join后的where，你清楚吗？

1.先看下示例表：ORACLE中，scott下的两张示例表

CREATE TABLE EMP
(
  EMPNO     NUMBER(4),
  ENAME     VARCHAR2(10 BYTE),
  JOB       VARCHAR2(9 BYTE),
  MGR       NUMBER(4),
  HIREDATE  DATE,
  SAL       NUMBER(7,2),
  COMM      NUMBER(7,2),
  DEPTNO    NUMBER(2)
)

CREATE TABLE DEPT
(
  DEPTNO  NUMBER(2),
  DNAME   VARCHAR2(14 BYTE),
  LOC     VARCHAR2(13 BYTE)
)

CREATE TABLE EMP
(
  EMPNO     NUMBER(4),
  ENAME     VARCHAR2(10 BYTE),
  JOB       VARCHAR2(9 BYTE),
  MGR       NUMBER(4),
  HIREDATE  DATE,
  SAL       NUMBER(7,2),
  COMM      NUMBER(7,2),
  DEPTNO    NUMBER(2)
)

CREATE TABLE DEPT
(
  DEPTNO  NUMBER(2),
  DNAME   VARCHAR2(14 BYTE),
  LOC     VARCHAR2(13 BYTE)
)

 

2.几条SQL引出的问题：（过滤条件在左表emp 中，sal != 3000）

--SQL1：全部当成join的条件
select e.* ,d.deptno , d.dname from emp e left join dept d
on e.deptno=d.deptno and sal != 3000  order by empno;

--SQL2：先过滤再join
select e.*,d.deptno ,d.dname from (select * from emp  where sal != 3000 ) e left join dept d
on e.deptno=d.deptno order by empno;
--SQL3：先过滤再join
select e.* ,d.deptno , d.dname from emp e left join dept d
on e.deptno=d.deptno where sal != 3000 order by empno;

--SQL1：全部当成join的条件
select e.* ,d.deptno , d.dname from emp e left join dept d 
on e.deptno=d.deptno and sal != 3000  order by empno;

--SQL2：先过滤再join
select e.*,d.deptno ,d.dname from (select * from emp  where sal != 3000 ) e left join dept d 
on e.deptno=d.deptno order by empno;
--SQL3：先过滤再join
select e.* ,d.deptno , d.dname from emp e left join dept d 
on e.deptno=d.deptno where sal != 3000 order by empno;

 SQL1生成的结果：

 

 SQL2和SQL3生成的结果一致：

 

SQL2或者3的执行计划是一致的：

执行计划
----------------------------------------------------------
Plan hash value: 1901738359

----------------------------------------------------------------------------
| Id  | Operation           | Name | Rows  | Bytes | Cost (%CPU)| Time     |
----------------------------------------------------------------------------
|   0 | SELECT STATEMENT    |      |    12 |   600 |     8  (25)| 00:00:01 |
|   1 |  SORT ORDER BY      |      |    12 |   600 |     8  (25)| 00:00:01 |
|*  2 |   HASH JOIN OUTER   |      |    12 |   600 |     7  (15)| 00:00:01 |
|*  3 |    TABLE ACCESS FULL| EMP  |    12 |   444 |     3   (0)| 00:00:01 |
|   4 |    TABLE ACCESS FULL| DEPT |     4 |    52 |     3   (0)| 00:00:01 |
----------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   2 - access("EMP"."DEPTNO"="D"."DEPTNO"(+))
   3 - filter("SAL"<>3000)

执行计划
----------------------------------------------------------
Plan hash value: 1901738359

----------------------------------------------------------------------------
| Id  | Operation           | Name | Rows  | Bytes | Cost (%CPU)| Time     |
----------------------------------------------------------------------------
|   0 | SELECT STATEMENT    |      |    12 |   600 |     8  (25)| 00:00:01 |
|   1 |  SORT ORDER BY      |      |    12 |   600 |     8  (25)| 00:00:01 |
|*  2 |   HASH JOIN OUTER   |      |    12 |   600 |     7  (15)| 00:00:01 |
|*  3 |    TABLE ACCESS FULL| EMP  |    12 |   444 |     3   (0)| 00:00:01 |
|   4 |    TABLE ACCESS FULL| DEPT |     4 |    52 |     3   (0)| 00:00:01 |
----------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   2 - access("EMP"."DEPTNO"="D"."DEPTNO"(+))
   3 - filter("SAL"<>3000)

  可见是先过滤sal!=3000,再join

 

而SQL1会把整个e.deptno=d.deptno and sal != 3000 当成join条件，一旦满足sal=3000 就会找不到对应的dept.

