以前已经简单介绍了MySQL的优化步骤,那么接下来天然而是就是经常使用的SQL优化,好比inseer、group by等经常使用SQL的优化,会涉及SQL语句内部细节(这正是我缺少的)。最后但愿本身能记录完成的一套MySQL优化博文!mysql
注:其中部分我并无所有实验(并不表明是错的),这里只至关于记录下,接下来会慢慢补充!sql
参考资料:《深刻浅出MySQL》(有须要PDF电子书的伙伴能够评论或者私信我)数据库
(1)对于MyISAM存储引擎的表,可使用:DISABLE KEYS 和 ENABLE KEYS 用来打开或者关闭 MyISAM 表非惟一索引的更新。工具
ALTER TABLE tbl_name DISABLE KEYS; loading the data ALTER TABLE tbl_name ENABLE KEYS;
(2)对于InnoDB引擎,有如下几种优化措施:性能
① 导入的数据按照主键的顺序保存:这是由于InnoDB引擎表示按照主键顺序保存的,若是能将插入的数据提早按照排序好天然能省去不少时间。测试
好比bulk_insert.txt文件是以表user主键的顺序存储的,导入的时间为15.23秒优化
mysql> load data infile 'mysql/bulk_insert.txt' into table user; Query OK, 126732 rows affected (15.23 sec) Records: 126732 Deleted: 0 Skipped: 0 Warnings: 0
没有按照主键排序的话,时间为:26.54秒ui
mysql> load data infile 'mysql/bulk_insert.txt' into table user; Query OK, 126732 rows affected (26.54 sec) Records: 126732 Deleted: 0 Skipped: 0 Warnings: 0
② 导入数据前执行SET UNIQUE_CHECKS=0,关闭惟一性校验,带导入以后再打开设置为1:校验会消耗时间,在数据量大的状况下须要考虑。spa
③ 导入前设置SET AUTOCOMMIT=0,关闭自动提交,导入后结束再设置为1:这是由于自动提交会消耗部分时间与资源,虽然消耗不是很大,可是在数据量大的状况下仍是得考虑。code
(1)尽可能使用多个值表的 INSERT 语句,这种方式将大大缩减客户端与数据库之间的链接、关闭等消耗。(同一客户的状况下),即:
INSERT INTO tablename values(1,2),(1,3),(1,4)
实验:插入8条数据到user表中(使用navicat客户端工具)
insert into user values(1,'test',replace(uuid(),'-','')); insert into user values(2,'test',replace(uuid(),'-','')); insert into user values(3,'test',replace(uuid(),'-','')); insert into user values(4,'test',replace(uuid(),'-','')); insert into user values(5,'test',replace(uuid(),'-','')); insert into user values(6,'test',replace(uuid(),'-','')); insert into user values(7,'test',replace(uuid(),'-','')); insert into user values(8,'test',replace(uuid(),'-',''));
获得反馈:
[SQL] insert into user values(1,'test',replace(uuid(),'-','')); 受影响的行: 1 时间: 0.033s [SQL] insert into user values(2,'test',replace(uuid(),'-','')); 受影响的行: 1 时间: 0.034s [SQL] insert into user values(3,'test',replace(uuid(),'-','')); 受影响的行: 1 时间: 0.056s [SQL] insert into user values(4,'test',replace(uuid(),'-','')); 受影响的行: 1 时间: 0.008s [SQL] insert into user values(5,'test',replace(uuid(),'-','')); 受影响的行: 1 时间: 0.008s [SQL] insert into user values(6,'test',replace(uuid(),'-','')); 受影响的行: 1 时间: 0.024s [SQL] insert into user values(7,'test',replace(uuid(),'-','')); 受影响的行: 1 时间: 0.004s [SQL] insert into user values(8,'test',replace(uuid(),'-','')); 受影响的行: 1 时间: 0.004s
总共的时间为0.171秒,接下来使用多值表形式:
insert into user values (9,'test',replace(uuid(),'-','')), (10,'test',replace(uuid(),'-','')), (11,'test',replace(uuid(),'-','')), (12,'test',replace(uuid(),'-','')), (13,'test',replace(uuid(),'-','')), (14,'test',replace(uuid(),'-','')), (15,'test',replace(uuid(),'-','')), (16,'test',replace(uuid(),'-',''));
获得反馈:
[SQL] insert into user values (9,'test',replace(uuid(),'-','')), (10,'test',replace(uuid(),'-','')), (11,'test',replace(uuid(),'-','')), (12,'test',replace(uuid(),'-','')), (13,'test',replace(uuid(),'-','')), (14,'test',replace(uuid(),'-','')), (15,'test',replace(uuid(),'-','')), (16,'test',replace(uuid(),'-','')); 受影响的行: 8 时间: 0.038s
获得时间为0.038,这样一来能够很明显节约时间优化SQL
(2)若是在不一样客户端插入不少行,可以使用INSERT DELAYED语句获得更高的速度,DELLAYED含义是让INSERT语句立刻执行,其实数据都被放在内存的队列中。并无真正写入磁盘。LOW_PRIORITY恰好相反。
(3)将索引文件和数据文件分在不一样的磁盘上存放(InnoDB引擎是在同一个表空间的)。
(4)若是批量插入,则能够增长bluk_insert_buffer_size变量值提供速度(只对MyISAM有用)
(5)当从一个文本文件装载一个表时,使用LOAD DATA INFILE,一般比INSERT语句快20倍。
在默认状况下,MySQL中的GROUP BY语句会对其后出现的字段进行默认排序(非主键状况),就比如咱们使用ORDER BY col1,col2,col3...因此咱们在后面跟上具备相同列(与GROUP BY后出现的col1,col2,col3...相同)ORDER BY子句并无影响该SQL的实际执行性能。
那么就会有这样的状况出现,咱们对查询到的结果是否已经排序不在意时,可使用ORDER BY NULL禁止排序达到优化目的。下面使用EXPLAIN命令分析SQL。
在user_1中执行select id, sum(money) form user_1 group by name时,会默认排序(注意group by后的column是非index才会体现group by的排序,若是是primary key,那以前说过了InnoDB默认是按照主键index排好序的)
mysql> select*from user_1; +----+----------+-------+ | id | name | money | +----+----------+-------+ | 1 | Zhangsan | 32 | | 2 | Lisi | 65 | | 3 | Wangwu | 44 | | 4 | Lijian | 100 | +----+----------+-------+ 4 rows in set
不由止排序,即不使用ORDER BY NULL时:有明显的Using filesort。
当使用ORDER BY NULL禁止排序后,Using filesort不存在
MySQL可使用一个索引来知足ORDER BY 子句的排序,而不须要额外的排序,可是须要知足如下几个条件:
(1)WHERE 条件和OREDR BY 使用相同的索引:即key_part1与key_part2是复合索引,where中使用复合索引中的key_part1
SELECT*FROM user WHERE key_part1=1 ORDER BY key_part1 DESC, key_part2 DESC;
(2)并且ORDER BY顺序和索引顺序相同:
SELECT*FROM user ORDER BY key_part1, key_part2;
(3)而且要么都是升序要么都是降序:
SELECT*FROM user ORDER BY key_part1 DESC, key_part2 DESC;
但如下几种状况则不使用索引:
(1)ORDER BY中混合ASC和DESC:
SELECT*FROM user ORDER BY key_part1 DESC, key_part2 ASC;
(2)查询行的关键字与ORDER BY所使用的不相同,即WHERE 后的字段与ORDER BY 后的字段是不同的
SELECT*FROM user WHERE key2 = ‘xxx’ ORDER BY key1;
(3)ORDER BY对不一样的关键字使用,即ORDER BY后的关键字不相同
SELECT*FROM user ORDER BY key1, key2;
当MySQL使用OR查询时,若是要利用索引的话,必须每一个条件列都使独立索引,而不是复合索引(多列索引),才能保证使用到查询的时候使用到索引。
好比咱们新建一张用户信息表user_info
mysql> select*from user_info; +---------+--------+----------+-----------+ | user_id | idcard | name | address | +---------+--------+----------+-----------+ | 1 | 111111 | Zhangsan | Kunming | | 2 | 222222 | Lisi | Beijing | | 3 | 333333 | Wangwu | Shanghai | | 4 | 444444 | Lijian | Guangzhou | +---------+--------+----------+-----------+ 4 rows in set
以后建立ind_name_id(user_id, name)复合索引、id_index(id_index)独立索引,idcard主键索引三个索引。
mysql> show index from user_info; +-----------+------------+-------------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+---------------+ | Table | Non_unique | Key_name | Seq_in_index | Column_name | Collation | Cardinality | Sub_part | Packed | Null | Index_type | Comment | Index_comment | +-----------+------------+-------------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+---------------+ | user_info | 0 | PRIMARY | 1 | idcard | A | 4 | NULL | NULL | | BTREE | | | | user_info | 1 | ind_name_id | 1 | user_id | A | 4 | NULL | NULL | | BTREE | | | | user_info | 1 | ind_name_id | 2 | name | A | 4 | NULL | NULL | YES | BTREE | | | | user_info | 1 | id_index | 1 | user_id | A | 4 | NULL | NULL | | BTREE | | | +-----------+------------+-------------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+---------------+ 4 rows in set
