由浅入深探究mysql索引结构原理、性能分析与优化（二）

时间  2019-11-07
标签 由浅入深 探究 mysql 索引 结构 原理 性能 分析 优化 栏目 MySQL 繁體版
原文   原文链接

(1.1)能正确的利用索引mysql

l  Where子句表达式顺序是(username)web

  1. mysql> explain select * from one where username='abgvwfnt';  sql

  2. +----+-------------+-------+------+---------------+----------+---------+-------+------+-------------+  性能

  3. | id | select_type | table | type | possible_keys | key      | key_len | ref   |rows | Extra       |  测试

  4. +----+-------------+-------+------+---------------+----------+---------+-------+------+-------------+  优化

  5. |  1 | SIMPLE      | one   | ref  | username      | username | 24      | const |5 | Using where |  编码

  6. +----+-------------+-------+------+---------------+----------+---------+-------+------+-------------+  orm

  7. 1 row in set (0.00 sec)  索引

l  Where子句表达式顺序是(username,password)it

  1. mysql> explain select * from one where username='abgvwfnt' and password='123456';  

  2. +----+-------------+-------+------+---------------+----------+---------+-------------+------+-------------+  

  3. | id | select_type | table | type | possible_keys | key      | key_len | ref | rows | Extra       |  

  4. +----+-------------+-------+------+---------------+----------+---------+-------------+------+-------------+  

  5. |  1 | SIMPLE      | one   | ref  | username      | username | 43      | const,const |    1 | Using where |  

  6. +----+-------------+-------+------+---------------+----------+---------+-------------+------+-------------+  

  7. 1 row in set (0.00 sec)  

l  Where子句表达式顺序是(username,password, last_login)

  1. mysql> explain select * from one where username='abgvwfnt' and password='123456'and last_login='1338251170';  

  2. +----+-------------+-------+------+---------------+----------+---------+-------------------+------+-------------+  

  3. | id | select_type | table | type | possible_keys | key      | key_len | ref| rows | Extra       |  

  4. +----+-------------+-------+------+---------------+----------+---------+-------------------+------+-------------+  

  5. |  1 | SIMPLE   | one   | ref  | username     | username | 83      | const,const,const |    1 | Using where |  

  6. +----+-------------+-------+------+---------------+----------+---------+-------------------+------+-------------+  

  7. 1 row in set (0.00 sec)  

上面能够看出type=ref 是多列索引，key_len分别是2四、4三、83，这说明用到的索引分别是(username), (username,password), (username,password, last_login );row分别是五、一、1检索的数据行都不多，由于这三个查询都按照索引前缀原则，能够利用到索引。

 

(1.2)不能正确的利用索引

l  Where子句表达式顺序是(password, last_login)

  1. mysql> explain select * from one where password='123456'and last_login='1338251170';  

  2. +----+-------------+-------+------+---------------+------+---------+------+-------+-------------+  

  3. | id | select_type | table | type | possible_keys | key  | key_len | ref  | rows| Extra       |  

  4. +----+-------------+-------+------+---------------+------+---------+------+-------+-------------+  

  5. |  1 | SIMPLE      | one   | ALL  | NULL          | NULL | NULL    | NULL | 20146 | Using where |  

  6. +----+-------------+-------+------+---------------+------+---------+------+-------+-------------+  

  7. 1 row in set (0.00 sec)  

l  Where 子句表达式顺序是(last_login)

  1. mysql> explain select * from one where last_login='1338252525';  

  2. +----+-------------+-------+------+---------------+------+---------+------+-------+-------------+  

  3. | id | select_type | table | type | possible_keys | key  | key_len | ref  | rows| Extra       |  

  4. +----+-------------+-------+------+---------------+------+---------+------+-------+-------------+  

  5. |  1 | SIMPLE      | one   | ALL  | NULL          | NULL | NULL    | NULL | 20146 | Using where |  

  6. +----+-------------+-------+------+---------------+------+---------+------+-------+-------------+  

  7. 1 row in set (0.00 sec)  

以上的两条语句都不是以username开始，这样是用不了索引，经过type=all(全表扫描)，key_len=null，rows都很大20146

Ps：one表里只有20003条数据，为何出现20146，这是优化器对表的一个估算值，不精确的。

l  Where 子句表达式虽然顺序是(username,password, last_login)或(username,password)但第一个是有范围’<’、’>’，’<=’，’>=’等出现

  1. mysql> explain select * from one where username>'abgvwfnt' and password ='123456'and last_login='1338251170';  

  2. +----+-------------+-------+------+---------------+------+---------+------+-------+-------------+  

  3. | id | select_type | table | type | possible_keys | key  | key_len | ref  | rows| Extra       |  

  4. +----+-------------+-------+------+---------------+------+---------+------+-------+-------------+  

  5. |  1 | SIMPLE      | one   | ALL  | username      | NULL | NULL    | NULL | 20146 | Using where |  

  6. +----+-------------+-------+------+---------------+------+---------+------+-------+-------------+  

  7. 1 row in set (0.00 sec)  

这个查询很明显是遍历全部表，一个索引都没用到，非第一列出现范围(password列或last_login列)，则能利用索引到首先出现范围的一列，也就是“where username='abgvwfnt' and password >'123456'and last_login='1338251170';”或则“where username='abgvwfnt' and password >'123456'and last_login<'1338251170';”索引长度ref_len=43,索引检索到password列，因此考虑多列索引的时候把那些查询语句用的比较的列放在最后(或非第一位)。

l  断层，便是where顺序(username, last_login)

  1. mysql> explain select * from one where username='abgvwfnt' and last_login='1338252525';  

