MySQL 中 where id in (1,2,3,4,...) 的效率问题讨论
MySQL
[求证&散分]MySQL 中 where id in (1,2,3,4,...) 的效率问题讨论
庆祝本月大版得分过万,兼把在这段论坛中经常被问到的一个问题拿出来大家讨论一下。
命题假设:
测试表如下
create table t_06 (
id int not null primary key,
c1 varchar(30),
i2 int
) engine = myisam;
delimiter //
CREATE PROCEDURE prepareData_t_06 ()
BEGIN
DECLARE i INT DEFAULT 1;
WHILE i < 500000 DO
insert into t_06 values (i,concat('AA',i),i);
SET i = i + 1;
END WHILE;
END;
//
delimiter ;
CALL prepareData_t_06();
select count(*) from t_06;
show index from t_06;
show table status like 't_06';
现在如果由程序在数组中提供 100 个 ID 如下
38068,238833,308799,274344,299910,309823,337486,290812,56393,64413,492511,260426,58626,450987,499220,187731,365665,212799,227788,255724,384636,465766,417310,313148,483987,328761,402876,237274,249183,174185,28187,189155,259940,67800,60065,340172,311667,354861,182,305523,115981,365082,213915,47894,131301,198754,358852,112496,404423,486725,233123,322936,325337,125932,299260,128791,295663,469897,120580,347766,34859,364204,37597,268974,351155,256955,214013,309192,412394,216800,30315,411242,16678,233247,359013,401666,30792,452394,408649,14159,5519,91705,227648,120966,319599,351170,68129,368701,233566,144256,156172,41972,499687,390955,6549,298079,498230,196397,239493,242037
如何实现查询效率比较高,请给出你的意思,测试方案和对比结果者加分。
select * from t_06 where id in (38068,238833,308799,274344,299910,309823,337486,290812,56393,64413,492511,260426,58626,450987,499220,187731,365665,212799,227788,255724,384636,465766,417310,313148,483987,328761,402876,237274,249183,174185,28187,189155,259940,67800,60065,340172,311667,354861,182,305523,115981,365082,213915,47894,131301,198754,358852,112496,404423,486725,233123,322936,325337,125932,299260,128791,295663,469897,120580,347766,34859,364204,37597,268974,351155,256955,214013,309192,412394,216800,30315,411242,16678,233247,359013,401666,30792,452394,408649,14159,5519,91705,227648,120966,319599,351170,68129,368701,233566,144256,156172,41972,499687,390955,6549,298079,498230,196397,239493,242037);
1、
直接用IN
select * from t_06 a inner join lsbtest b on a.id=b.id1;
2、用LSBTEST
DROP TABLE IF EXISTS `lsbtest`;
CREATE TABLE `lsbtest` (
`id1` bigint(10) DEFAULT NULL
) ENGINE=InnoDB DEFAULT CHARSET=utf8;
/*Data for the table `lsbtest` */
LOCK TABLES `lsbtest` WRITE;
insert into `lsbtest`(`id1`) values (1),(2),(38068),(238833),(308799),(274344),(299910),(309823),(337486),(290812),(56393),(64413),(492511),(260426),(58626),(450987),(499220),(187731),(365665),(212799),(227788),(255724),(384636),(465766),(417310),(313148),(483987),(328761),(402876),(237274),(249183),(174185),(28187),(189155),(259940),(67800),(60065),(340172),(311667),(354861),(182),(305523),(115981),(365082),(213915),(47894),(131301),(198754),(358852),(112496),(404423),(486725),(233123),(322936),(325337),(125932),(299260),(128791),(295663),(469897),(120580),(347766),(34859),(364204),(37597),(268974),(351155),(256955),(214013),(309192),(412394),(216800),(30315),(411242),(16678),(233247),(359013),(401666),(30792),(452394),(408649),(14159),(5519),(91705),(227648),(120966),(319599),(351170),(68129),(368701),(233566),(144256),(156172),(41972),(499687),(390955),(6549),(298079),(498230),(196397),(239493),(242037);
UNLOCK TABLES;
select * from t_06 a inner join lsbtest b on a.id=b.id1;
