Mysql 范围查询优化

  Range查询:用单独的Index的一个或多个index值来检索表的子集行数据,当然包含多个index。

  1:一个index (单一部分)的range access 方法:(eg : 指的这种key (column1 ))

  单独的index,index值间隔可以方便的由对应的where子句的条件表示,所有我们称值为range条件而不是间隔;

  单独index的range条件的定义:

     1.1:对于btree和hash索引,index和一个常量值通过 =, <=>,in(),is null,或者 IS not null操作符做比较;

     1.2:另外,对于Btree索引,index和一个常量值通过 <,>,<=,>=,between,!=,或者<>操作符做比较;

     1.3: 对于所有类型的index,多范围条件通过 or and关键字组合形式;

 

  '常量值'在之前的描述中意味着:

    2.1: 查询字符串的常量形式;

    2.2: const 或者system表的一列(也只有一列)的自连接(join);

    2.3: 不相关子查询的结果;

    2.4: 有上面类型子表达式完全组成的任意表达式;

 

   where子句范围查找的例子:

SELECT * FROM t1
  WHERE key_col > 1
  AND key_col < 10;

SELECT * FROM t1
  WHERE key_col = 1
  OR key_col IN (15,18,20);

SELECT * FROM t1
  WHERE key_col LIKE 'ab%'
  OR key_col BETWEEN 'bar' AND 'foo';

  一些非常量值可能在传播阶段转换为常量;

  Mysql尝试对于在where子句中的任何可能的index提取range condition,在提取过程中,不能构造的范围条件被舍去,产生重叠的条件被组合,产生空范围的条件被舍去。

  如下:

SELECT * FROM t1 WHERE
  (key1 < 'abc' AND (key1 LIKE 'abcde%' OR key1 LIKE '%b')) OR
  (key1 < 'bar' AND nonkey = 4) OR
  (key1 < 'uux' AND key1 > 'z');

key1为index, nonkey非index;

key1的提取过程如下:

  1:从原始的where子句开始:

(key1 < 'abc' AND (key1 LIKE 'abcde%' OR key1 LIKE '%b')) OR
(key1 < 'bar' AND nonkey = 4) OR
(key1 < 'uux' AND key1 > 'z')

 2:舍去nokey = 4和key1 like ‘%d’,因为他们不能用作范围scan,正确的舍去方法是用true代替她们,

(key1 < 'abc' AND (key1 LIKE 'abcde%' OR TRUE)) OR
(key1 < 'bar' AND TRUE) OR
(key1 < 'uux' AND key1 > 'z')

3:  以下条件总是true or false;

   1:(key1 LIKE 'abcde%' OR TRUE) is always true

   2:(key1 < 'uux' AND key1 > 'z') is always false

     

 (key1 < 'abc' AND TRUE) OR (key1 < 'bar' AND TRUE) OR (FALSE)

 4: 去掉不必要的true 和false条件:

(key1 < 'abc') OR (key1 < 'bar')

 5:组合那些重叠的区间成一个:

(key1 < 'bar')

 

   范围条件的提取算法可以处理内嵌的and /or任意深度的构造,他的数去不依赖与他们出现在where子句中的顺序;

   

2:多part index的范围查询:

   index的多个部分的范围条件是上面的扩展,被一个或几个key元组来限制条件:

  eg: 一个联合索引定义:key1(key_part1,key_part2,key_part3),key元组以key order显示如下:

key_part1  key_part2  key_part3
  NULL       1          'abc'
  NULL       1          'xyz'
  NULL       2          'foo'
   1         1          'abc'
   1         1          'xyz'
   1         2          'abc'
   2         1          'aaa'

 

     条件 key_part1 = 1 定义的区间:

(1,-inf,-inf) <= (key_part1,key_part2,key_part3) < (1,+inf,+inf)

    这区间包括第四,第五,第六元组。

    相对,条件 key_part3 = 'abc'不能定义一个单独的区间并且不能被区间scan方法使用(最左前缀index);

   

 3:多值的区间优化:

   以下col_name 是index 列:

col_name IN(val1, ..., valN)
col_name = val1 OR ... OR col_name = valN

    如果col_name 等于这些值中的任意一个返回true,

   1:如果col_name 为unique index,则范围的行评估为1,因为最多一个值只能对应一行;