测试一:OR链接两个有单独索引的字段,整个SQL查询才会用到索引(index_merge),而且咱们知道OR其实是把每一个结果最后UNION一块儿的。
mysql> explain select*from user_info where user_id=1 or idcard='222222'; +----+-------------+-----------+------------+-------------+------------------------------+---------------------+---------+------+------+----------+----------------------------------------------------+ | id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra | +----+-------------+-----------+------------+-------------+------------------------------+---------------------+---------+------+------+----------+----------------------------------------------------+ | 1 | SIMPLE | user_info | NULL | index_merge | PRIMARY,ind_name_id,id_index | ind_name_id,PRIMARY | 4,62 | NULL | 2 | 100 | Using sort_union(ind_name_id,PRIMARY); Using where | +----+-------------+-----------+------------+-------------+------------------------------+---------------------+---------+------+------+----------+----------------------------------------------------+ 1 row in set
测试二:OR使用复合索引的字段name,与没有索引的address,整个SQL都是ALL全表扫描的
mysql> explain select*from user_info where name='Zhangsan' or address='Beijing'; +----+-------------+-----------+------------+------+---------------+------+---------+------+------+----------+-------------+ | id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra | +----+-------------+-----------+------------+------+---------------+------+---------+------+------+----------+-------------+ | 1 | SIMPLE | user_info | NULL | ALL | NULL | NULL | NULL | NULL | 4 | 43.75 | Using where | +----+-------------+-----------+------------+------+---------------+------+---------+------+------+----------+-------------+ 1 row in set
交换OR位置而且使用另外的复合索引的列,也是ALL全表扫描:
mysql> explain select*from user_info where address='Beijing' or user_id=1; +----+-------------+-----------+------------+------+----------------------+------+---------+------+------+----------+-------------+ | id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra | +----+-------------+-----------+------------+------+----------------------+------+---------+------+------+----------+-------------+ | 1 | SIMPLE | user_info | NULL | ALL | ind_name_id,id_index | NULL | NULL | NULL | 4 | 43.75 | Using where | +----+-------------+-----------+------------+------+----------------------+------+---------+------+------+----------+-------------+ 1 row in set
使用嵌套查询有时候可使用更有效的JOIN链接代替,这是由于MySQL中不须要在内存中建立临时表完成SELECT子查询与主查询两部分查询工做。可是并非全部的时候都成立,最好是在on关键字后面的列有索引的话,效果会更好!
