mysql中的JOIN用法总结

join是mysql中一个基础的关键词，一般在多表连接查询中使用，这里做一下总结

1、JOIN的语法格式

table_references:
table_reference [, table_reference] ...

table_reference:
    table_factor
  | join_table

table_factor:
    tbl_name [[AS] alias]
        [{USE|IGNORE|FORCE} INDEX (key_list)]
  | ( table_references )
  | { OJ table_reference LEFT OUTER JOIN table_reference
        ON conditional_expr }

join_table:
    table_reference [INNER | CROSS] JOIN table_factor [join_condition]
  | table_reference STRAIGHT_JOIN table_factor
  | table_reference STRAIGHT_JOIN table_factor ON condition
  | table_reference LEFT [OUTER] JOIN table_reference join_condition
  | table_reference NATURAL [LEFT [OUTER]] JOIN table_factor
  | table_reference RIGHT [OUTER] JOIN table_reference join_condition
  | table_reference NATURAL [RIGHT [OUTER]] JOIN table_factor

join_condition:
    ON conditional_expr
  | USING (column_list)

 2、JOIN解析说明

我们先准备实验例子（mysql 版本：mysql  Ver 14.12 Distrib 5.0.95, for redhat-linux-gnu (i686) using readline 5.1）

CREATE TABLE `test_join_a` (
  `id` bigint(20) NOT NULL DEFAULT '0',
  `name` varchar(200),  
  PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;

CREATE TABLE `test_join_b` (
  `id` bigint(20) NOT NULL DEFAULT '0',
  `sex` varchar(200),  
  PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;

CREATE TABLE `test_join_c` (
  `id` bigint(20) NOT NULL DEFAULT '0',
  `class` varchar(200),  
  PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;

insert into test_join_a (id) values (1);
insert into test_join_a (id, name) values (2, "abc");
insert into test_join_a (id, name) values (3, "abcd");
insert into test_join_a (id, name) values (4, "abcde");

insert into test_join_b (id) values (1);
insert into test_join_b (id, sex) values (2, "abc");
insert into test_join_b (id, sex) values (3, "abc");

insert into test_join_c (id) values (1);
insert into test_join_c (id, class) values (2, "abc");
insert into test_join_c (id, class) values (3, "abc");

a、关于LEFT JOIN

    对于 A LEFT JOIN (B， C)  ON join_condition ，A表m，B表n， C表t 条记录   【说明：由于未发现专业术语描述A，B，C三表区别，此处，将A表暂称基础表，B,C表称附加表】
    * 以A表为基础表，从B表和C表中挑选符合ON join_condition条件的数据，如果没有符合的，则附加表(B,C)的列对应的行设置为NULL。
    * 通过这种连接方式，查询后的数据条数，最多为 m * n * t条， 最少为 m条   
    
    同理，对于(A, B) LEFT JOIN C ON join_condition
    * 以A， B表为基础表，从C表中挑选符合ON join_condition条件的数据，如果没有符合的，则附加表(C)的列对应的行设置为NULL。
    * 通过这种连接方式，查询后的数据条数，最多为 m * n * t条， 最少为 m * n条
    
    在LEFT JOIN 和 RIGHT JOIN 中，ON 子句后还有 WHERE 子句的筛选条件，区别在于
    1、where子句可以省略，ON 子句不能省略，如果不用ON join_condition，则可以使用ON 1=1 或者 ON 1
    2、ON join_condtion 会以基础表的笛卡尔积为基础，附加表没有符合ON join_condtion的记录，则将附加表列对应的行设为NULL。
    3、where 子句会将 ON join_condition生成的虚拟表做筛选。所以通过where子句后，获取的记录数，最少为 0条

