Pretty similar to MYSQL - select first 4 records for each category in a table but there isn\'t an accepted answer and the one answer there doesn\'t make much sense so i\'m a
This should do what you are looking for.
SET @I=0;
SET @C='';
SELECT ID, Name, Category FROM (
SELECT B.*,
IF(@C != B.Category, @I:=1, @I:=@I+1) AS RowNum,
@C:=B.Category
FROM (
SELECT ID, Name, Category FROM Products GROUP BY Name, Category ORDER BY Category
) AS B HAVING RowNum <= 2
) AS A
Something along these lines will work:
SELECT id, name, category
FROM (
SELECT *,
IF( @prev <> category,
@rownum := 1,
@rownum := @rownum+1
) AS rank,
@prev := category,
@rownum
FROM (
SELECT * FROM products
ORDER BY category, rand()
) random_prodcts
) products_ranked
WHERE rank <= 2;
It orders them randomly within the categories, then pulls them out tracking how many it's got from each.
Not sure how nicely it will scale though.
EDIT: Tried it with a few thousand records and it seems ok.
FULL SQL:
SELECT t.id, t.name, t.category
FROM the_table AS t
, (SELECT MIN(id) AS id_min FROM the_table GROUP BY category) AS t1
, (SELECT MAX(id) AS id_max FROM the_table GROUP BY category) AS t2
WHERE (t.id=t1.id_min) OR (t.id=t2.id_max)
The query picks the min and the max of each category and return both of them (or only one if it is the same).
You said "The order of the selected products is of no importance, they might as well be random", so this method should be ok.