標籤:
眾所周知,在MySQL中,如果直接 ORDER BY RAND() 的話,效率非常差,因為會多次執行。事實上,如果等值查詢也是用 RAND() 的話也如此,我們先來看看下面這幾個SQL的不同執行計畫和執行耗時。
首先,看下建表DDL,這是一個沒有顯式自增主鍵的InnoDB表:
[[email protected]]> show create table t_innodb_random\G*************************** 1. row ***************************Table: t_innodb_randomCreate Table: CREATE TABLE `t_innodb_random` (`id` int(10) unsigned NOT NULL,`user` varchar(64) NOT NULL DEFAULT ‘‘,KEY `idx_id` (`id`)) ENGINE=InnoDB DEFAULT CHARSET=latin1
往這個表裡灌入一些測試資料,至少10萬以上, id 欄位也是亂序的。
[[email protected]]> select count(*) from t_innodb_random\G*************************** 1. row ***************************count(*): 393216
1、常量等值檢索:
[[email protected]]> explain select id from t_innodb_random where id = 13412\G*************************** 1. row ***************************id: 1select_type: SIMPLEtable: t_innodb_randomtype: refpossible_keys: idx_idkey: idx_idkey_len: 4ref: constrows: 1Extra: Using index[[email protected]]> select id from t_innodb_random where id = 13412;1 row in set (0.00 sec)
可以看到執行計畫很不錯,是常量等值查詢,速度非常快。
2、使用RAND()函數乘以常量,求得隨機數後檢索:
[[email protected]]> explain select id from t_innodb_random where id = round(rand()*13241324)\G*************************** 1. row ***************************id: 1select_type: SIMPLEtable: t_innodb_randomtype: indexpossible_keys: NULLkey: idx_idkey_len: 4ref: NULLrows: 393345Extra: Using where; Using index[[email protected]]> select id from t_innodb_random where id = round(rand()*13241324)\GEmpty set (0.26 sec)
可以看到執行計畫很糟糕,雖然是只掃描索引,但是做了全索引掃描,效率非常差。因為WHERE條件中包含了RAND(),使得MySQL把它當做變數來處理,無法用常量等值的方式查詢,效率很低。
我們把常量改成取t_innodb_random表的最大id值,再乘以RAND()求得隨機數後檢索看看什麼情況:
[[email protected]]> explain select id from t_innodb_random where id = round(rand()*(select max(id) from t_innodb_random))\G*************************** 1. row ***************************id: 1select_type: PRIMARYtable: t_innodb_randomtype: indexpossible_keys: NULLkey: idx_idkey_len: 4ref: NULLrows: 393345Extra: Using where; Using index*************************** 2. row ***************************id: 2select_type: SUBQUERYtable: NULLtype: NULLpossible_keys: NULLkey: NULLkey_len: NULLref: NULLrows: NULLExtra: Select tables optimized away[[email protected]]> select id from t_innodb_random where id = round(rand()*(select max(id) from t_innodb_random))\GEmpty set (0.27 sec)
可以看到,執行計畫依然是全索引掃描,執行耗時也基本相當。
3、改造成普通子查詢模式 ,這裡有兩次子查詢
[[email protected]]> explain select id from t_innodb_random where id = (select round(rand()*(select max(id) from t_innodb_random)) as nid)\G*************************** 1. row ***************************id: 1select_type: PRIMARYtable: t_innodb_randomtype: indexpossible_keys: NULLkey: idx_idkey_len: 4ref: NULLrows: 393345Extra: Using where; Using index*************************** 2. row ***************************id: 3select_type: SUBQUERYtable: NULLtype: NULLpossible_keys: NULLkey: NULLkey_len: NULLref: NULLrows: NULLExtra: Select tables optimized away[[email protected]]> select id from t_innodb_random where id = (select round(rand()*(select max(id) from t_innodb_random)) as nid)\GEmpty set (0.27 sec)
可以看到,執行計畫也不好,執行耗時較慢。
4、改造成JOIN關聯查詢,不過最大值還是用常量表示
[[email protected]]> explain select id from t_innodb_random t1 join (select round(rand()*13241324) as id2) as t2 where t1.id = t2.id2\G*************************** 1. row ***************************id: 1select_type: PRIMARYtable: <derived2>type: systempossible_keys: NULLkey: NULLkey_len: NULLref: NULLrows: 1Extra:*************************** 2. row ***************************id: 1select_type: PRIMARYtable: t1type: refpossible_keys: idx_idkey: idx_idkey_len: 4ref: constrows: 1Extra: Using where; Using index*************************** 3. row ***************************id: 2select_type: DERIVEDtable: NULLtype: NULLpossible_keys: NULLkey: NULLkey_len: NULLref: NULLrows: NULLExtra: No tables used[[email protected]]> select id from t_innodb_random t1 join (select round(rand()*13241324) as id2) as t2 where t1.id = t2.id2\GEmpty set (0.00 sec)
