MySQL count(*)之索引选择

覆盖索引对于一些统计问题，如下：

MySQL > show create table test1 \G*************************** 1. row ***************************       Table: test1Create Table: CREATE TABLE `test1` (  `id` bigint(16) NOT NULL AUTO_INCREMENT,  `order_seq` bigint(16) NOT NULL,  PRIMARY KEY (`id`),  KEY `idx_id` (`id`),  KEY `idx_id_ordseq` (`id`,`order_seq`)) ENGINE=InnoDB AUTO_INCREMENT=15002212 DEFAULT CHARSET=utf8mb41 row in set (0.00 sec)MySQL > explain select count(*) from test1 where id>10000 and id<20000;+----+-------------+-------+------------+-------+------------------------------+--------+---------+------+------+----------+--------------------------+| id | select_type | table | partitions | type  | possible_keys                | key    | key_len | ref  | rows | filtered | Extra                    |+----+-------------+-------+------------+-------+------------------------------+--------+---------+------+------+----------+--------------------------+|  1 | SIMPLE      | test1 | NULL       | range | PRIMARY,idx_id,idx_id_ordseq | idx_id | 8       | NULL | 9999 |   100.00 | Using where; Using index |+----+-------------+-------+------------+-------+------------------------------+--------+---------+------+------+----------+--------------------------+1 row in set, 1 warning (0.00 sec)

innodb存储引擎选择了id字段的辅助索引，而不是聚集索引来统计，更不是联合索引。原因是辅助索引远小于聚集索引，选择辅助索引可以减少IO资源消耗。

而另外一个统计场景：

select count(*) from test1 where order_seq > 1502131212577 and order_seq< 202007080947244761;

test1表建有id和 order_seq 字段的联合索引。

MySQL > show create table test1 \G*************************** 1. row ***************************       Table: test1Create Table: CREATE TABLE `test1` (  `id` bigint(16) NOT NULL AUTO_INCREMENT,  `order_seq` bigint(16) NOT NULL,  PRIMARY KEY (`id`),  KEY `idx_id` (`id`),  KEY `idx_id_ordseq` (`id`,`order_seq`)) ENGINE=InnoDB AUTO_INCREMENT=15002212 DEFAULT CHARSET=utf8mb41 row in set (0.00 sec)MySQL > explain select count(*) from test1 where order_seq > 1502131212577 and order_seq< 202007080947244761;+----+-------------+-------+------------+-------+---------------+---------------+---------+------+----------+----------+--------------------------+| id | select_type | table | partitions | type  | possible_keys | key           | key_len | ref  | rows     | filtered | Extra                    |+----+-------------+-------+------------+-------+---------------+---------------+---------+------+----------+----------+--------------------------+|  1 | SIMPLE      | test1 | NULL       | index | NULL          | idx_id_ordseq | 16      | NULL | 15068082 |    11.11 | Using where; Using index |+----+-------------+-------+------------+-------+---------------+---------------+---------+------+----------+----------+--------------------------+1 row in set, 1 warning (0.00 sec)

这里使用条件 order_seq 查询，一般情况下使用不了联合索引的，但是这个案例中的查询，利用到覆盖索引的信息。possible_keys依然为null，但是key是idx_id_ordseq，extra里出现Using index，表示为覆盖索引。