我在使用错误（低效）索引的 MySQL 查询中遇到问题。
桌子：
mysql> describe ADDRESS_BOOK;
+---------------+--------------+------+-----+---------+----------------+
| Field | Type | Null | Key | Default | Extra |
+---------------+--------------+------+-----+---------+----------------+
| ADD_BOOK_ID | bigint(20) | NO | PRI | NULL | auto_increment |
| COMPANY_ID | bigint(20) | NO | MUL | NULL | |
| ADDRESS_NAME | varchar(150) | NO | MUL | NULL | |
| CLEAN_NAME | varchar(150) | NO | MUL | NULL | |
| ADDRESS_KEY_1 | varchar(150) | NO | MUL | NULL | |
| ADDRESS_KEY_2 | varchar(150) | NO | MUL | NULL | |
+---------------+--------------+------+-----+---------+----------------+
CLEAN_NAME 是普通 ADDRESS_NAME 的“已清理”版本，其中除 [a-zA-Z] 之外的所有内容都已删除，ADDRESS_KEY1 和 ADDRESS_KEY2 是 ADDRESS_NAME 中最长的两个单词，除 [a-zA-Z] 之外的所有内容都已删除。
这些是我的索引（玩弄它试图找到最好的）：
mysql> SHOW INDEX FROM ADDRESS_BOOK;
+--------------+------------+-------------------+--------------+---------------+-----------+-------------+----------+--------+------+------------+---------+---------------+
| Table | Non_unique | Key_name | Seq_in_index | Column_name | Collation | Cardinality | Sub_part | Packed | Null | Index_type | Comment | Index_comment |
+--------------+------------+-------------------+--------------+---------------+-----------+-------------+----------+--------+------+------------+---------+---------------+
| ADDRESS_BOOK | 0 | PRIMARY | 1 | ADD_BOOK_ID | A | 37847 | NULL | NULL | | BTREE | | |
| ADDRESS_BOOK | 1 | FK_ADDRESS_BOOK_2 | 1 | COMPANY_ID | A | 36 | NULL | NULL | | BTREE | | |
| ADDRESS_BOOK | 1 | IDX_ADDRESS_NAME | 1 | ADDRESS_NAME | A | 37847 | NULL | NULL | | BTREE | | |
| ADDRESS_BOOK | 1 | FX_ADDRESS_KEYS | 1 | CLEAN_NAME | A | 37847 | NULL | NULL | | BTREE | | |
| ADDRESS_BOOK | 1 | FX_ADDRESS_KEYS | 2 | ADDRESS_KEY_1 | A | 37847 | NULL | NULL | | BTREE | | |
| ADDRESS_BOOK | 1 | FX_ADDRESS_KEYS | 3 | ADDRESS_KEY_2 | A | 37847 | NULL | NULL | | BTREE | | |
| ADDRESS_BOOK | 1 | FX_ADDRESS_KEYS | 4 | COMPANY_ID | A | 37847 | NULL | NULL | | BTREE | | |
| ADDRESS_BOOK | 1 | FK_ADDRESS_2 | 1 | ADDRESS_KEY_2 | A | 18923 | NULL | NULL | | BTREE | | |
| ADDRESS_BOOK | 1 | FK_CLEAN | 1 | CLEAN_NAME | A | 37847 | NULL | NULL | | BTREE | | |
| ADDRESS_BOOK | 1 | FK_ADDRESS_1 | 1 | ADDRESS_KEY_1 | A | 37847 | NULL | NULL | | BTREE | | |
+--------------+------------+-------------------+--------------+---------------+-----------+-------------+----------+--------+------+------------+---------+---------------+
现在我的查询是：
select * from ADDRESS_BOOK addressboo0_
where (addressboo0_.CLEAN_NAME like concat('trad', '%')
or addressboo0_.ADDRESS_KEY_1 like concat('trad', '%')
or addressboo0_.ADDRESS_KEY_2 like concat('trad', '%'))
and addressboo0_.COMPANY_ID=1
order by addressboo0_.CLEAN_NAME asc
limit 200
系统中有来自不同公司的用户，因此查询应该只返回用户所在公司的地址簿条目。
对此的解释是
+----+-------------+--------------+------+----------------------------------------------------------------------+-------------------+---------+-------+------+-----------------------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+--------------+------+----------------------------------------------------------------------+-------------------+---------+-------+------+-----------------------------+
| 1 | SIMPLE | addressboo0_ | ref | FK_ADDRESS_BOOK_2,FX_ADDRESS_KEYS,FK_ADDRESS_2,FK_CLEAN,FK_ADDRESS_1 | FK_ADDRESS_BOOK_2 | 8 | const | 4108 | Using where; Using filesort |
+----+-------------+--------------+------+----------------------------------------------------------------------+-------------------+---------+-------+------+-----------------------------+
我知道 MySQL 不能在查询或查询上使用多列索引，但正如您所见，它使用的是 COMPANY (FK_ADDRESS_BOOK_2) 的索引，而不是字符串列的任何索引！
如果我将公司从查询中取出，它将使用其他索引：
+----+-------------+--------------+-------------+----------------------------------------------------+------------------------------------+-------------+------+------+-----------------------------------------------------------------------------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+--------------+-------------+----------------------------------------------------+------------------------------------+-------------+------+------+-----------------------------------------------------------------------------------+
| 1 | SIMPLE | addressboo0_ | index_merge | FX_ADDRESS_KEYS,FK_ADDRESS_2,FK_CLEAN,FK_ADDRESS_1 | FK_CLEAN,FK_ADDRESS_1,FK_ADDRESS_2 | 452,452,452 | NULL | 1089 | Using sort_union(FK_CLEAN,FK_ADDRESS_1,FK_ADDRESS_2); Using where; Using filesort |
+----+-------------+--------------+-------------+----------------------------------------------------+------------------------------------+-------------+------+------+-----------------------------------------------------------------------------------+
如果我对不同的公司使用相同的查询（包括公司），它会突然使用多列索引：
+----+-------------+--------------+-------+----------------------------------------------------------------------+-----------------+---------+------+------+-------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+--------------+-------+----------------------------------------------------------------------+-----------------+---------+------+------+-------------+
| 1 | SIMPLE | addressboo0_ | index | FK_ADDRESS_BOOK_2,FX_ADDRESS_KEYS,FK_ADDRESS_2,FK_CLEAN,FK_ADDRESS_1 | FX_ADDRESS_KEYS | 1364 | NULL | 492 | Using where |
+----+-------------+--------------+-------+----------------------------------------------------------------------+-----------------+---------+------+------+-------------+
因此，对于公司 1，它有 266 个结果，而对于公司 16，它有 437 个结果。公司 1 总共有 4109 个条目，而公司 16 有 7745 个条目。
所以我比较困惑。为什么 MySQL 为一家公司使用多列索引 FX_ADDRESS_KEYS 而为另一家公司使用相当低效的 FK_ADDRESS_BOOK_2 （基本上遍历该公司的每一行）。
如何改进查询/索引？如果我删除 ADDRESS_KEY_1 和 ADDRESS_KEY_2 的 or ，它正在使用 FX_ADDRESS_KEYS 索引，但我无法在名称中搜索字符串。如果我使用 '%trade%' 之类的东西，就不能使用索引。