I have some kind of impossible request :).
I have a table where one of the columns is named type
. I would like to SELECT 3 records for each type in tha
When the table is large and collection is more unpredictable, the row numbering needs to be ordered by type in the inner query for the side-effecting variables to work.
select id, title, type
from (select id, title, type,
@r := CASE WHEN @g = type THEN @r+1 ELSE 1 END r,
@g := type
from tbl
order by type, title) as x
where row_number <= 3
# order by type, title
Another way to do this without using side effecting variables, if no two records are exactly the same on (title, type, id), is given below. This uses only standard ANSI SQL92 SQL. It may be slower than the above though.
select A.id, A.title, A.type
from tbl A
left join tbl B on
A.title = B.title and
(A.type < B.type or
(A.type = B.type and A.id < A.id))
group by A.id, A.title, A.type
having count(B.title) <= 2
If you have an index on (type, title)
, and you know the possible values for type
, I believe that dynamic SQL is the way to go (for once) for best performance.
For each possible value of type
, add a union all and a select for that specific type. The final query will look like the following query:
(select * from t1 where type = 'string1' order by title limit 3)
union all
(select * from t1 where type = 'string2' order by title limit 3)
union all
(select * from t1 where type = 'string3' order by title limit 3);
It executes in less than 1 second on a table with 1,000,000 rows, whereas the others solutions (Martins & Cyberkiwis) takes roughly 11 seconds.
The difference is because the unioned query above can fetch the first three title entries for each type and then stop, whereas the simulated analytics function has to scan the entire table.
Check out this article. Given:
+--------+------------+-------+
| type | variety | price |
+--------+------------+-------+
| apple | gala | 2.79 |
| apple | fuji | 0.24 |
| apple | limbertwig | 2.87 |
| orange | valencia | 3.59 |
| orange | navel | 9.36 |
| pear | bradford | 6.05 |
| pear | bartlett | 2.14 |
| cherry | bing | 2.55 |
| cherry | chelan | 6.33 |
+--------+------------+-------+
Query:
select type, variety, price
from fruits
where (
select count(*) from fruits as f
where f.type = fruits.type and f.price < fruits.price
) <= 2;
select id, title, type
from (select id, title, type,
@num := if(@group = type, @num + 1, 1) as row_number,
@group := type as dummy
from your_table
order by type, title) as x
where row_number <= 3
(Uses a different article on the same site as Martin Wickman's answer!)