這時候執行計畫就非常完美了,和最開始的常量等值查詢是一樣的了,執行耗時也非常之快。
這種方法雖然很好,但是有可能查詢不到記錄,改造範圍尋找,但結果LIMIT 1就可以了:
[[email protected]]> explain select id from t_innodb_random where id > (select round(rand()*(select max(id) from t_innodb_random)) as nid) limit 1\G*************************** 1. row ***************************id: 1select_type: PRIMARYtable: t_innodb_randomtype: indexpossible_keys: NULLkey: idx_idkey_len: 4ref: NULLrows: 393345Extra: Using where; Using index*************************** 2. row ***************************id: 3select_type: SUBQUERYtable: NULLtype: NULLpossible_keys: NULLkey: NULLkey_len: NULLref: NULLrows: NULLExtra: Select tables optimized away[[email protected]]> select id from t_innodb_random where id > (select round(rand()*(select max(id) from t_innodb_random)) as nid) limit 1\G*************************** 1. row ***************************id: 13011 row in set (0.00 sec)
可以看到,雖然執行計畫也是全索引掃描,但是因為有了LIMIT 1,只需要找到一條記錄,即可終止掃描,所以效率還是很快的。
小結:
從資料庫中隨機取一條記錄時,可以把RAND()產生隨機數放在JOIN子查詢中以提高效率。
5、再來看看用ORDRR BY RAND()方式一次取得多個隨機值的方式:
[[email protected]]> explain select id from t_innodb_random order by rand() limit 1000\G*************************** 1. row ***************************id: 1select_type: SIMPLEtable: t_innodb_randomtype: indexpossible_keys: NULLkey: idx_idkey_len: 4ref: NULLrows: 393345Extra: Using index; Using temporary; Using filesort[[email protected]]> select id from t_innodb_random order by rand() limit 1000;1000 rows in set (0.41 sec)
全索引掃描,產生排序暫存資料表,太差太慢了。
6、把隨機數放在子查詢裡看看:
[[email protected]]> explain select id from t_innodb_random where id > (select rand() * (select max(id) from t_innodb_random) as nid) limit 1000\G*************************** 1. row ***************************id: 1select_type: PRIMARYtable: t_innodb_randomtype: indexpossible_keys: NULLkey: idx_idkey_len: 4ref: NULLrows: 393345Extra: Using where; Using index*************************** 2. row ***************************id: 3select_type: SUBQUERYtable: NULLtype: NULLpossible_keys: NULLkey: NULLkey_len: NULLref: NULLrows: NULLExtra: Select tables optimized away[[email protected]]> select id from t_innodb_random where id > (select rand() * (select max(id) from t_innodb_random) as nid) limit 1000\G1000 rows in set (0.04 sec)
嗯,提速了不少,這個看起來還不賴:)
7、仿照上面的方法,改成JOIN和隨機數子查詢關聯
[[email protected]]> explain select id from t_innodb_random t1 join (select rand() * (select max(id) from t_innodb_random) as nid) t2 on t1.id > t2.nid limit 1000\G*************************** 1. row ***************************id: 1select_type: PRIMARYtable: <derived2>type: systempossible_keys: NULLkey: NULLkey_len: NULLref: NULLrows: 1Extra:*************************** 2. row ***************************id: 1select_type: PRIMARYtable: t1type: rangepossible_keys: idx_idkey: idx_idkey_len: 4ref: NULLrows: 196672Extra: Using where; Using index*************************** 3. row ***************************id: 2select_type: DERIVEDtable: NULLtype: NULLpossible_keys: NULLkey: NULLkey_len: NULLref: NULLrows: NULLExtra: No tables used*************************** 4. row ***************************id: 3select_type: SUBQUERYtable: NULLtype: NULLpossible_keys: NULLkey: NULLkey_len: NULLref: NULLrows: NULLExtra: Select tables optimized away[[email protected]]> select id from t_innodb_random t1 join (select rand() * (select max(id) from t_innodb_random) as nid) t2 on t1.id > t2.nid limit 1000\G1000 rows in set (0.00 sec)
可以看到,全索引檢索,發現符合記錄的條件後,直接取得1000行,這個方法是最快的。
綜上,想從MySQL資料庫中隨機取一條或者N條記錄時,最好把RAND()產生隨機數放在JOIN子查詢中以提高效率。
上面說了那麼多的廢話,最後簡單說下,就是把下面這個SQL:
SELECT id FROM table ORDER BY RAND() LIMIT n;
改造成下面這個:
SELECT id FROM table t1 JOIN (SELECT RAND() * (SELECT MAX(id) FROM table) AS nid) t2 ON t1.id > t2.nid LIMIT n;
如果想要達到完全隨機,還可以改成下面這種寫法:
SELECT id FROM table t1 JOIN (SELECT round(RAND() * (SELECT MAX(id) FROM table)) AS nid FROM table LIMIT n) t2 ON t1.id = t2.nid;
就可以享受在SQL中直接取得隨機數了,不用再在程式中構造一串隨機數去檢索了。
From: http://imysql.com/2014/07/04/mysql-optimization-case-rand-optimize.shtml
[MySQL最佳化案例]系列 — RAND()最佳化