  

所以凡遇见join后加的是左边表的过滤

select e.* ,d.deptno , d.dname from emp e left join dept d
on e.deptno=d.deptno where sal != 3000 order by empno;

select e.* ,d.deptno , d.dname from emp e left join dept d 
on e.deptno=d.deptno where sal != 3000 order by empno; 

 

按照执行计划就可以理解成：

--先过滤再join
select e.*,d.deptno ,d.dname from (select * from emp  where sal != 3000 ) e left join dept d
on e.deptno=d.deptno order by empno;

--先过滤再join
select e.*,d.deptno ,d.dname from (select * from emp  where sal != 3000 ) e left join dept d 
on e.deptno=d.deptno order by empno;

 所以，对于过滤的是左边表emp的条件sal!=3000,若是将过滤条件放在join条件之中，例如SQL1，会把它当成整个JOIN的条件，若是将过滤条件当成where放在join后，例如SQL3,就是先过滤再join，结果自然就不一样了。

 

3.还是几条SQL引出的问题：（过滤条件在右表 dept中，d.loc!='NEW YORK'）

 

--SQL 4
select e.* ,  d.loc from emp e left join dept d on e.deptno=d.deptno and d.loc!='NEW YORK' order by empno;
--SQL 5
select e.*, d.loc from  emp e left join (select * from dept where dept.loc!='NEW YORK' ) d on e.deptno=d.deptno order by empno;
--SQL 6
select e.* , d.loc from emp e left join dept d on e.deptno=d.deptno where d.loc!='NEW YORK' order by empno;

--SQL 4
select e.* ,  d.loc from emp e left join dept d on e.deptno=d.deptno and d.loc!='NEW YORK' order by empno;
--SQL 5
select e.*, d.loc from  emp e left join (select * from dept where dept.loc!='NEW YORK' ) d on e.deptno=d.deptno order by empno;
--SQL 6
select e.* , d.loc from emp e left join dept d on e.deptno=d.deptno where d.loc!='NEW YORK' order by empno;

 SQL4的执行计划：

执行计划
----------------------------------------------------------
Plan hash value: 1901738359

----------------------------------------------------------------------------
| Id  | Operation           | Name | Rows  | Bytes | Cost (%CPU)| Time     |
----------------------------------------------------------------------------
|   0 | SELECT STATEMENT    |      |    14 |   672 |     8  (25)| 00:00:01 |
|   1 |  SORT ORDER BY      |      |    14 |   672 |     8  (25)| 00:00:01 |
|*  2 |   HASH JOIN OUTER   |      |    14 |   672 |     7  (15)| 00:00:01 |
|   3 |    TABLE ACCESS FULL| EMP  |    14 |   518 |     3   (0)| 00:00:01 |
|*  4 |    TABLE ACCESS FULL| DEPT |     3 |    33 |     3   (0)| 00:00:01 |
----------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   2 - access("E"."DEPTNO"="D"."DEPTNO"(+))
   4 - filter("D"."LOC"(+)<>'NEW YORK')

执行计划
----------------------------------------------------------
Plan hash value: 1901738359

----------------------------------------------------------------------------
| Id  | Operation           | Name | Rows  | Bytes | Cost (%CPU)| Time     |
----------------------------------------------------------------------------
|   0 | SELECT STATEMENT    |      |    14 |   672 |     8  (25)| 00:00:01 |
|   1 |  SORT ORDER BY      |      |    14 |   672 |     8  (25)| 00:00:01 |
|*  2 |   HASH JOIN OUTER   |      |    14 |   672 |     7  (15)| 00:00:01 |
|   3 |    TABLE ACCESS FULL| EMP  |    14 |   518 |     3   (0)| 00:00:01 |
|*  4 |    TABLE ACCESS FULL| DEPT |     3 |    33 |     3   (0)| 00:00:01 |
----------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   2 - access("E"."DEPTNO"="D"."DEPTNO"(+))
   4 - filter("D"."LOC"(+)<>'NEW YORK')

 

可见是先filter("D"."LOC"(+)<>'NEW YORK')，再join。相当于SQL5

所以SQL4和SQL5的执行结果是一样的。

 

但是SQL6的结果就很戏剧了，SQL6执行结果

 

 

 先join再过滤：

 select * from (select e.* ,d.loc from emp e left join dept d on e.deptno=d.deptno) where loc!='NEW YORK' order by empno

 

4.更经典的（过滤两张表共同的列）：

select e.* ,d.deptno, d.dname from emp e left join dept d on e.deptno=d.deptno where d.deptno!=30 order by empno;

select e.* ,d.deptno, d.dname from emp e left join dept d on e.deptno=d.deptno where d.deptno!=30 order by empno;

 

执行计划
----------------------------------------------------------
Plan hash value: 3357797783

----------------------------------------------------------------------------
| Id  | Operation           | Name | Rows  | Bytes | Cost (%CPU)| Time     |
----------------------------------------------------------------------------
|   0 | SELECT STATEMENT    |      |     9 |   450 |     8  (25)| 00:00:01 |
|   1 |  SORT ORDER BY      |      |     9 |   450 |     8  (25)| 00:00:01 |
|*  2 |   HASH JOIN         |      |     9 |   450 |     7  (15)| 00:00:01 |
|*  3 |    TABLE ACCESS FULL| DEPT |     3 |    39 |     3   (0)| 00:00:01 |
|*  4 |    TABLE ACCESS FULL| EMP  |     9 |   333 |     3   (0)| 00:00:01 |
----------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   2 - access("E"."DEPTNO"="D"."DEPTNO")
   3 - filter("D"."DEPTNO"<>30)
   4 - filter("E"."DEPTNO"<>30)

