SQL常用函数 层次化查询(START BY … CONNECT BY PRIOR)

层次化查询,即树型结构查询,是SQL中经常用到的功能之一,通常由根节点,父节点,子节点,叶节点组成,其语法如下:

SELECT [LEVEL] ,column,expression,...
FROM table_name
[WHERE where_clause] [[START WITH start_condition] [CONNECT BY PRIOR prior_condition]];

LEVEL:为伪列,用于表示树的层次
start_condition：层次化查询的起始条件
prior_condition：定义父节点和子节点之间的关系

--使用start with ...connect by prior 从根节点开始遍历

SQL> select empno,mgr,ename,job from emp
2 start with empno = 7839
3 connect by prior empno = mgr;

EMPNO MGR ENAME JOB
---------- ---------- ---------- ---------
7839 KING PRESIDENT
7566 7839 JONES MANAGER
7788 7566 SCOTT ANALYST
7876 7788 ADAMS CLERK
7902 7566 FORD ANALYST
7369 7902 SMITH CLERK
7698 7839 BLAKE MANAGER
7499 7698 ALLEN SALESMAN
7521 7698 WARD SALESMAN
7654 7698 MARTIN SALESMAN
7844 7698 TURNER SALESMAN

EMPNO MGR ENAME JOB
---------- ---------- ---------- ---------
7900 7698 JAMES CLERK
7782 7839 CLARK MANAGER
7934 7782 MILLER CLERK

14 rows selected.

树型结构遍历过程(通过上面的查询来描述)

1).从根节点开始(即where_clause中的条件,如果为非根节点则分根节点作为根节点开始遍历,如上例empno = 7839)
2).遍历根节点(得到empno = 7839记录的相关信息)
3).判断该节点是否存在由子节点，如果则访问最左侧未被访问的子节点,转到),否则下一步
如上例中prior_condition为empno = mgr,即子节点的mgr等于父节点的empno,在此时mgr为7839的记录
4).当节点为叶节点，则访问完毕,否则,转到)
5).返回到该节点的父节点,转到)

--伪列level的使用

--注意connect by prior empno = mgr 的理解
--prior表示前一条记录,即下一条返回记录的mgr应当等于前一条记录的empno

SQL> select level,empno,mgr,ename,job from emp
2 start with ename = 'KING'
3 connect by prior empno = mgr
4 order by level;

LEVEL EMPNO MGR ENAME JOB
---------- ---------- ---------- ---------- ---------
1 7839 KING PRESIDENT
2 7566 7839 JONES MANAGER
2 7698 7839 BLAKE MANAGER
2 7782 7839 CLARK MANAGER
3 7902 7566 FORD ANALYST
3 7521 7698 WARD SALESMAN
3 7900 7698 JAMES CLERK
3 7934 7782 MILLER CLERK
3 7499 7698 ALLEN SALESMAN
3 7788 7566 SCOTT ANALYST
3 7654 7698 MARTIN SALESMAN

LEVEL EMPNO MGR ENAME JOB
---------- ---------- ---------- ---------- ---------
3 7844 7698 TURNER SALESMAN
4 7876 7788 ADAMS CLERK
4 7369 7902 SMITH CLERK

--获得层次数
SQL> select count(distinct level) "Level" from emp
2 start with ename = 'KING'
3 connect by prior empno = mgr;

Level
----------
4

--格式化层次查询结果(使用左填充* level - 1个空格)
SQL> col Ename for a30
SQL> select level,
2 lpad(' ',2 * level - 1) || ename as "Ename",
3 job
4 from emp
5 start with ename = 'KING'
6 connect by prior empno = mgr;

LEVEL Ename JOB
---------- ------------------------------ ---------
1 KING PRESIDENT
2 JONES MANAGER
3 SCOTT ANALYST
4 ADAMS CLERK
3 FORD ANALYST
4 SMITH CLERK
2 BLAKE MANAGER
3 ALLEN SALESMAN
3 WARD SALESMAN
3 MARTIN SALESMAN
3 TURNER SALESMAN

LEVEL Ename JOB
---------- ------------------------------ ---------
3 JAMES CLERK
2 CLARK MANAGER
3 MILLER CLERK

14 rows selected.

