当来自不同表的字段处于 OR 条件时,MySQL 不使用索引

标签 mysql join indexing

看到这个:Can you use index in mysql using "col1 OR col2"? 但我想我可以问这个相关的问题,以使讨论简单明了。

我有两个表:listsagree_and_disagree_count

lists 表有两个 timestamp 字段 modified_onfirst_publish_date。这两个字段都有自己的索引。

agree_and_disagree_count 表还有一个 timestamp 字段 date_created,该字段也已建立索引。

在同一个表中的字段的以下 OR 语句中,使用了两个索引。

mysql>  EXPLAIN SELECT * FROM lists l
    ->  WHERE (l.modified_on > '2013-01-07 12:50:51' OR
    ->   l.first_publish_date > '2013-01-07 12:50:51')\G
*************************** 1. row ***************************
           id: 1
  select_type: SIMPLE
        table: l
         type: index_merge
possible_keys: modifiedon,first_publish_date
          key: modifiedon,first_publish_date
      key_len: 4,9
          ref: NULL
         rows: 2
        Extra: Using sort_union(modifiedon,first_publish_date); Using where

但是,如果字段来自不同的表,则看起来没有使用索引:

mysql> EXPLAIN SELECT * FROM lists l
    -> JOIN agree_and_disagree_count adc ON adc.list_id=l.list_id
    -> WHERE (l.modified_on > '2013-01-07 12:50:51' OR
    -> adc.date_created > '2013-01-07 12:50:51')\G
*************************** 1. row ***************************
           id: 1
  select_type: SIMPLE
        table: adc
         type: ALL
possible_keys: list_id_type_id,idx_list_id,idx_date_created
          key: NULL
      key_len: NULL
          ref: NULL
         rows: 5114907
        Extra:
*************************** 2. row ***************************
           id: 1
  select_type: SIMPLE
        table: l
         type: eq_ref
possible_keys: PRIMARY,modifiedon
          key: PRIMARY
      key_len: 4
          ref: adc.list_id
         rows: 1
        Extra: Using where

为什么?


我不知道如何在评论中输入格式良好的回复,因此以下是对 @Iserni 的回复,他非常好心地帮助我:

我做了以下事情:

use test;

-- create tables
create table lists (
list_id INT UNSIGNED NOT NULL AUTO_INCREMENT,
modified_on TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
first_publish_date DATETIME NULL DEFAULT NULL,
primary key (list_id)
) ENGINE=InnoDB;

create table agree_and_disagree_count (
id INT UNSIGNED NOT NULL AUTO_INCREMENT,
list_id int unsigned not null,
date_created TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
primary key (id)
) ENGINE=InnoDB;

-- create indices
create index index_modified_on on lists (modified_on);
create index index_first_publish_date on lists (first_publish_date);

create index index_data_created on agree_and_disagree_count (date_created);

-- load data
insert into lists (list_id, modified_on, first_publish_date) values 
(1, now(), now()),
(2, now(), now()),
(3, now(), now());

insert into agree_and_disagree_count (list_id, date_created) values 
(1, now()), (1, now()), (2, now());

-- query
EXPLAIN SELECT * FROM lists l JOIN agree_and_disagree_count adc 
ON adc.list_id=l.list_id WHERE (l.modified_on > '2013-01-01 12:50:51'
OR adc.date_created > '2013-01-07 11:50:51');

+----+-------------+-------+------+----------------------------------+------+---------+------+------+--------------------------------+
| id | select_type | table | type | possible_keys                    | key  | key_len | ref  | rows | Extra                          |
+----+-------------+-------+------+----------------------------------+------+---------+------+------+--------------------------------+
|  1 | SIMPLE      | l     | ALL  | PRIMARY,index_modified_on        | NULL | NULL    | NULL |    3 |                                |
|  1 | SIMPLE      | adc   | ALL  | index_data_created,index_list_id | NULL | NULL    | NULL |    3 | Using where; Using join buffer |
+----+-------------+-------+------+----------------------------------+------+---------+------+------+--------------------------------+

-- create new indexes suggested by Iserni
CREATE INDEX adc_test_ndx ON agree_and_disagree_count (list_id, date_created);
CREATE INDEX l_test_ndx ON lists (list_id, modified_on);

-- query
EXPLAIN SELECT * FROM lists l JOIN agree_and_disagree_count adc 
ON adc.list_id=l.list_id WHERE (l.modified_on > '2013-01-01 12:50:51'
OR adc.date_created > '2013-01-07 11:50:51');

