SQL 三种基本Join

时间:2025-05-10 22:07:31

Join是关系型数据库系统的重要操作之一,SQL常用Join:内联接、外联接和交叉联接等。

这里讨论一下这常用的三种连接。

测试环境:db2 v10.1, linux

表定义:

 --用户
CREATE TABLE USER
(
USERID INTEGER NOT NULL,
COMPANYID INTEGER,
TELNO VARCHAR(12)
); --公司
CREATE TABLE COMPANY
(
COMPANYID INTEGER NOT NULL,
TELNO VARCHAR(12)
);

数据:

--USER

USERID      COMPANYID   TELNO
----------- ----------- ------------
11 2 777777
22 3 123456
33 4 567890 --COMPANY COMPANYID TELNO
----------- ------------
2 888888

1. inner join

[db2inst1@win ~]$ db2 "select * from user inner join company on user.companyid=company.companyid"

USERID      COMPANYID   TELNO        COMPANYID   TELNO
----------- ----------- ------------ ----------- ------------
11 2 777777 2 888888 1 record(s) selected.

注意:内联接(Inner join)满足交换律:“A inner join B” 和 “B inner join A” 是相等的。

查看访问计划:

Optimized Statement:
-------------------
SELECT
Q2.USERID AS "USERID",
Q2.COMPANYID AS "COMPANYID",
Q2.TELNO AS "TELNO",
Q1.COMPANYID AS "COMPANYID",
Q1.TELNO AS "TELNO"
FROM
DB2INST1.COMPANY AS Q1,
DB2INST1.USER AS Q2
WHERE
(Q2.COMPANYID = Q1.COMPANYID) Access Plan:
-----------
Total Cost: 13.5566
Query Degree: 1 Rows
RETURN
( 1)
Cost
I/O
|
1
HSJOIN
( 2)
13.5566
2
/-----+------\
3 1
TBSCAN TBSCAN
( 3) ( 4)
6.77858 6.77776
1 1
| |
3 1
TABLE: DB2INST1 TABLE: DB2INST1
USER COMPANY
Q2 Q1

用HSJOIN的方式进行,DB2对此进行了重写。

2. outer join

外部联接保存一个或两个输入表的所有行,即使无法找到匹配联接谓词的行。

[db2inst1@win ~]$ db2 "select * from user left outer join company on user.companyid=company.companyid"

USERID      COMPANYID   TELNO        COMPANYID   TELNO
----------- ----------- ------------ ----------- ------------
11 2 777777 2 888888
33 4 567890 - -
22 3 123456 - - 3 record(s) selected.
[db2inst1@win ~]$ db2 "select * from user right outer join company on user.companyid=company.companyid"

USERID      COMPANYID   TELNO        COMPANYID   TELNO
----------- ----------- ------------ ----------- ------------
11 2 777777 2 888888 1 record(s) selected.

查看访问计划:

Optimized Statement:
-------------------
SELECT
Q2.USERID AS "USERID",
Q2.COMPANYID AS "COMPANYID",
Q2.TELNO AS "TELNO",
Q1.COMPANYID AS "COMPANYID",
Q1.TELNO AS "TELNO"
FROM
DB2INST1.COMPANY AS Q1
LEFT OUTER JOIN DB2INST1.USER AS Q2
ON (Q2.COMPANYID = Q1.COMPANYID) Access Plan:
-----------
Total Cost: 13.5566
Query Degree: 1 Rows
RETURN
( 1)
Cost
I/O
|
1
HSJOIN<
( 2)
13.5566
2
/-----+------\
3 1
TBSCAN TBSCAN
( 3) ( 4)
6.77858 6.77776
1 1
| |
3 1
TABLE: DB2INST1 TABLE: DB2INST1
USER COMPANY
Q2 Q1

这里也是用的HSJOIN。

3. cross join

交叉联接(cross join)执行两个表的笛卡尔积(就是把表A和表B的数据进行一个N*M的组合)。也就是说,它匹配一个表与另一个表中的每一行;我们不能通过使用ON子句在交叉联接指定谓词,虽然我们可以使用WHERE子句来实现相同的结果,这是交叉联接基本上是作为一个内部联接了。

[db2inst1@win ~]$ db2 "select * from user cross join company"

USERID      COMPANYID   TELNO        COMPANYID   TELNO
----------- ----------- ------------ ----------- ------------
11 2 777777 2 888888
22 3 123456 2 888888
33 4 567890 2 888888 3 record(s) selected.

查看访问计划:

Optimized Statement:
-------------------
SELECT
Q2.USERID AS "USERID",
Q2.COMPANYID AS "COMPANYID",
Q2.TELNO AS "TELNO",
Q1.COMPANYID AS "COMPANYID",
Q1.TELNO AS "TELNO"
FROM
DB2INST1.COMPANY AS Q1,
DB2INST1.USER AS Q2 Access Plan:
-----------
Total Cost: 13.5563
Query Degree: 1 Rows
RETURN
( 1)
Cost
I/O
|
3
NLJOIN
( 2)
13.5563
2
/-----+------\
1 3
TBSCAN TBSCAN
( 3) ( 4)
6.77776 6.77858
1 1
| |
1 3
TABLE: DB2INST1 TABLE: DB2INST1
COMPANY USER
Q1 Q2

看到这里使用的是NLJOIN。