--从非根节点开始遍历(只需修改start with 中的条件即可)
SQL> select level,
2 lpad(' ',2 * level - 1) || ename as "Ename",
3 job
4 from emp
5 start with ename = 'SCOTT'
6 connect by prior empno = mgr;

LEVEL Ename JOB
---------- ------------------------------ ---------
1 SCOTT ANALYST
2 ADAMS CLERK

--从下向上遍历(交换connect by prior中的条件即可,使用mgr = empno)
--注意connect by prior mgr = empno 的理解
--prior表示前一条记录,即下一条返回记录的empno应当等于前一条记录的mgr

SQL> select level,
2 lpad(' ',2 * level - 1) || ename as "Ename",
3 job
4 from emp
5 start with ename = 'SCOTT'
6 connect by prior mgr = empno;

LEVEL Ename JOB
---------- ------------------------------ ---------
1 SCOTT ANALYST
2 JONES MANAGER
3 KING PRESIDENT

--从下向上遍历(也可以将prior置于等号右边,得到相同的结果)

SQL> select level,
2 lpad(' ',2 * level - 1) || ename as "Ename",
3 job
4 from emp
5 start with ename = 'SCOTT'
6 connect by empno = prior mgr;

LEVEL Ename JOB
---------- ------------------------------ ---------
1 SCOTT ANALYST
2 JONES MANAGER
3 KING PRESIDENT

--从层次查询中删除节点和分支
SQL> select level,
2 lpad(' ',2 * level - 1) || ename as "Ename"
3 ,job
4 from emp
5 where ename != 'SCOTT' --通过where子句来过滤SCOTT用户，但SCOTT的下属ADAMS并没有过滤掉
6 start with empno = 7839
7 connect by prior empno = mgr;

LEVEL Ename JOB
---------- -------------------- ---------
1 KING PRESIDENT
2 JONES MANAGER
4 ADAMS CLERK
3 FORD ANALYST
4 SMITH CLERK
2 BLAKE MANAGER
3 ALLEN SALESMAN
3 WARD SALESMAN
3 MARTIN SALESMAN
3 TURNER SALESMAN
3 JAMES CLERK

LEVEL Ename JOB
---------- -------------------- ---------
2 CLARK MANAGER
3 MILLER CLERK

13 rows selected.

--通过将过滤条件由where 子句的内容移动到connect by prior 子句中过滤掉SCOTT及其下属
SQL> select level,
2 lpad(' ',2 * level - 1) || ename as "Ename"
3 ,job
4 from emp
5 start with empno = 7839
6 connect by prior empno = mgr and ename != 'SCOTT';

LEVEL Ename JOB
---------- -------------------- ---------
1 KING PRESIDENT
2 JONES MANAGER
3 FORD ANALYST
4 SMITH CLERK
2 BLAKE MANAGER
3 ALLEN SALESMAN
3 WARD SALESMAN
3 MARTIN SALESMAN
3 TURNER SALESMAN
3 JAMES CLERK
2 CLARK MANAGER

LEVEL Ename JOB
---------- -------------------- ---------
3 MILLER CLERK

12 rows selected.

--在层次化查询中增加过滤条件或使用子查询
SQL> select level,
2 lpad(' ',2 * level - 1) || ename as "Ename"
3 ,job
4 from emp
5 where sal > 2500
6 start with empno = 7839
7 connect by prior empno = mgr
8 ;

LEVEL Ename JOB
---------- -------------------- ---------
1 KING PRESIDENT
2 JONES MANAGER
3 SCOTT ANALYST
3 FORD ANALYST
2 BLAKE MANAGER

SQL> select level,
2 lpad(' ',2 * level - 1) || ename as "Ename"
3 ,job
4 from emp
5 where sal > (select avg(sal) from emp)
6 start with empno = 7839
7 connect by prior empno = mgr ;

LEVEL Ename JOB
---------- -------------------- ---------
1 KING PRESIDENT
2 JONES MANAGER
3 SCOTT ANALYST
3 FORD ANALYST
2 BLAKE MANAGER
2 CLARK MANAGER

6 rows selected.