Question

Please note that the database structure probably can’t be changed without a lot of work due to the amount of users and data on it.

The “friends” table is basically like this:

> show create table `friends`
CREATE TABLE `friends` (
    `id` int(10) unsigned NOT NULL auto_increment,
    `user1` int(10) unsigned NOT NULL,
    `user2` int(10) unsigned NOT NULL,
    PRIMARY KEY (`id`),
    UNIQUE KEY `user1_2` (`user1`,`user2`),
    KEY `user1` (`user1`),
    KEY `user2` (`user2`)
) ENGINE=InnoDB AUTO_INCREMENT=747908

To fetch a user’s friends, I have three options:

1. Individually select user2 where user1 equals the user’s ID, and vice versa, then combine the results in PHP.
2. SELECT IF(user1=@userid,user2,user1) FROM friends WHERE @userid IN (user1,user2)
3. SELECT user2 FROM friends WHERE user1=@userid
UNION SELECT user1 FROM friends WHERE user2=@userid

I tried timing options 2 and 3, and this is where I have a problem: The first time I run it, option 2 takes about 400ms whereas option 3 only takes less then 1ms. Every other time, however, opton 2 takes 0.6ms and option 2 takes 0.8ms.

What should I do? Which option is actually faster? The EXPLAIN queries return this:

id   select_type  table      type  possible_keys key     key_len ref   rows   Extra
1    SIMPLE       friends    index NULL          user1_2 8       NULL  386438 Using where; Using index

id   select_type  table      type  possible_keys key     key_len ref   rows   Extra
1    PRIMARY      friends    ref   user1,user1_2 user1_2 4       const 8      Using index
2    UNION        friends    ref   user2         user2   4       const 8
NULL UNION RESULT <union1,2> ALL   NULL          NULL    NULL    NULL  NULL

Answer 1

As usual when benchmarking, beware of caches.

Measure your SELECT queries using SQL_NO_CACHE clause (see the SELECT syntax).