mysql - 为什么 MySQL 不在 WHERE IF 子句中使用索引？

Question

因为这个设置：

mysql> show global variables like '%indexes';
+-------------------------------+-------+
| Variable_name                 | Value |
+-------------------------------+-------+
| log_queries_not_using_indexes | ON    | 
+-------------------------------+-------+

慢查询日志不断收到：

# Time: 120607 16:58:30
# User@Host: xbtit[xbtit] @  [123.30.53.244]
# Query_time: 0  Lock_time: 0  Rows_sent: 1  Rows_examined: 16006
SELECT * FROM xbtit_files WHERE IF(soha_id is null OR soha_id = '', info_hash, soha_id)='6d63dd4ab199190b531752067414d4d6e6568f90';

试图解释这个查询：

mysql> EXPLAIN SELECT * FROM xbtit_files WHERE IF(soha_id is null OR soha_id = '', info_hash, soha_id)='6d63dd4ab199190b531752067414d4d6e6568f90';
+----+-------------+-------------+------+---------------+------+---------+------+-------+-------------+
| id | select_type | table       | type | possible_keys | key  | key_len | ref  | rows  | Extra       |
+----+-------------+-------------+------+---------------+------+---------+------+-------+-------------+
|  1 | SIMPLE      | xbtit_files | ALL  | NULL          | NULL | NULL    | NULL | 16006 | Using where | 
+----+-------------+-------------+------+---------------+------+---------+------+-------+-------------+

令我惊讶的是为什么 MySQL 不使用索引：

mysql> show index from xbtit_files;
+-------------+------------+-----------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+
| Table       | Non_unique | Key_name  | Seq_in_index | Column_name | Collation | Cardinality | Sub_part | Packed | Null | Index_type | Comment |
+-------------+------------+-----------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+
| xbtit_files |          0 | PRIMARY   |            1 | info_hash   | A         |       16006 |     NULL | NULL   |      | BTREE      |         | 
| xbtit_files |          1 | filename  |            1 | filename    | A         |       16006 |     NULL | NULL   | YES  | BTREE      |         | 
| xbtit_files |          1 | category  |            1 | category    | A         |           1 |     NULL | NULL   |      | BTREE      |         | 
| xbtit_files |          1 | uploader  |            1 | uploader    | A         |          16 |     NULL | NULL   |      | BTREE      |         | 
| xbtit_files |          1 | bin_hash  |            1 | bin_hash    | A         |       16006 |       20 | NULL   |      | BTREE      |         | 
| xbtit_files |          1 | ix_sohaid |            1 | soha_id     | A         |       16006 |     NULL | NULL   | YES  | BTREE      |         | 
+-------------+------------+-----------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+

FORCE INDEX也不起作用：

mysql> EXPLAIN SELECT * FROM xbtit_files force index (PRIMARY) WHERE IF(soha_id is null OR soha_id = '', info_hash, soha_id)='6d63dd4ab199190b531752067414d4d6e6568f90';
+----+-------------+-------------+------+---------------+------+---------+------+-------+-------------+
| id | select_type | table       | type | possible_keys | key  | key_len | ref  | rows  | Extra       |
+----+-------------+-------------+------+---------------+------+---------+------+-------+-------------+
|  1 | SIMPLE      | xbtit_files | ALL  | NULL          | NULL | NULL    | NULL | 16006 | Using where | 
+----+-------------+-------------+------+---------------+------+---------+------+-------+-------------+

我必须将此查询拆分为 2 个操作吗？

Answer 1

在MySQL中，您不能在表达式上创建索引，并且优化器不够聪明，无法针对两个索引拆分您的查询。

用这个：

SELECT  *
FROM    xbtit_files 
WHERE   soha_id = '6d63dd4ab199190b531752067414d4d6e6568f90'
UNION ALL
SELECT  *
FROM    xbtit_files 
WHERE   soha_id = ''
        AND info_hash = '6d63dd4ab199190b531752067414d4d6e6568f90'
UNION ALL
SELECT  *
FROM    xbtit_files 
WHERE   soha_id IS NULL
        AND info_hash = '6d63dd4ab199190b531752067414d4d6e6568f90'

每个查询都使用自己的索引。

您可以将其组合成一个查询：

SELECT  *
FROM    xbtit_files 
WHERE   (
        soha_id = '6d63dd4ab199190b531752067414d4d6e6568f90'
        OR
        (soha_id = '' AND info_hash = '6d63dd4ab199190b531752067414d4d6e6568f90')
        OR
        (soha_id IS NULL AND info_hash = '6d63dd4ab199190b531752067414d4d6e6568f90')
        )

并为此创建一个复合索引以(soha_id, info_hash)使其快速工作。

MySQL还能够使用 将来自两个索引的结果合并在一起，index_merge因此即使您没有创建复合索引，您也有可能在第二个查询的计划中看到这一点。

Answer 2

您可以阅读本文以了解为什么OR运算符不适用于索引数据库。

Answer 3

因为函数是黑盒子：http ://use-the-index-luke.com/sql/where-clause/functions/case-insensitive-search

编辑- 给你的上下文太少了，抱歉。

相关部分是：

It is a trap we all fall into. We instantly recognize the relation between 
LAST_NAME and UPPER(LAST_NAME) and expect the database to “see” it as well.
In fact, the optimizer’s picture is more like that:

SELECT first_name, last_name, phone_number
  FROM employees
 WHERE BLACKBOX(...) = 'WINAND';

The UPPER function is just a black box. The parameters to the function are
not relevant because there is no general relationship between the function’s
parameters and the result.

这适用于所有函数：UPPER、IF 等等……

MySQL 被划掉是因为该问题的解决方案（在页面下方进行了描述）不适用于 MySQL。

Answer 4

在可能会降低性能的地方使用函数（LEFT函数除外）。试试这个查询

SELECT * FROM xbtit_files WHERE 
((soha_id is null OR soha_id = '') AND (info_hash='6d63dd4ab199190b531752067414d4d6e6568f90')) OR
( (soha_id='6d63dd4ab199190b531752067414d4d6e6568f90'))

Answer 5

主键基于 torrent 的哈希值，但您可以添加字段 id 并使用主键定义它，
如下所示：

ALTER TABLE `xbtit_files` DROP PRIMARY KEY;
ALTER TABLE `xbtit_files` ADD `id` INT NOT NULL AUTO_INCREMENT PRIMARY KEY FIRST;
ALTER TABLE `xbtit_files` ADD UNIQUE (`info_hash`);

不要忘记将字段 info_hash 放入UNIQUE