关联字段字符集相同，但是校验规则不一样，导致表关联时做隐式转换，无法用到关联字段上的索引


同类型的case还有 关联字段字符集不一样，导致表关联时做隐式转换，无法用到关联字段上的索引

案例分析如下：

有2个表act_hi_comment，bpm_task_result，建表语句如下

CREATE TABLE `act_hi_comment` (
  `ID_` varchar(64) COLLATE utf8_bin NOT NULL,
  `TYPE_` varchar(255) COLLATE utf8_bin DEFAULT NULL,
  `TIME_` datetime(3) NOT NULL,
  `USER_ID_` varchar(255) COLLATE utf8_bin DEFAULT NULL,
  `TASK_ID_` varchar(64) COLLATE utf8_bin DEFAULT NULL,
  `PROC_INST_ID_` varchar(64) COLLATE utf8_bin DEFAULT NULL,
  `ACTION_` varchar(255) COLLATE utf8_bin DEFAULT NULL,
  `MESSAGE_` varchar(4000) COLLATE utf8_bin DEFAULT NULL,
  `FULL_MSG_` longblob,
  PRIMARY KEY (`ID_`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8 COLLATE=utf8_bin;

CREATE TABLE `bpm_task_result` (
  `id` bigint(20) NOT NULL AUTO_INCREMENT,
  `taskId` varchar(50) DEFAULT NULL COMMENT '任务ID',
  `processId` varchar(50) DEFAULT NULL COMMENT '流程ID',
  `createTime` timestamp NOT NULL DEFAULT CURRENT_TIMESTAMP,
  `state` int(11) NOT NULL COMMENT '1-审批通过 2-驳回修改 3-审批不通过 4-审批作废',
  PRIMARY KEY (`id`),
  UNIQUE KEY `bpm_task_result_taskId_uindex` (`taskId`) USING BTREE
) ENGINE=InnoDB AUTO_INCREMENT=15916 DEFAULT CHARSET=utf8;

有如下SQL，分析执行计划
EXPLAIN SELECT * from act_hi_comment LEFT JOIN bpm_task_result ON act_hi_comment.TASK_ID_ = bpm_task_result.taskId;



使用desc extended，查看附加信息：

mysql> show warnings\G
*************************** 1. row ***************************
  Level: Warning
   Code: 1739
Message: Cannot use ref access on index 'bpm_task_result_taskId_uindex' due to type or collation conversion on field 'taskId'
*************************** 2. row ***************************
  Level: Note
   Code: 1003
Message: /* select#1 */ select `personnel`.`act_hi_comment`.`ID_` AS `ID_`,`personnel`.`act_hi_comment`.`TYPE_` AS `TYPE_`,`personnel`.`act_hi_comment`.`TIME_` AS `TIME_`,`personnel`.`act_hi_comment`.`USER_ID_` AS `USER_ID_`,`personnel`.`act_hi_comment`.`TASK_ID_`
AS `TASK_ID_`,`personnel`.`act_hi_comment`.`PROC_INST_ID_` AS `PROC_INST_ID_`,`personnel`.`act_hi_comment`.`ACTION_` AS `ACTION_`,`personnel`.`act_hi_comment`.`MESSAGE_` AS `MESSAGE_`,`personnel`.`act_hi_comment`.`FULL_MSG_` AS `FULL_MSG_`,`personnel`.`bpm_task_result`.`id`
AS `id`,`personnel`.`bpm_task_result`.`taskId` AS `taskId`,`personnel`.`bpm_task_result`.`processId` AS `processId`,`personnel`.`bpm_task_result`.`createTime` AS `createTime`,`personnel`.`bpm_task_result`.`state` AS `state` from `personnel`.`act_hi_comment`
left join `personnel`.`bpm_task_result` on((`personnel`.`act_hi_comment`.`TASK_ID_` = `personnel`.`bpm_task_result`.`taskId`)) where 1

明确了用不上关联字段taskid上的索引bpm_task_result_taskId_uindex的原因是collation conversion on field 'taskId'

找到原因了，那么就想解决办法：
方法一：将字符集校验规则修改成一致

在测试环境拷贝一张`activiti`.act_hi_comment，修改表的字符集校验规则，并且将已存在的数据做转换，SQL如下
ALTER TABLE `activiti`.`act_hi_comment` CONVERT TO CHARACTER SET utf8 COLLATE utf8_general_ci;

修改后的表结构如下：
 CREATE TABLE `act_hi_comment` (
  `ID_` varchar(64) NOT NULL,
  `TYPE_` varchar(255) DEFAULT NULL,
  `TIME_` datetime NOT NULL,
  `USER_ID_` varchar(255) DEFAULT NULL,
  `TASK_ID_` varchar(64) DEFAULT NULL,
  `PROC_INST_ID_` varchar(64) DEFAULT NULL,
  `ACTION_` varchar(255) DEFAULT NULL,
  `MESSAGE_` varchar(4000) DEFAULT NULL,
  `FULL_MSG_` longblob,
  PRIMARY KEY (`ID_`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8 ROW_FORMAT=COMPACT;

可以看到字符集校验规则已经变得和bpm_task_result一致了，再次查看执行计划：
EXPLAIN SELECT * from `activiti`.act_hi_comment a LEFT JOIN bpm_task_result b ON a.TASK_ID_ = b.taskId;



可以看到已经使用上了bpm_task_result_taskId_uindex这个索引

方法二：显示转换字符集校验规则：

很遗憾，目前convert和cast函数中不支持指定字符集校验规则，详见官方文档https://dev.mysql.com/doc/refman/5.7/en/cast-functions.html，但是可以直接在字段后面跟COLLATE utf8_general_ci，显示指定校验规则，所以SQL可以改成如下形式：

EXPLAIN SELECT * from act_hi_comment a LEFT JOIN bpm_task_result b ON a.TASK_ID_ COLLATE utf8_general_ci = b.taskId;



可以发现，也达到了同样的效果