执行计划
----------------------------------------------------------
Plan hash value: 3357797783

----------------------------------------------------------------------------
| Id  | Operation           | Name | Rows  | Bytes | Cost (%CPU)| Time     |
----------------------------------------------------------------------------
|   0 | SELECT STATEMENT    |      |     9 |   450 |     8  (25)| 00:00:01 |
|   1 |  SORT ORDER BY      |      |     9 |   450 |     8  (25)| 00:00:01 |
|*  2 |   HASH JOIN         |      |     9 |   450 |     7  (15)| 00:00:01 |
|*  3 |    TABLE ACCESS FULL| DEPT |     3 |    39 |     3   (0)| 00:00:01 |
|*  4 |    TABLE ACCESS FULL| EMP  |     9 |   333 |     3   (0)| 00:00:01 |
----------------------------------------------------------------------------

Predicate Information (identified by operation id):
---------------------------------------------------

   2 - access("E"."DEPTNO"="D"."DEPTNO")
   3 - filter("D"."DEPTNO"<>30)
   4 - filter("E"."DEPTNO"<>30)

 

此时2张表都被过滤后（ 3 - filter("D"."DEPTNO"<>30)  4 - filter("E"."DEPTNO"<>30)）再JOIN。

 

查询结果：

 

 

呵呵，小心你的where和join吧，先想清楚业务需求，再想清楚SQL原理，才能查出正确的结果，否则就悲剧了。

附件scott.rar是导出的emp和dept的备份。oracle导出版本是 Release 10.2.0.1.0，低于这个版本的，将无法导入我导出的这个文件。

 

附SQL：

--全部当成join的条件
select e.* ,d.deptno , d.dname from emp e left join dept d
on e.deptno=d.deptno and sal != 3000  order by empno;

--先过滤再join
select e.*,d.deptno ,d.dname from (select * from emp  where sal != 3000 ) e left join dept d
on e.deptno=d.deptno order by empno;
--先过滤再join
select e.* ,d.deptno , d.dname from emp e left join dept d
on e.deptno=d.deptno where sal != 3000 order by empno;


--SQL 4
select e.* ,  d.loc from emp e left join dept d on e.deptno=d.deptno and d.loc!='NEW YORK' order by empno;
--SQL 5
select e.*, d.loc from  emp e left join (select * from dept where dept.loc!='NEW YORK' ) d on e.deptno=d.deptno order by empno;
--SQL 6
select e.* ,d.loc from emp e left join dept d on e.deptno=d.deptno where d.loc!='NEW YORK' order by empno;
select * from (select e.* ,d.loc from emp e left join dept d on e.deptno=d.deptno) where loc!='NEW YORK' order by empno


select e.* , d.deptno,d.dname from emp e left join dept d on e.deptno=d.deptno and d.deptno!=30 order by empno;

--先过滤再join
select e.*,d.deptno,d.dname from  emp e left join (select * from dept where dept.deptno!=30 ) d on e.deptno=d.deptno order by empno;

select e.* ,d.deptno, d.dname from emp e left join dept d on e.deptno=d.deptno where d.deptno!=30 order by empno;

--全部当成join的条件
select e.* ,d.deptno , d.dname from emp e left join dept d 
on e.deptno=d.deptno and sal != 3000  order by empno;

--先过滤再join
select e.*,d.deptno ,d.dname from (select * from emp  where sal != 3000 ) e left join dept d 
on e.deptno=d.deptno order by empno;
--先过滤再join
select e.* ,d.deptno , d.dname from emp e left join dept d 
on e.deptno=d.deptno where sal != 3000 order by empno;


--SQL 4
select e.* ,  d.loc from emp e left join dept d on e.deptno=d.deptno and d.loc!='NEW YORK' order by empno;
--SQL 5
select e.*, d.loc from  emp e left join (select * from dept where dept.loc!='NEW YORK' ) d on e.deptno=d.deptno order by empno;
--SQL 6
select e.* ,d.loc from emp e left join dept d on e.deptno=d.deptno where d.loc!='NEW YORK' order by empno;
select * from (select e.* ,d.loc from emp e left join dept d on e.deptno=d.deptno) where loc!='NEW YORK' order by empno


select e.* , d.deptno,d.dname from emp e left join dept d on e.deptno=d.deptno and d.deptno!=30 order by empno;

--先过滤再join
select e.*,d.deptno,d.dname from  emp e left join (select * from dept where dept.deptno!=30 ) d on e.deptno=d.deptno order by empno;

select e.* ,d.deptno, d.dname from emp e left join dept d on e.deptno=d.deptno where d.deptno!=30 order by empno;