1、
select * from t_06 where id in (38068,238833,308799,274344,299910,309823,337486,290812,56393,64413,492511,260426,58626,450987,499220,187731,365665,212799,227788,255724,384636,465766,417310,313148,483987,328761,402876,237274,249183,174185,28187,189155,259940,67800,60065,340172,311667,354861,182,305523,115981,365082,213915,47894,131301,198754,358852,112496,404423,486725,233123,322936,325337,125932,299260,128791,295663,469897,120580,347766,34859,364204,37597,268974,351155,256955,214013,309192,412394,216800,30315,411242,16678,233247,359013,401666,30792,452394,408649,14159,5519,91705,227648,120966,319599,351170,68129,368701,233566,144256,156172,41972,499687,390955,6549,298079,498230,196397,239493,242037);
(100 row(s) returned)
Execution Time : 00:00:00:000
Transfer Time : 00:00:00:000
Total Time : 00:00:00:000
2、
select * from t_06 a inner join lsbtest b on a.id=b.id1;
lsbtest:没有索引
(102 row(s) returned)
Execution Time : 00:00:00:000
Transfer Time : 00:00:00:000
Total Time : 00:00:00:000
如果直接运行1、2则两者在速度上基本相近
但2如果加上建表、插入数据,则
DROP TABLE IF EXISTS `lsbtest`;
CREATE TABLE `lsbtest` (
`id1` bigint(10) DEFAULT NULL
) ENGINE=InnoDB DEFAULT CHARSET=utf8;
/*Data for the table `lsbtest` */
insert into `lsbtest`(`id1`) values (1),(2),(38068),(238833),(308799),(274344),(299910),(309823),(337486),(290812),(56393),(64413),(492511),(260426),(58626),(450987),(499220),(187731),(365665),(212799),(227788),(255724),(384636),(465766),(417310),(313148),(483987),(328761),(402876),(237274),(249183),(174185),(28187),(189155),(259940),(67800),(60065),(340172),(311667),(354861),(182),(305523),(115981),(365082),(213915),(47894),(131301),(198754),(358852),(112496),(404423),(486725),(233123),(322936),(325337),(125932),(299260),(128791),(295663),(469897),(120580),(347766),(34859),(364204),(37597),(268974),(351155),(256955),(214013),(309192),(412394),(216800),(30315),(411242),(16678),(233247),(359013),(401666),(30792),(452394),(408649),(14159),(5519),(91705),(227648),(120966),(319599),(351170),(68129),(368701),(233566),(144256),(156172),(41972),(499687),(390955),(6549),(298079),(498230),(196397),(239493),(242037);
select * from t_06 a inner join lsbtest b on a.id=b.id1;
Total Time : 00:00:01:094
Total Time : 00:00:00:250
Total Time : 00:00:00:015
大约1.359秒
用临时表的话,应该把创建临时表和删除临时表的时间算上
试试这个不创建表的方法(没有环境测试,不知道性能如何):
select
a.*
from
t_06 as a
inner join
(select 38068 as id
union all select 238833 union all select 308799
union all select 274344
union all select 299910
union all select 309823
union all select 337486
union all select 290812
union all select 56393
union all select 64413
union all select 492511
union all select 260426
union all select 58626
union all select 450987
union all select 499220
union all select 187731
union all select 365665
union all select 212799
union all select 227788
union all select 255724
union all select 384636
union all select 465766
union all select 417310
union all select 313148
union all select 483987
union all select 328761
union all select 402876
union all select 237274
union all select 249183
union all select 174185
union all select 28187
union all select 189155
union all select 259940
union all select 67800
union all select 60065
union all select 340172