    2:否则,优化器估计范围行数得使用index潜入和index的统计信息;

    index潜入,优化器植入一个潜入在每个range的结尾并且用该范围的行数来估计,eg:col_name IN (10, 20, 30) ,优化器植入两个潜入在每个区间中来估计行数;index潜入提供精确的行估计,但是当表达式中的比较值越多,每个潜入对需要话费更多时间。用index统计信息缺乏精确但是较快。

30727 rows in set (5.37 sec)

mysql> explain select * from employees where first_name like 'm%' ;
+----+-------------+-----------+------+---------------+------+---------+------+--------+-------------+
| id | select_type | table     | type | possible_keys | key  | key_len | ref  | rows   | Extra       |
+----+-------------+-----------+------+---------------+------+---------+------+--------+-------------+
|  1 | SIMPLE      | employees | ALL  | idx_fn_ln     | NULL | NULL    | NULL | 299290 | Using where |
+----+-------------+-----------+------+---------------+------+---------+------+--------+-------------+
1 row in set (0.00 sec)

mysql> explain select * from employees where first_name like 'mart%' ;
+----+-------------+-----------+-------+---------------+-----------+---------+------+------+-----------------------+
| id | select_type | table     | type  | possible_keys | key       | key_len | ref  | rows | Extra                 |
+----+-------------+-----------+-------+---------------+-----------+---------+------+------+-----------------------+
|  1 | SIMPLE      | employees | range | idx_fn_ln     | idx_fn_ln | 16      | NULL | 1635 | Using index condition |
+----+-------------+-----------+-------+---------------+-----------+---------+------+------+-----------------------+
1 row in set (0.00 sec)

第一个条件查找从执行计划上看走的扫表(all),where条件得到的结果为所以数据行的1/10;第二条增加like条件的匹配精度,走的ICP range查找,得到的数据行大约是1/30,mysql

会根据查找的数据范围(多少)决定走index range查找还是直接扫表(all)

mysql> desc select * from employees where first_name like 'ma%' and last_name like 'he%';
+----+-------------+-----------+-------+---------------+-----------+---------+------+-------+-----------------------+
| id | select_type | table     | type  | possible_keys | key       | key_len | ref  | rows  | Extra                 |
+----+-------------+-----------+-------+---------------+-----------+---------+------+-------+-----------------------+
|  1 | SIMPLE      | employees | range | idx_fn_ln     | idx_fn_ln | 34      | NULL | 38174 | Using index condition |
+----+-------------+-----------+-------+---------------+-----------+---------+------+-------+-----------------------+
1 row in set (0.04 sec)
mysql> desc select * from employees where first_name like 'ma%' or last_name like 'he%';
+----+-------------+-----------+------+---------------+------+---------+------+--------+-------------+
| id | select_type | table     | type | possible_keys | key  | key_len | ref  | rows   | Extra       |
+----+-------------+-----------+------+---------------+------+---------+------+--------+-------------+
|  1 | SIMPLE      | employees | ALL  | idx_fn_ln     | NULL | NULL    | NULL | 299290 | Using where |
+----+-------------+-----------+------+---------------+------+---------+------+--------+-------------+
1 row in set (0.02 sec)

 

mysql> desc select * from employees where first_name in('mart','mori','moon');
+----+-------------+-----------+-------+---------------+-----------+---------+------+------+-----------------------+
| id | select_type | table     | type  | possible_keys | key       | key_len | ref  | rows | Extra                 |
+----+-------------+-----------+-------+---------------+-----------+---------+------+------+-----------------------+
|  1 | SIMPLE      | employees | range | idx_fn_ln     | idx_fn_ln | 16      | NULL |  710 | Using index condition |
+----+-------------+-----------+-------+---------------+-----------+---------+------+------+-----------------------+
1 row in set (0.02 sec)

 

mysql> explain select * from employees where emp_no = 11;
+----+-------------+-------+------+---------------+------+---------+------+------+-----------------------------------------------------+
| id | select_type | table | type | possible_keys | key  | key_len | ref  | rows | Extra                                               |
+----+-------------+-------+------+---------------+------+---------+------+------+-----------------------------------------------------+
|  1 | SIMPLE      | NULL  | NULL | NULL          | NULL | NULL    | NULL | NULL | Impossible WHERE noticed after reading const tables |
+----+-------------+-------+------+---------------+------+---------+------+------+-----------------------------------------------------+
1 row in set (0.01 sec)

Mysql发现where条件不可能成立 ,返回null

posted @ 2015-05-20 15:34  TomSun*star  阅读(10828)  评论(0编辑  收藏  举报