  2. +----+-------------+-------+------+---------------+----------+---------+-------+------+-------------+  

  3. | id | select_type | table | type | possible_keys | key      | key_len | ref   | rows | Extra       |  

  4. +----+-------------+-------+------+---------------+----------+---------+-------+------+-------------+  

  5. |  1 | SIMPLE   | one   | ref  | username   | username | 24     | const |5 | Using where |  

  6. +----+-------------+-------+------+---------------+----------+---------+-------+------+-------------+  

  7. 1 row in set (0.00 sec)  

注意这里的key_len=24=8*3(8是username的长度，3是utf8编码)，rows=5，和下面一条sql语句搜索出来同样

  1. mysql>  select * from one where username='abgvwfnt';  

  2. +-------+----------+----------+-------+------------+  

  3. | id    | username | password | level | last_login |  

  4. +-------+----------+----------+-------+------------+  

  5. |  3597 | abgvwfnt | 234567   |     0 | 1338251420 |  

  6. |  7693 | abgvwfnt | 456789   |     0 | 1338251717 |  

  7. | 11789 | abgvwfnt | 456789   |     0 | 1338251992 |  

  8. | 15885 | abgvwfnt | 456789   |     0 | 1338252258 |  

  9. | 19981 | abgvwfnt | 456789   |     0 | 1338252525 |  

  10. +-------+----------+----------+-------+------------+  

  11. 5 rows in set (0.00 sec)  

  12.  

  13. mysql>  select * from one where username='abgvwfnt' and last_login='1338252525';  

  14. +-------+----------+----------+-------+------------+  

  15. | id    | username | password | level | last_login |  

  16. +-------+----------+----------+-------+------------+  

  17. | 19981 | abgvwfnt | 456789   |     0 | 1338252525 |  

  18. +-------+----------+----------+-------+------------+  

  19. 1 row in set (0.00 sec)  

这个就是要的返回结果，因此能够知道断层(username,last_login)，这样只用到username索引，把用到索引的数据再从新检查last_login条件，这个相对全表查询来讲仍是有性能上优化，这也是不少sql优化文章中提到的where 范围查询要放在最后(这不绝对，但能够利用一部分索引)

(1.3)若是一个查询where子句中确实不须要password列，那就用“补洞”。

  1. mysql> select distinct(password) from one;  

  2. +----------+  

  3. | password |  

  4. +----------+  

  5. | 234567   |  

  6. | 345678   |  

  7. | 456789   |  

  8. | 123456   |  

  9. +----------+  

  10. 4 rows in set (0.08 sec)

能够看出password列中只有这几个值，固然在现实中不可能密码有这么多同样的，再说数据也可能不断更新，这里只是举例说明补洞的方法

  1. mysql> explain select * from one where username='abgvwfnt' and password in('123456','234567','345678','456789')

  2. and last_login='1338251170';  

  3. +----+-------------+-------+-------+---------------+----------+---------+------+------+-------------+  

  4. | id | select_type | table | type  | possible_keys | key      | key_len | ref  | rows | Extra       |  

  5. +----+-------------+-------+-------+---------------+----------+---------+------+------+-------------+  

  6. |  1 | SIMPLE    | one | range | username    | username| 83      | NULL |4 | Using where |  

  7. +----+-------------+-------+-------+---------------+----------+---------+------+------+-------------+  

  8. 1 row in set (0.00 sec)  

能够看出ref=83 全部的索引都用到了，type=range是由于用了in子句。

这个被“补洞”列中的值应该是有限的，可预知的，如性别，其值只有男和女(加多一个不男不女也无妨)。

“补洞”方法也有瓶颈，当不少列，且须要补洞的相应列(能够多列)的值虽有限但不少(如中国城市)的时候，优化器在优化时组合起来的数量是很大，这样的话就要作好基准测试和性能分析，权衡得失，取得一个合理的优化方法。

(1.4)like

  1. mysql> explain select * from one where username like 'abgvwfnt%';  

  2. +----+-------------+-------+-------+---------------+----------+---------+------+------+-------------+  

  3. | id | select_type | table | type  | possible_keys | key      | key_len | ref  |  

  4. rows | Extra       |  

  5. +----+-------------+-------+-------+---------------+----------+---------+------+------+-------------+  

  6. |  1 | SIMPLE      | one   | range | username      | username | 24      | NULL |  

  7. 5 | Using where |  

  8. +----+-------------+-------+-------+---------------+----------+---------+------+------+-------------+  

  9. 1 row in set (0.00 sec)  

  10. mysql> explain select * from one where username like '%abgvwfnt%';  

  11. +----+-------------+-------+------+---------------+------+---------+------+-------+-------------+  

  12. | id | select_type | table | type | possible_keys | key  | key_len | ref  | rows| Extra       |  

  13. +----+-------------+-------+------+---------------+------+---------+------+-------+-------------+  

  14. |  1 | SIMPLE      | one   | ALL  | NULL          | NULL | NULL    | NULL | 20259 | Using where |  

  15. +----+-------------+-------+------+---------------+------+---------+------+-------+-------------+  

  16. 1 row in set (0.01 sec)  

对比就知道like操做abgvwfnt%能用到索引，%abgvwfnt%用不到

---------------------------------------------------------------------------------------------

相关文章
相关标签/搜索
每日一句
    每一个你不满意的现在，都有一个你没有努力的曾经。
本站公众号
   欢迎关注本站公众号,获取更多信息
联系我们 最近搜索 最新文章 沪ICP备13005482号-10 MyBatis教程 SQL 教程 MySQL教程 Java 教程 Thymeleaf 教程 Hibernate教程 Spring教程 Redis教程