union all select 311667
union all select 354861
union all select 182
union all select 305523
union all select 115981
union all select 365082
union all select 213915
union all select 47894
union all select 131301
union all select 198754
union all select 358852
union all select 112496
union all select 404423
union all select 486725
union all select 233123
union all select 322936
union all select 325337
union all select 125932
union all select 299260
union all select 128791
union all select 295663
union all select 469897
union all select 120580
union all select 347766
union all select 34859
union all select 364204
union all select 37597
union all select 268974
union all select 351155
union all select 256955
union all select 214013
union all select 309192
union all select 412394
union all select 216800
union all select 30315
union all select 411242
union all select 16678
union all select 233247
union all select 359013
union all select 401666
union all select 30792
union all select 452394
union all select 408649
union all select 14159
union all select 5519
union all select 91705
union all select 227648
union all select 120966
union all select 319599
union all select 351170
union all select 68129
union all select 368701
union all select 233566
union all select 144256
union all select 156172
union all select 41972
union all select 499687
union all select 390955
union all select 6549
union all select 298079
union all select 498230
union all select 196397
union all select 239493
union all select 242037
) as b
on
a.id=b.id
临时表:
DELIMITER $$
DROP PROCEDURE IF EXISTS `zz`.`cxzfnewa`$$
CREATE DEFINER=`root`@`localhost` PROCEDURE `cxzfnewa`(i varchar(5000))
BEGIN
DECLARE ai INT DEFAULT 1;
DROP TABLE IF EXISTS `t_06`;
create table t_06 (
id int not null primary key,
c1 varchar(30),
i2 int
) engine = myisam;
WHILE ai < 500000 DO
insert into t_06 values (ai,concat('AA',ai),ai);
SET ai = ai + 1;
END WHILE;
DROP TABLE IF EXISTS `lsb2`;
CREATE TABLE `lsb2` (
`id` bigint(10) DEFAULT NULL
) ENGINE=InnoDB DEFAULT CHARSET=utf8;
set ai=1;
while ai<=1000 do
insert into lsb2 values (ai);
set ai=ai+1;
end while;
DROP TABLE IF EXISTS `lsbtest`;
CREATE TABLE `lsbtest` (
`id1` bigint(10) DEFAULT NULL
) ENGINE=InnoDB DEFAULT CHARSET=utf8;
insert into lsbtest
select
mid(mid(i,id,length(i)),2,LOCATE(',',mid(i,id,length(i)),2)-2)
from lsb2 where length(i)>=id and mid(i,id,1)=','
and length(mid(i,id,length(i)))>=2;
select * from t_06 a inner join lsbtest b on a.id=b.id1;
END$$
DELIMITER ;
Total Time : 00:00:47:719
上一个调用:
call cfzfnew(',38068,238833,308799,274344,299910,309823,337486,290812,56393,64413,492511,260426,58626,450987,499220,187731,365665,212799,227788,255724,384636,465766,417310,313148,483987,328761,402876,237274,249183,174185,28187,189155,259940,67800,60065,340172,311667,354861,182,305523,115981,365082,213915,47894,131301,198754,358852,112496,404423,486725,233123,322936,325337,125932,299260,128791,295663,469897,120580,347766,34859,364204,37597,268974,351155,256955,214013,309192,412394,216800,30315,411242,16678,233247,359013,401666,30792,452394,408649,14159,5519,91705,227648,120966,319599,351170,68129,368701,233566,144256,156172,41972,499687,390955,6549,298079,498230,196397,239493,242037,');
show warnings;