 　注意： NULL=NULL 既不符合 join的on condition 也不符合 where子句的筛选   
    
     下面是两个实验过程，a表4条记录，b表3条，c表3条：

　　1、可以看到生成记录条数4 * 3 * 3 = 36 条

mysql> select a.id as aid, a.name, b.id as bid, b.sex, c.id as cid, c.class from test_join_a a left join (test_join_b b, test_join_c c) on 1=1;
+-----+-------+------+------+------+-------+
| aid | name  | bid  | sex  | cid  | class |
+-----+-------+------+------+------+-------+
|   1 | NULL  |    1 | NULL |    1 | NULL  | 
|   1 | NULL  |    1 | NULL |    2 | abc   | 
|   1 | NULL  |    1 | NULL |    3 | abc   | 
|   1 | NULL  |    2 | abc  |    1 | NULL  | 
|   1 | NULL  |    2 | abc  |    2 | abc   | 
|   1 | NULL  |    2 | abc  |    3 | abc   | 
|   1 | NULL  |    3 | abc  |    1 | NULL  | 
|   1 | NULL  |    3 | abc  |    2 | abc   | 
|   1 | NULL  |    3 | abc  |    3 | abc   | 
|   2 | abc   |    1 | NULL |    1 | NULL  | 
|   2 | abc   |    1 | NULL |    2 | abc   | 
|   2 | abc   |    1 | NULL |    3 | abc   | 
|   2 | abc   |    2 | abc  |    1 | NULL  | 
|   2 | abc   |    2 | abc  |    2 | abc   | 
|   2 | abc   |    2 | abc  |    3 | abc   | 
|   2 | abc   |    3 | abc  |    1 | NULL  | 
|   2 | abc   |    3 | abc  |    2 | abc   | 
|   2 | abc   |    3 | abc  |    3 | abc   | 
|   3 | abcd  |    1 | NULL |    1 | NULL  | 
|   3 | abcd  |    1 | NULL |    2 | abc   | 
|   3 | abcd  |    1 | NULL |    3 | abc   | 
|   3 | abcd  |    2 | abc  |    1 | NULL  | 
|   3 | abcd  |    2 | abc  |    2 | abc   | 
|   3 | abcd  |    2 | abc  |    3 | abc   | 
|   3 | abcd  |    3 | abc  |    1 | NULL  | 
|   3 | abcd  |    3 | abc  |    2 | abc   | 
|   3 | abcd  |    3 | abc  |    3 | abc   | 
|   4 | abcde |    1 | NULL |    1 | NULL  | 
|   4 | abcde |    1 | NULL |    2 | abc   | 
|   4 | abcde |    1 | NULL |    3 | abc   | 
|   4 | abcde |    2 | abc  |    1 | NULL  | 
|   4 | abcde |    2 | abc  |    2 | abc   | 
|   4 | abcde |    2 | abc  |    3 | abc   | 
|   4 | abcde |    3 | abc  |    1 | NULL  | 
|   4 | abcde |    3 | abc  |    2 | abc   | 
|   4 | abcde |    3 | abc  |    3 | abc   | 
+-----+-------+------+------+------+-------+
36 rows in set (0.00 sec)

　

　　2、筛选符合条件的记录，如果没有符合的，则设置为NULL，参看实验表中aid=1，3, 4情况。

mysql> select a.id as aid, a.name, b.id as bid, b.sex, c.id as cid, c.class from test_join_a a left join (test_join_b b, test_join_c c) on 
a.name=b.sex;
+-----+-------+------+------+------+-------+
| aid | name  | bid  | sex  | cid  | class |
+-----+-------+------+------+------+-------+
|   1 | NULL  | NULL | NULL | NULL | NULL  | 
|   2 | abc   |    2 | abc  |    1 | NULL  | 
|   2 | abc   |    2 | abc  |    2 | abc   | 
|   2 | abc   |    2 | abc  |    3 | abc   | 
|   2 | abc   |    3 | abc  |    1 | NULL  | 
|   2 | abc   |    3 | abc  |    2 | abc   | 
|   2 | abc   |    3 | abc  |    3 | abc   | 
|   3 | abcd  | NULL | NULL | NULL | NULL  | 
|   4 | abcde | NULL | NULL | NULL | NULL  | 
+-----+-------+------+------+------+-------+
9 rows in set (0.00 sec)