+----+-------------+-------+-------+-----------------------------------------------+--------------+---------+------+------+--------------------------------+
| id | select_type | table | type  | possible_keys                                 | key          | key_len | ref  | rows | Extra                          |
+----+-------------+-------+-------+-----------------------------------------------+--------------+---------+------+------+--------------------------------+
|  1 | SIMPLE      | adc   | index | index_data_created,index_list_id,adc_test_ndx | adc_test_ndx | 8       | NULL |    3 | Using index                    |
|  1 | SIMPLE      | l     | ALL   | PRIMARY,index_modified_on,l_test_ndx          | NULL         | NULL    | NULL |    3 | Using where; Using join buffer |
+----+-------------+-------+-------+-----------------------------------------------+--------------+---------+------+------+--------------------------------+

看起来完整扫描仍在进行?


请查看 @Iserni 在接受的答案下方的评论。

最佳答案

一旦list_id上的JOIN开始,MySQL就无法将列表标识符与查询中使用并保存在索引中的字段“链接”,因为没有索引同时包含 ID 这些字段;并且您无法对索引运行联接。

尝试创建这两个索引:

CREATE INDEX adc_test_ndx ON agree_and_disagree_count (list_id, date_created);
CREATE INDEX l_test_ndx ON lists (list_id, modified_on);

然后使用 EXPLAIN 重试。在我的系统上,我尝试创建两个最小的表,每个表有两条记录:

mysql> EXPLAIN SELECT * FROM lists l JOIN agree_and_disagree_count adc 
ON adc.list_id=l.list_id WHERE (l.modified_on > '2013-01-01 12:50:51'
OR adc.date_created > '2013-01-07 11:50:51')\G
*************************** 1. row ***************************
           id: 1
  select_type: SIMPLE
        table: l
         type: ALL
possible_keys: NULL
          key: NULL
      key_len: NULL
          ref: NULL
         rows: 2
        Extra:
*************************** 2. row ***************************
           id: 1
  select_type: SIMPLE
        table: adc
         type: ALL
possible_keys: NULL
          key: NULL
      key_len: NULL
          ref: NULL
         rows: 2
        Extra: Using where; Using join buffer
2 rows in set (0.00 sec)

然后:

mysql> CREATE INDEX adc_test_ndx ON agree_and_disagree_count (list_id, date_created);
Query OK, 0 rows affected (0.00 sec)
Records: 0  Duplicates: 0  Warnings: 0

mysql> CREATE INDEX l_test_ndx ON lists (list_id, modified_on);
Query OK, 0 rows affected (0.00 sec)
Records: 0  Duplicates: 0  Warnings: 0

mysql> EXPLAIN SELECT * FROM lists l JOIN agree_and_disagree_count adc 
ON adc.list_id=l.list_id WHERE (l.modified_on > '2013-01-01 12:50:51'
OR adc.date_created > '2013-01-07 11:50:51')\G
*************************** 1. row ***************************
           id: 1
  select_type: SIMPLE
        table: l
         type: index
possible_keys: l_test_ndx
          key: l_test_ndx
      key_len: 9
          ref: NULL
         rows: 2
        Extra: Using index
*************************** 2. row ***************************
           id: 1
  select_type: SIMPLE
        table: adc
         type: ref
possible_keys: adc_test_ndx
          key: adc_test_ndx
      key_len: 5
          ref: test.l.list_id
         rows: 1
        Extra: Using where; Using index
2 rows in set (0.00 sec)

mysql>

正如预期的那样,我现在在两个表上都建立了索引。

你能发布表的结构和现有索引,或者更好的是,一个 SQLFiddle 吗?

关于当来自不同表的字段处于 OR 条件时,MySQL 不使用索引,我们在Stack Overflow上找到一个类似的问题: https://stackoverflow.com/questions/14204651/

相关文章:

python - 根据子串匹配列

mysql - 这个多表连接应该怎么做呢?

sql - 索引不会提高性能

php - 更新查询有效但不插入查询

php - 使用 php 创建登录页面 - 看不到错误所在

mysql - MySql 存储过程中的事务

sql-server - 禁用所有非聚集索引

mysql - 查询两天内有效记录并按最近一个月的每一天进行统计

sql-server - SQL Server 条件 JOIN 语句

python - 如何在Python中使用图片网址数组替换索引?