IN方法:
call cfzxin('38068,238833,308799,274344,299910,309823,337486,290812,56393,64413,492511,260426,58626,450987,499220,187731,365665,212799,227788,255724,384636,465766,417310,313148,483987,328761,402876,237274,249183,174185,28187,189155,259940,67800,60065,340172,311667,354861,182,305523,115981,365082,213915,47894,131301,198754,358852,112496,404423,486725,233123,322936,325337,125932,299260,128791,295663,469897,120580,347766,34859,364204,37597,268974,351155,256955,214013,309192,412394,216800,30315,411242,16678,233247,359013,401666,30792,452394,408649,14159,5519,91705,227648,120966,319599,351170,68129,368701,233566,144256,156172,41972,499687,390955,6549,298079,498230,196397,239493,242037');
show warnings;
Total Time : 00:00:18:234
DELIMITER $$
DROP PROCEDURE IF EXISTS `zz`.`cfzxin`$$
CREATE DEFINER=`root`@`localhost` PROCEDURE `cfzxin`(i varchar(8000))
BEGIN
DECLARE ai INT DEFAULT 1;
DROP TABLE IF EXISTS `t_06`;
create table t_06 (
id int not null primary key,
c1 varchar(30),
i2 int
) engine = myisam;
WHILE ai < 500000 DO
insert into t_06 values (ai,concat('AA',ai),ai);
SET ai = ai + 1;
END WHILE;
set @qq=concat('select * from t_06 where id in(',i,')');
select @qq;
prepare dd from @qq;
execute dd;
END$$
DELIMITER ;
用临时表方法,去掉生成LSB2(即生成1-1000数字的代码),加入主键
Total Time : 00:00:18:438
DELIMITER $$
DROP PROCEDURE IF EXISTS `zz`.`cfzfnewa`$$
CREATE DEFINER=`root`@`localhost` PROCEDURE `cfzfnewa`(i varchar(5000))
BEGIN
DECLARE ai INT DEFAULT 1;
DROP TABLE IF EXISTS `t_06`;
create table t_06 (
id int not null primary key,
c1 varchar(30),
i2 int
) engine = myisam;
WHILE ai < 500000 DO
insert into t_06 values (ai,concat('AA',ai),ai);
SET ai = ai + 1;
END WHILE;
DROP TABLE IF EXISTS `lsbtest`;
CREATE TABLE `lsbtest` (
`id1` bigint(10) primary key
) ENGINE=InnoDB DEFAULT CHARSET=utf8;
insert into lsbtest
select
mid(mid(i,id,length(i)),2,LOCATE(',',mid(i,id,length(i)),2)-2)
from lsb2 where length(i)>=id and mid(i,id,1)=','
and length(mid(i,id,length(i)))>=2;
select a.* from t_06 a inner join lsbtest b on a.id=b.id1;
END$$
DELIMITER ;
与IN方法时间相差不多。
妙!
我的看法是:如果前面要查询的源表的数据量很大时,则把in这部分放进临时表并建立索引来关联的速度是最快的(当然源查询表对应字段也有索引),如果要把 查询动态化的话,我的看法跟WWWWA 一样,就是用存储过程的方式,把in那部分的值清单作为参数传进去,若in的清单比较多,则在存储过程内部的最后 对临时表进行人工处理(先truncate,再drop)。
其实这种需求,关键还要看源查询表的数据量及in的值清单的数据量。
其实在特大的高压下,你会发现IN的效率很高!
最近在帮客户解决了一个问题,就是把自己内部的临时表直接转成IN了。能够承受的压力更大!
从8楼,9楼的试验结果上,似乎还看不出 mysql 用临时表做 join 会比 直接 IN 有效率上的提高。
各位可以在其它数据库上试试
个人总结一下以上的方法有:
1。 直接使用 IN (38068,238833,308799 ... )
2。 将 (38068,238833,308799 ... ) 放入临时表,然后用 JOIN
3。 直接在程序中 执行多个 select * where id = 38068; select * from where id=238833; ...
4。 使用 inner join (select 38068 union all select 238833 union all ... )
关注此帖子,也遇到这个问题
目前是直接用的in的,源数据表有20万数据,in后面的数字有一百多个,感觉有些吃不清,建临时表并join没有试过
1 觉得楼主要这种测试比较意义不太大,不同数据分布查询速度会有一点点不同。
2 在MySQL 中,应该只有 in 和 join 表之间比较。
3 join 表比较多的时候会比 in 慢,因为 in 是在 buffer 中寻找对应内容;
直接使用 【fill_record】函数就完成;join 是需要创建临时表的额外消耗。
那要是同时查询的个数多呢 像这样
。。。。。。。。。。。。。。。。in(1,2,3,4,5,6,7,8,9,10);
。。。。。。。。。。。。。。。。in(12,22,32,42,52,62,72,82,92,102);
。。。。。。。。。。。。。。。。in(11,21,31,41,51,61,7,81,9,10);
。。。。。。。。。。。。。。。。in(13,2,3,4,53,6,7,8,9,10);
。。。。。。。。。。。。。。。。in(1,2,32,4,5,6,74,84,9,10);
。。。。。。。。。。。。。。。。in(15,25,3,4,5,6,7,8,69,10);
我这儿使用后的结果,效率很慢,慢慢的jsp页面就卡死了
这个问题怎么解决 用什么可以替换in