Question

here is my 2 tables, id for a inner join event. i want to do this thing:

in table b, there has 10 albums, i want random get out 4 albums. then each album select one record, the record is random position in the album.

so that i will get 4 records back (these 4 records with no duplicate id), then take these 4 records for a inner join query, to get the title from table a.

here just little records just for test. in the fact, i have 300,000 records in table a and 2,000,000 records in table b.

table a

+-----+-------+
| id  | title | 
+-----+-------+
| 1   | a1    |
+-----+-------+
| 2   | a2    |
+-----+-------+
| 3   | a3    |
+-----+-------+
| 4   | a4    |
+-----+-------+
| 5   | a5    |
+-----+-------+
| 6   | a6    |
+-----+-------+

table b

+-----+--------+
| id  | album  | 
+-----+--------+
|  1  | album1 | 
+-----+--------+
|  2  | album1 | 
+-----+--------+
|  3  | album1 | 
+-----+--------+
|  6  | album1 | 
+-----+--------+
|  2  | album2 | 
+-----+--------+
|  3  | album2 | 
+-----+--------+
|  5  | album3 | 
+-----+--------+
|  6  | album3 | 
+-----+--------+
|  3  | album4 | 
+-----+--------+
|  2  | album5 | 
+-----+--------+
|  4  | album5 | 
+-----+--------+
|  5  | album5 | 
+-----+--------+
|  1  | album6 | 
+-----+--------+
|  3  | album6 | 
+-----+--------+
|  2  | album7 | 
+-----+--------+
|  4  | album7 | 
+-----+--------+
|  1  | album8 | 
+-----+--------+
|  5  | album8 | 
+-----+--------+
|  3  | album9 | 
+-----+--------+
|  2  | album10| 
+-----+--------+
|  5  | album10| 
+-----+--------+

I am not good at mysql query. In my mind I would do

select * from b group by album order by random() limit 0,4 

get back 4 album, then do a inner join query (this query not correct, how to check the b.id no duplicate?)

select * from b inner join a on b.id = a.id where (select id from b where b.album = '".$row['album']."'  order by random() limit 1) 

I need an easy and quicker method, the best is just use one query. many thanks.

Answer 1

AFAIR, “ORDER BY RAND()” is extremely slow solutions, especially on tables like you have (2 million+ records), so I’d recommend looking at something similar to these kind of articles first: http://www.greggdev.com/web/articles.php?id=6

So, you should know the number of records in your table before running the query and then do something like:

"SELECT * FROM `album` LIMIT 1 OFFSET " . rand(0,$count)

This will return you 1 random row a bit more efficiently, I believe.

Also, I think it’s not a good idea to store album references as string in tracks table, you’d rather use a proper integer foreign key album_id referenced to albums.id. Then you can join both tables much fatser. If I were you, I’d do first:

ALTER TABLE `tracks` add column `album_id` int;
UPDATE `tracks` SET `album_id` = SUBSTRING(`album`,5);

Then, after doing this and combining with the solution above, launch something like:

"SELECT * FROM `album` INNER JOIN `tracks`ON `tracks`.`album_id` = `albums`.`id` LIMIT 1 OFFSET " . rand(0,$count)