Question

Before I explain my problem, I will quickly go over how the table structure is:

Type: MySQL
Posts/topic Table:

 int      int**     UNIX Time     Int        Int
--------------------------------------------------
| id | category | postdate | topic_id | is_topic |
--------------------------------------------------
|  1 |    a     | 12345678 |    1     |     1    |
--------------------------------------------------
|  2 |    a     | 12345678 |    1     |     0    |
--------------------------------------------------
|  3 |    b     | 12345678 |    3     |     1    |
--------------------------------------------------
|  4 |    b     | 12345678 |    3     |     0    |
--------------------------------------------------
|  5 |    c     | 12345678 |    5     |     1    |
--------------------------------------------------
|  6 |    c     | 12345678 |    5     |     0    |
--------------------------------------------------

**I’m using letters to make is easier to read the table

I am trying to retrieve the 4 newest rows for each category, and I am able to get the 1 newest from each category with GROUP BY, but I have no idea how to get multiple for each category.

I have tried something like this:

SELECT *
FROM posts p
WHERE NOT EXISTS
(
    SELECT *
    FROM posts
    WHERE category = p.category
    LIMIT 4
)

I also tried working with some of the other answers people supplied for some other answers here on SO, but I did not seem to be able to make them fit my purpose.

I am completely lost here, as mysql is not a strong side of mine when it comes to these more complex queries.

Any help or pointers in the right directions would really be appreciated!

Thanks!

UPDATE: Please note that the number of categories is not be static, and will change.

Answer 1

Something like this would probably do :

select * from
    (select 
         @rank:=CASE WHEN @ranked <> category THEN 1 ELSE @rank+1 END as rank,
         id,
         category,
         postdate,
         topic_id,
         is_topic
         @ranked:=category
     from 
          (select @rank := -1) a, 
          (select @ranked :=- -1) b, 
          (select * from posts order by category, postdate desc) c
    ) ranked_posts
where ranked_posts.rank <= 4

Basically what is happening here is that I am trying to create a “ranking” function present in other engines (MS SQL comes in mind).

The query goes through all the posts, ordered by category, postdate and adds a “ranking” number to every row resetting the rank when the category changes. Like:

rank | category  
1          a     
2          a     
...
100        a     
1          b 
2          b
1          c
2          c
3          c

You do that inside a “sub query” to mimic a table then just select the rows that have rank <= 4, (the ones you need). If you need more or less you can adjust that number.

One thing to keep in mind is that the ordering is important or the ranks will get all screwed up. The categories have to be “grouped”, hence the ORDER BY category and then the groups ordered by your criteria postdate desc.