how to modify my query in order to have LIMIT 0,3 return not 3 rows but 3 questions regardless of how many nested comments is has?

changes with @michael’s suggestion

    SELECT qa.*,
           user_profiles.*,
           c.*,
           n.pid,
           v.*,
           Ifnull(n.ans_count, 0) AS ans_count
    FROM   (SELECT * FROM forum_qa LIMIT 0, 3) qa
           JOIN user_profiles
             ON user_id = qa.forum_qa_author_id
           LEFT JOIN (SELECT *
                      FROM   votes) AS v
             ON qa.forum_qa_id = v.forum_qa_id_fk
           LEFT JOIN (SELECT forum_cm_id,
                             forum_cm_author_id,
                             forum_qa_id_fk,
                             forum_cm_text,
                             forum_cm_timestamp,
                             forum_cm_flag,
                             first_name  AS forum_cm_first_name,
                             last_name   AS forum_cm_last_name,
                             facebook_id AS forum_cm_fb_id,
                             picture     AS forum_cm_picture,
                             moderator   AS forum_cm_moderator
                      FROM   forum_cm
                             JOIN user_profiles
                               ON user_id = forum_cm_author_id) AS c
             ON qa.forum_qa_id = c.forum_qa_id_fk
           LEFT JOIN (SELECT forum_qa_parent_id AS pid,
                             COUNT(*)           AS ans_count
                      FROM   forum_qa
                      WHERE  forum_qa_parent_id IS NOT NULL
                      GROUP  BY forum_qa_parent_id) AS n
             ON qa.forum_qa_id = n.pid
    WHERE  qa.forum_qa_id LIKE "%"
           AND qa.forum_qa_parent_id IS NULL
    ORDER  BY qa.forum_qa_timestamp DESC

(HOPEFULLY) THE SOLUTION

After reading here

http://www.mysqlperformanceblog.com/2006/09/01/order-by-limit-performance-optimization/

I decided to index the field I wanted to ORDER BY (forum_qa_type) — once I did this, the query is returning the correct number of questions.

Thanks @Michael for helping out. Will update here if this stops working.

Answer 1

Assuming forum_qa is the table holding the questions, you can place its SELECT * into a subquery with its own limit. This is of course untested, but should work in principle.

SELECT     qa.*,
           user_profiles.*,
           c.*,
           n.pid,
           v.*,
           Ifnull(n.ans_count, 0) AS ans_count
    FROM   (SELECT * FROM forum_qa LIMIT 0, 3) qa 
          JOIN user_profiles
             ON user_profiles.user_id = qa.forum_qa_author_id
           LEFT JOIN (SELECT *.....
           -- etc...

Additional references to the table name forum_qa will need to be changed to its new alias qa.