　　3、针对 (A, B) LEFT JOIN C ON join_condition 的情况，筛选符合条件的记录，以A，B表的笛卡尔积为基础，查找C中符合 on join_condition的记录数，如果没有，则将C中对应列的行值设置未NULL。

mysql> select a.id as aid, a.name, b.id as bid, b.sex, c.id as cid, c.class from (test_join_a a, test_join_b b) left join  test_join_c c on b.id>5;
+-----+-------+-----+------+------+-------+
| aid | name  | bid | sex  | cid  | class |
+-----+-------+-----+------+------+-------+
|   1 | NULL  |   1 | NULL | NULL | NULL  | 
|   1 | NULL  |   2 | abc  | NULL | NULL  | 
|   1 | NULL  |   3 | abc  | NULL | NULL  | 
|   2 | abc   |   1 | NULL | NULL | NULL  | 
|   2 | abc   |   2 | abc  | NULL | NULL  | 
|   2 | abc   |   3 | abc  | NULL | NULL  | 
|   3 | abcd  |   1 | NULL | NULL | NULL  | 
|   3 | abcd  |   2 | abc  | NULL | NULL  | 
|   3 | abcd  |   3 | abc  | NULL | NULL  | 
|   4 | abcde |   1 | NULL | NULL | NULL  | 
|   4 | abcde |   2 | abc  | NULL | NULL  | 
|   4 | abcde |   3 | abc  | NULL | NULL  | 
+-----+-------+-----+------+------+-------+
12 rows in set (0.00 sec)

　　4、可以看出ON condition的执行在WHERE之前，ON condition会形成虚拟表。

mysql> explain select a.id as aid, a.name, b.id as bid, b.sex, c.id as cid, c.class from test_join_a a left join (test_join_b b, test_join_c c) on b.id>5 where b.id>5;
+----+-------------+-------+-------+---------------+---------+---------+------+------+-------------------+
| id | select_type | table | type  | possible_keys | key     | key_len | ref  | rows | Extra             |
+----+-------------+-------+-------+---------------+---------+---------+------+------+-------------------+
|  1 | SIMPLE      | b     | range | PRIMARY       | PRIMARY | 8       | NULL |    1 | Using where       | 
|  1 | SIMPLE      | c     | ALL   | NULL          | NULL    | NULL    | NULL |    3 | Using join buffer | 
|  1 | SIMPLE      | a     | ALL   | NULL          | NULL    | NULL    | NULL |    4 | Using join buffer | 
+----+-------------+-------+-------+---------------+---------+---------+------+------+-------------------+
3 rows in set (0.00 sec)

　　5、WHERE 对虚拟表的筛选和ON条件是不一样的，ON condition会保存基础表的每条数据，附加表中没有符合的，则将附加表列设置为NULL，WHERE 只是筛选符合WHERE 子句的记录。 (可以从下面的第三次查询的结果看出，查询的是通过ON condition形成的虚拟表)

mysql> select a.id as aid, a.name, b.id as bid, b.sex, c.id as cid, c.class from test_join_a a left join (test_join_b b, test_join_c c) on b.id>5;              
+-----+-------+------+------+------+-------+
| aid | name  | bid  | sex  | cid  | class |
+-----+-------+------+------+------+-------+
|   1 | NULL  | NULL | NULL | NULL | NULL  | 
|   2 | abc   | NULL | NULL | NULL | NULL  | 
|   3 | abcd  | NULL | NULL | NULL | NULL  | 
|   4 | abcde | NULL | NULL | NULL | NULL  | 
+-----+-------+------+------+------+-------+
4 rows in set (0.00 sec)

