Mysql 在使用索引条件自行加入 table 时不要使用索引

Question

当自加入一个 table 并将条件作为索引列时，mysql 将进行完整的 table 扫描。通过更改连接顺序，我们可以使用索引。

Table EmployeeDetails 架构

EmployeeDetails | CREATE TABLE `EmployeeDetails` (
  `CompanyId` bigint(19) NOT NULL DEFAULT '0',
  `EmployeeId` bigint(19) NOT NULL DEFAULT '0',
  `ParentEmployeeId` bigint(19) DEFAULT NULL,
  PRIMARY KEY (`EmployeeId`),
  KEY `EmployeeDetails_FK1_IDX` (`CompanyId`),
  KEY `EmployeeDetails_FK2_IDX` (`ParentEmployeeId`),
  CONSTRAINT `EmployeeDetails_FK1` FOREIGN KEY (`CompanyId`) REFERENCES `CompanyDetails` (`CompanyId`) ON DELETE CASCADE,
  CONSTRAINT `EmployeeDetails_FK2` FOREIGN KEY (`ParentEmployeeId`) REFERENCES `EmployeeDetails` (`EmployeeId`) ON DELETE CASCADE,
) ENGINE=InnoDB DEFAULT CHARSET=utf8 |

Table CompanyDetails 架构

| CompanyDetails | CREATE TABLE `CompanyDetails` (
  `CompanyId` bigint(19) NOT NULL DEFAULT '0',
  `NAME` varchar(25) NOT NULL DEFAULT '',
  `DESCRIPTION` varchar(255) DEFAULT NULL,
  PRIMARY KEY (`CompanyId`),
) ENGINE=InnoDB DEFAULT CHARSET=utf8 
|

第一次查询 - 首先 select 将扫描 EmployeeDetails table 中的所有记录并进行自连接（不会使用标准中使用的索引 KEY EmployeeDetails_FK1_IDX (CompanyId)）

mysql> select * from EmployeeDetails INNER JOIN `EmployeeDetails` `t1` ON `EmployeeDetails`.`EmployeeId`=`t1`.`ParentEmployeeId` where `EmployeeDetails`.`CompanyId` IN(123);
Empty set (0.10 sec)

mysql> explain select * from EmployeeDetails INNER JOIN `EmployeeDetails` `t1` ON `EmployeeDetails`.`EmployeeId`=`t1`.`ParentEmployeeId` where `EmployeeDetails`.`CompanyId` IN(123)\G
*************************** 1. row ***************************
           id: 1
  select_type: SIMPLE
        table: t1
   partitions: NULL
         type: ALL
possible_keys: EmployeeDetails_FK2_IDX
          key: NULL
      key_len: NULL
          ref: NULL
         rows: 7999
     filtered: 100.00
        Extra: Using where
*************************** 2. row ***************************
           id: 1
  select_type: SIMPLE
        table: EmployeeDetails
   partitions: NULL
         type: eq_ref
possible_keys: PRIMARY,EmployeeDetails_FK1_IDX
          key: PRIMARY
      key_len: 8
          ref: db.t1.ParentEmployeeId
         rows: 1
     filtered: 5.00
        Extra: Using where
2 rows in set, 1 warning (0.01 sec)

第二个查询 - 第一个 select 使用 EmployeeDetails_FK1_IDX(CompanyId`) 索引的 EmployeeDetails 记录，仅扫描符合条件的记录并进行自连接

mysql> select * from EmployeeDetails INNER JOIN `EmployeeDetails` `t1` ON `t1`.`EmployeeId`=`EmployeeDetails`.`ParentEmployeeId` where `EmployeeDetails`.`CompanyId` IN(123);
Empty set (0.00 sec)

mysql> explain select * from EmployeeDetails INNER JOIN `EmployeeDetails` `t1` ON `t1`.`EmployeeId`=`EmployeeDetails`.`ParentEmployeeId` where `EmployeeDetails`.`CompanyId` IN(123)\G
*************************** 1. row ***************************
           id: 1
  select_type: SIMPLE
        table: EmployeeDetails
   partitions: NULL
         type: ref
possible_keys: EmployeeDetails_FK1_IDX,EmployeeDetails_FK2_IDX
          key: EmployeeDetails_FK1_IDX
      key_len: 8
          ref: const
         rows: 54
     filtered: 100.00
        Extra: Using where
*************************** 2. row ***************************
           id: 1
  select_type: SIMPLE
        table: t1
   partitions: NULL
         type: eq_ref
possible_keys: PRIMARY
          key: PRIMARY
      key_len: 8
          ref: db.EmployeeDetails.ParentEmployeeId
         rows: 1
     filtered: 100.00
        Extra: NULL
2 rows in set, 1 warning (0.00 sec)

为什么会这样？

Answer 1

从 EXPLAIN 中可以看出，"EmployeeDetails_FK2_IDX" 被列为可能的键。如果 MySQL 优化器不使用它，那只是它不会减少查询时间。

尝试用数据填充 table（我检查过它并且它按原样工作），您应该看到 MySQL 优化器将使用索引进行连接而不查看它们的订单。因为连接顺序对 MySQL 优化器算法没有任何影响。

同时勾选 STRAIGHT_JOIN

Mysql 在使用索引条件自行加入 table 时不要使用索引

Mysql don't use index while self joining a table with indexed criteria

mysql

mysql-5.6

mysql-5.7