【MySQL】性能优化之 覆盖索引

原文：http://blog.itpub.net/22664653/viewspace-774667/
 
一个包含查询所需的字段的索引称为 covering index 覆盖索引。MySQL只需要通过索引就可以返回查询所需要的数据，而不必在查到索引之后进行回表操作，减少IO，提供效率。
   当你对一个sql 使用explain statement 查看一个sql的执行计划时，在EXPLAIN的Extra列出现Using Index提示时，就说明该select查询使用了覆盖索引。

【使用场景】
生产过程中遇到的一个例子，且先不讨论 count(字段)还是 count(*)  
root@yang 06:38:34>select count(o.order_id) as cnt from `order` o  WHERE o.settle_type = 2   and o.order_status = 2  and o.buyer_id=1979459339672858;      
 
+------+
| cnt  |
+------+
| 7574 |
+------+
1 row in set (0.22 sec)

root@yang 06:36:38>explain select count(o.order_id) as cnt from `order` o  WHERE o.settle_type = 2   and o.order_status = 2  and o.buyer_id=1979459339672858; 
+----+-------------+-------+-------------+-----------------------+-----------------------+---------+------+------+-----------------------------------------------------+
| id | select_type | table | type        | possible_keys         | key                   | key_len | ref  | rows | Extra                                    
          |
+----+-------------+-------+-------------+-----------------------+-----------------------+---------+------+------+-----------------------------------------------------+
|  1 | SIMPLE      | o     | index_merge | buyer_id,order_status | buyer_id,order_status | 9,1     | NULL | 3852 | Using intersect(buyer_id,order_status); Using
where |
+----+-------------+-------+-------------+-----------------------+-----------------------+---------+------+------+-----------------------------------------------------+
1 row in set (0.00 sec)

上述select语句的执行计划使用了index_merge 索引聚合，整体sql执行花费0.22s 。根据查询条件对表结构进行如下调整：
root@yang 06:33:14>alter table `ordert` add key ind_od_bid_st_ty_oid(`buyer_id`,`order_status`,`settle_type`,`order_id`);
Query OK, 0 rows affected (3.00 sec)
Records: 0  Duplicates: 0  Warnings: 0

root@yang 06:38:50>explain select count(o.order_id) as cnt from `ordert` o  WHERE o.settle_type = 2   and o.order_status = 2  and o.buyer_id=1979459339672858;
+----+-------------+-------+------+----------------------+----------------------+---------+-------------------+-------+--------------------------+
| id  | select_type | table | type | possible_keys        | key                  | key_len | ref               | rows  | Extra                    |
+----+-------------+-------+------+----------------------+----------------------+---------+-------------------+-------+--------------------------+
|  1 | SIMPLE      | o     | ref  | ind_od_bid_st_ty_oid | ind_od_bid_st_ty_oid | 11      | const,const,const | 15242 | Using where; Using index |
+----+-------------+-------+------+----------------------+----------------------+---------+-------------------+-------+--------------------------+
1 row in set (0.00 sec)
执行计划使用了 using index --覆盖索引而且执行时间由0.22s，下降到小于0.01s。
root@yang 06:39:06>select count(o.order_id) as cnt from `ordert` o  WHERE o.settle_type = 2   and o.order_status = 2  and o.buyer_id=1979459339672858;      
 
+------+
| cnt  |
+------+
| 7574 |
+------+
1 row in set (0.00 sec)

【覆盖索引的限制】
遇到以下情况，执行计划不会选择覆盖查询：
1select选择的字段中含有不在索引中的字段 ，也即索引没有覆盖全部的列。
2 where 条件中不能含有对索引进行like的操作。

root@odbsyunying 08:18:15>explain select count(*) as cnt from `ordert` o  WHERE o.settle_type = 2 
 > and o.order_status = 2  
 > and o.buyer_id like '1979459339672858' \G
*************************** 1. row ***************************
                id: 1
  select_type: SIMPLE
           table: o
           type: index
possible_keys: ind_od_bid_st_ty_oid
            key: ind_od_bid_st_ty_oid
      key_len: 19
             ref: NULL
            rows: 767466  ---覆盖索引扫描
           Extra: Using where; Using index
1 row in set (0.00 sec)