为什么使用 MySQL 时带有“exists”的 sql 运行速度比“in”慢

我是 MySQL 优化的新手，我发现了一个惊人的事情：使用 'exists' 的 sql 比使用 'in' 运行得慢！

以下是我的 DDL：

mysql> `show create table order\G`;
*************************** 1. row ***************************
       Table: order
Create Table: CREATE TABLE `order` (
  `id` int(4) NOT NULL AUTO_INCREMENT,
  `employee_id` int(4) NOT NULL,
  `price` decimal(7,2) NOT NULL,
  `trade_date` timestamp NOT NULL DEFAULT CURRENT_TIMESTAMP,
  PRIMARY KEY (`id`),
  KEY `fk_employee_id` (`employee_id`),
  CONSTRAINT `fk_employee_id` FOREIGN KEY (`employee_id`) REFERENCES `employee` (`id`)
) ENGINE=InnoDB AUTO_INCREMENT=231001 DEFAULT CHARSET=utf8



mysql> `show create table order_detail\G`;
*************************** 1. row ***************************
       Table: order_detail
Create Table: CREATE TABLE `order_detail` (
  `id` int(4) NOT NULL AUTO_INCREMENT,
  `menu_id` int(4) NOT NULL,
  `order_id` int(4) NOT NULL,
  `amount` int(4) NOT NULL,
  PRIMARY KEY (`id`),
  KEY `fk_menu_id` (`menu_id`),
  KEY `fk_order_id` (`order_id`),
  CONSTRAINT `fk_menu_id` FOREIGN KEY (`menu_id`) REFERENCES `menu` (`id`),
  CONSTRAINT `fk_order_id` FOREIGN KEY (`order_id`) REFERENCES `order` (`id`)
) ENGINE=InnoDB AUTO_INCREMENT=1037992 DEFAULT CHARSET=utf8



Query Solution 1: use exists
---------------------------------

mysql> `SELECT count(`order`.id) FROM `order` WHERE EXISTS ( SELECT 1 FROM order_detail WHERE order_detail.order_id = `order`.id GROUP BY (order_detail.order_id) HAVING COUNT(order_id) > 5 );`
+-------------------+
| count(`order`.id) |
+-------------------+
|             92054 |
+-------------------+
1 row in set (***6.53 sec***)

mysql> `explain SELECT count(`order`.id) FROM `order` WHERE EXISTS ( SELECT 1 FROM order_detail WHERE order_detail.order_id = `order`.id GROUP BY (order_detail.order_id) HAVING COUNT(order_id) > 5 )\G;`
*************************** 1. row ***************************
           id: 1
  select_type: PRIMARY
        table: order
         type: index
possible_keys: NULL
          key: fk_employee_id
      key_len: 4
          ref: NULL
         rows: 231032
        Extra: Using where; Using index
*************************** 2. row ***************************
           id: 2
  select_type: DEPENDENT SUBQUERY
        table: order_detail
         type: ref
possible_keys: fk_order_id
          key: fk_order_id
      key_len: 4
          ref: performance_test.order.id
         rows: 3
        Extra: Using where; Using index
2 rows in set (0.01 sec)


Query solution 2: use in
------------------------

mysql> `SELECT count(`order`.id) FROM `order` WHERE `order`.id IN ( SELECT order_detail.order_id FROM order_detail GROUP BY (order_detail.order_id) HAVING COUNT(order_id) > 5 ) ;`
+-------------------+
| count(`order`.id) |
+-------------------+
|             92054 |
+-------------------+
1 row in set (***3.88 sec***)

mysql> `explain SELECT count(`order`.id) FROM `order` WHERE `order`.id IN ( SELECT order_detail.order_id FROM order_detail GROUP BY (order_detail.order_id) HAVING COUNT(order_id) > 5 ) \G;`<br>
*************************** 1. row ***************************
           id: 1
  select_type: PRIMARY
        table: order
         type: index
possible_keys: NULL
          key: fk_employee_id
      key_len: 4
          ref: NULL
         rows: 231032
        Extra: Using where; Using index
*************************** 2. row ***************************
           id: 2
  select_type: SUBQUERY
        table: order_detail
         type: index
possible_keys: fk_order_id
          key: fk_order_id
      key_len: 4
          ref: NULL
         rows: 1036314
        Extra: Using index
2 rows in set (0.00 sec)

我认为你有点困惑，你有一个错误的想法，“EXISTS”比“IN”工作得更快，我试图让你理解原因。

EXISTS 返回一个布尔值，并且将在第一个匹配时返回一个布尔值。因此，如果您正在处理重复项/多个项，则根据数据和需求，“EXISTS”的执行速度将比“IN”或“JOIN”更快。

而“IN”是 OR 子句的语法糖。虽然它非常方便，但处理比较的大量值（1,000以北）时存在问题。在重复/多个的情况下，“IN”会检查所有存在的值，这自然会比“EXISTS”消耗更多的执行时间，这就是为什么“IN”总是比“EXISTS”慢的原因。

我希望我能澄清你的困惑。