mysql> select a.id as aid, a.name, b.id as bid, b.sex, c.id as cid, c.class from test_join_a a left join (test_join_b b, test_join_c c) on b.id>5 where b.id>5;
Empty set (0.00 sec)

mysql> select a.id as aid, a.name, b.id as bid, b.sex, c.id as cid, c.class from test_join_a a left join (test_join_b b, test_join_c c) ON a.name=b.sex where b.sex is NULL;
+-----+-------+------+------+------+-------+
| aid | name  | bid  | sex  | cid  | class |
+-----+-------+------+------+------+-------+
|   1 | NULL  | NULL | NULL | NULL | NULL  | 
|   3 | abcd  | NULL | NULL | NULL | NULL  | 
|   4 | abcde | NULL | NULL | NULL | NULL  | 
+-----+-------+------+------+------+-------+

  6、NULL=NULL 既不符合 join的on condition 也不符合 where子句的筛选。

mysql> select a.id as aid, a.name, b.id as bid, b.sex, c.id as cid, c.class from test_join_a a left join (test_join_b b, test_join_c c) ON a.name=b.sex;       
+-----+-------+------+------+------+-------+
| aid | name  | bid  | sex  | cid  | class |
+-----+-------+------+------+------+-------+
|   1 | NULL  | NULL | NULL | NULL | NULL  | 
|   2 | abc   |    2 | abc  |    1 | NULL  | 
|   2 | abc   |    2 | abc  |    2 | abc   | 
|   2 | abc   |    2 | abc  |    3 | abc   | 
|   2 | abc   |    3 | abc  |    1 | NULL  | 
|   2 | abc   |    3 | abc  |    2 | abc   | 
|   2 | abc   |    3 | abc  |    3 | abc   | 
|   3 | abcd  | NULL | NULL | NULL | NULL  | 
|   4 | abcde | NULL | NULL | NULL | NULL  | 
+-----+-------+------+------+------+-------+
9 rows in set (0.00 sec)

mysql> select a.id as aid, a.name, b.id as bid, b.sex, c.id as cid, c.class from test_join_a a left join (test_join_b b, test_join_c c) ON a.name=b.sex where a.name=b.sex;
+-----+------+------+------+------+-------+
| aid | name | bid  | sex  | cid  | class |
+-----+------+------+------+------+-------+
|   2 | abc  |    2 | abc  |    1 | NULL  | 
|   2 | abc  |    3 | abc  |    1 | NULL  | 
|   2 | abc  |    2 | abc  |    2 | abc   | 
|   2 | abc  |    3 | abc  |    2 | abc   | 
|   2 | abc  |    2 | abc  |    3 | abc   | 
|   2 | abc  |    3 | abc  |    3 | abc   | 
+-----+------+------+------+------+-------+
6 rows in set (0.00 sec)

b、关于RIGHT JOIN

　　right join和 left join类似，区别在于
　　1、对于 A right join (B, C) on join_condition ，会以 B,C表为基础表，A表为附加表
　　2、对于 (B, C) right join A on join_condition ，会以 A 表为基础表，B,C表为附加表

C、关于JOIN

JOIN与LEFT JOIN 和RIGHT JOIN的区别在于 JOIN 不区别左右基础表，可以认为他的ON condition 退化成where 子句形式，
例如：对于下面的SQL

select a.id as aid, a.name, b.id as bid, b.sex, c.id as cid, c.class from (test_join_a a ,test_join_b b) join test_join_c c ON a.name=b.sex;

结果上等价于

select a.id as aid, a.name, b.id as bid, b.sex, c.id as cid, c.class from (test_join_a a ,test_join_b b) join test_join_c c ON  1=1 where a.name=b.sex;

同时对于 A JOIN (B, C) 等价于 （A，B) JOIN C

关于JOIN的解析过程，我们可以理解成， 将A,B,C表做笛卡尔积，构成新的虚拟表， ON condition转化成where 子句，结合后面的where 子句做筛选，得出筛选结果。