好比在表major中major_id是有索引的:
select * from student u left join major m on u.major_id=m.major_id where m.major_id is null;
而经过嵌套查询时,在内存中建立临时表完成SELECT子查询与主查询两部分查询工做,会有必定的消耗
select * from student u where major_id not in (select major_id from major);
SQL提示(SQL HINT)是优化数据库的一个重要手段,就是往SQL语句中加入一些人为的提示来达到优化目的。下面是一些经常使用的SQL提示:
(1)USE INDEX:使用USE INDEX是但愿MySQL去参考索引列表,就可让MySQL不须要考虑其余可用索引,其实也就是possible_keys属性下参考的索引值
mysql> explain select* from user_info use index(id_index,ind_name_id) where user_id>0; +----+-------------+-----------+------------+------+----------------------+------+---------+------+------+----------+-------------+ | id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra | +----+-------------+-----------+------------+------+----------------------+------+---------+------+------+----------+-------------+ | 1 | SIMPLE | user_info | NULL | ALL | ind_name_id,id_index | NULL | NULL | NULL | 4 | 100 | Using where | +----+-------------+-----------+------------+------+----------------------+------+---------+------+------+----------+-------------+ 1 row in set mysql> explain select* from user_info use index(id_index) where user_id>0; +----+-------------+-----------+------------+------+---------------+------+---------+------+------+----------+-------------+ | id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra | +----+-------------+-----------+------------+------+---------------+------+---------+------+------+----------+-------------+ | 1 | SIMPLE | user_info | NULL | ALL | id_index | NULL | NULL | NULL | 4 | 100 | Using where | +----+-------------+-----------+------------+------+---------------+------+---------+------+------+----------+-------------+ 1 row in set
(2)IGNORE INDEX忽略索引
咱们使用user_id判断,用不到其余索引时,能够忽略索引。即与USE INDEX相反,从possible_keys中减去不须要的索引,可是实际环境中不多使用。
mysql> explain select* from user_info ignore index(primary,ind_name_id,id_index) where user_id>0; +----+-------------+-----------+------------+------+---------------+------+---------+------+------+----------+-------------+ | id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra | +----+-------------+-----------+------------+------+---------------+------+---------+------+------+----------+-------------+ | 1 | SIMPLE | user_info | NULL | ALL | NULL | NULL | NULL | NULL | 4 | 33.33 | Using where | +----+-------------+-----------+------------+------+---------------+------+---------+------+------+----------+-------------+ 1 row in set
(3)FORCE INDEX强制索引
好比where user_id > 0,可是user_id在表中都是大于0的,天然就会进行ALL全表搜索,可是使用FORCE INDEX虽然执行效率不是最高(where user_id > 0条件决定的)但MySQL仍是使用索引。
mysql> explain select* from user_info where user_id>0; +----+-------------+-----------+------------+------+----------------------+------+---------+------+------+----------+-------------+ | id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra | +----+-------------+-----------+------------+------+----------------------+------+---------+------+------+----------+-------------+ | 1 | SIMPLE | user_info | NULL | ALL | ind_name_id,id_index | NULL | NULL | NULL | 4 | 100 | Using where | +----+-------------+-----------+------------+------+----------------------+------+---------+------+------+----------+-------------+ 1 row in set
以后强制使用独立索引id_index(user_id):
mysql> explain select* from user_info force index(id_index) where user_id>0; +----+-------------+-----------+------------+-------+---------------+----------+---------+------+------+----------+-----------------------+ | id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | filtered | Extra | +----+-------------+-----------+------------+-------+---------------+----------+---------+------+------+----------+-----------------------+ | 1 | SIMPLE | user_info | NULL | range | id_index | id_index | 4 | NULL | 4 | 100 | Using index condition | +----+-------------+-----------+------------+-------+---------------+----------+---------+------+------+----------+-----------------------+ 1 row in set
(1)不少时候数据库的性能是因为不合适(是指效率不高,可能会致使锁表等)的SQL语句形成,本篇博文只是介绍简单的SQL优化
(2)其中有些优化在真正开发中是用不到的,可是一旦出问题性能降低的时候须要去一一分析。