Question

This a simplified version of my SQL table for questions and answers, post_id is the primary key.

 post_id | ref_post_id | title  | date (unix timestamp)
   1     |     0       | Title1 |  10 
   2     |     1       |        |  20 
   3     |     0       | Title2 |  30 

When ref_post_id is 0, that means it’s a question. When it is not 0 it’s an answer where the number corresponds to the question’s post_id. That means that the second row is an answer to the first row in my example above.

What I want to do is to get only the last row of every question. If the question does not have any answers I want the question row, but if it has answers I want the last answer but I want to join in the title that belongs to the question row.

I’ve googled and tried myself for a while now. Hopefully you understand my problem!

Thanks!

Answer 1

I would reccommend to split your table in two different tables, one for the questions and one for the answer.

However, the query you need could be written as this:

SELECT posts.*
FROM
  posts INNER JOIN
    (SELECT IF(ref_post_id=0,post_id,ref_post_id) as id, Max(Date) as maxdate
     FROM posts
     GROUP BY id) p
  ON
    (p.maxdate = posts.date) AND
    (p.id =IF(posts.ref_post_id=0,posts.post_id,posts.ref_post_id))

To select max date for every question/answer you have to group by ref_post_id, except the case in which ref_post_id = 0 where you have to group by post_id, hence the IF.

When you join the posts table with itself, using the subquery, you have to use the same trick.

EDIT: if you need to mark a question as deleted, then you have to remove all deleted questions and answers from the select above, and you have to use one more subquery to select only undeleted questions:

SELECT posts.*
FROM
  posts INNER JOIN
    (SELECT IF(ref_post_id=0,post_id,ref_post_id) as id, Max(Date) as maxdate
     FROM posts
     WHERE EXISTS (SELECT null 
                   FROM posts posts_sub
                   WHERE
                     posts_sub.post_id =
                     IF(posts.ref_post_id=0,posts.post_id,posts.ref_post_id)
                   AND posts_sub.deleted =0)
     GROUP BY id) p
  ON
    (p.maxdate = posts.date) AND
    (p.id =IF(posts.ref_post_id=0,posts.post_id,posts.ref_post_id))