我们来看一下实验情况：

mysql> explain select a.id as aid, a.name, b.id as bid, b.sex, c.id as cid, c.class from (test_join_a a ,test_join_b b) join test_join_c c ON a.name=b.sex; 
+----+-------------+-------+------+---------------+------+---------+------+------+--------------------------------+
| id | select_type | table | type | possible_keys | key  | key_len | ref  | rows | Extra                          |
+----+-------------+-------+------+---------------+------+---------+------+------+--------------------------------+
|  1 | SIMPLE      | b     | ALL  | NULL          | NULL | NULL    | NULL |    3 |                                | 
|  1 | SIMPLE      | c     | ALL  | NULL          | NULL | NULL    | NULL |    3 | Using join buffer              | 
|  1 | SIMPLE      | a     | ALL  | NULL          | NULL | NULL    | NULL |    4 | Using where; Using join buffer | 
+----+-------------+-------+------+---------------+------+---------+------+------+--------------------------------+
3 rows in set (0.00 sec)

mysql> explain select a.id as aid, a.name, b.id as bid, b.sex, c.id as cid, c.class from (test_join_a a ,test_join_b b) join test_join_c c ON 1=1 where a.name=b.sex;
+----+-------------+-------+------+---------------+------+---------+------+------+--------------------------------+
| id | select_type | table | type | possible_keys | key  | key_len | ref  | rows | Extra                          |
+----+-------------+-------+------+---------------+------+---------+------+------+--------------------------------+
|  1 | SIMPLE      | b     | ALL  | NULL          | NULL | NULL    | NULL |    3 |                                | 
|  1 | SIMPLE      | c     | ALL  | NULL          | NULL | NULL    | NULL |    3 | Using join buffer              | 
|  1 | SIMPLE      | a     | ALL  | NULL          | NULL | NULL    | NULL |    4 | Using where; Using join buffer | 
+----+-------------+-------+------+---------------+------+---------+------+------+--------------------------------+
3 rows in set (0.00 sec)

可以看出，两个sql的explain结果是一样的，拥有相同的执行计划

我们再看看这两个sql的执行结果

mysql> select a.id as aid, a.name, b.id as bid, b.sex, c.id as cid, c.class from (test_join_a a ,test_join_b b) join test_join_c c ON a.name=b.sex;                 
+-----+------+-----+------+-----+-------+
| aid | name | bid | sex  | cid | class |
+-----+------+-----+------+-----+-------+
|   2 | abc  |   2 | abc  |   1 | NULL  | 
|   2 | abc  |   3 | abc  |   1 | NULL  | 
|   2 | abc  |   2 | abc  |   2 | abc   | 
|   2 | abc  |   3 | abc  |   2 | abc   | 
|   2 | abc  |   2 | abc  |   3 | abc   | 
|   2 | abc  |   3 | abc  |   3 | abc   | 
+-----+------+-----+------+-----+-------+
6 rows in set (0.00 sec)

mysql> select a.id as aid, a.name, b.id as bid, b.sex, c.id as cid, c.class from (test_join_a a ,test_join_b b) join test_join_c c ON 1=1 where a.name=b.sex; 
+-----+------+-----+------+-----+-------+
| aid | name | bid | sex  | cid | class |
+-----+------+-----+------+-----+-------+
|   2 | abc  |   2 | abc  |   1 | NULL  | 
|   2 | abc  |   3 | abc  |   1 | NULL  | 
|   2 | abc  |   2 | abc  |   2 | abc   | 
|   2 | abc  |   3 | abc  |   2 | abc   | 
|   2 | abc  |   2 | abc  |   3 | abc   | 
|   2 | abc  |   3 | abc  |   3 | abc   | 
+-----+------+-----+------+-----+-------+
6 rows in set (0.00 sec)