oracle中 start with connect by 用法

权姣

　　oracle 提供了start with connect by 语法结构可以实现递归查询。

1. 一个简单举例:
SQL> select *  from test;

BILL_MONTH           DAY_NUMBER MSISDN
-------------------- ---------- --------------------
200803                        1 13800
200803                        3 13800
200803                        2 13800
200803                        2 13801
200803                        4 13804
200803                        5 13804
200803                        7 13804
200803                        8 13804
200803                        6 13802
200803                        6 13801
200803                        7 13801
200803                        8 13801

12 rows selected

SQL>
SQL> select * from test
  2       start with day_number=1
  3       connect by  prior day_number=day_number-1 and prior msisdn= msisdn
  4      ;

BILL_MONTH           DAY_NUMBER MSISDN
-------------------- ---------- --------------------
200803                        1 13800
200803                        2 13800
200803                        3 13800

SQL>


上面的语句查找出了从1开始，并且day_number 逐渐+1 递增的，并且 msisdn 相同的哪些个数据.


2. start with  connect by 语法结构
 如上面说看到的 例子， 其语法结构为  start with condition  connect by  condition （含 prior 关键字)
start with conditon 给出的seed 数据的范围, connect by  后面给出了递归查询的条件,prior 关键字表示父数据，prior 条件表示子数据需要满足父数据的什么条件【要的时候有两种写法：connect by prior day_number=day_number-1 或 connect by day_number=prior day_number-1，前一种写法表示采用自上而下的搜索方式（先找父节点然后找子节点），后一种写法表示采用自下而上的搜索方式（先找叶子节点然后找父节点）】。

在下面的这个start with connect by 结构中，就表示 查找出了从1开始，并且day_number 逐渐+1 递增的，并且 msisdn 相同的那些个数据.

start with day_number=1
     connect by  prior day_number=day_number-1 and prior msisdn= msisdn

3.  执行计划
对于这个特殊的语法结构，我们来看看它的执行计划。
通过下面的执行计划，我们可以看出，对于简单的访问一个对象的递归查询，实际上oracle 要三次访问要查询的对象。因此，这一个告诉我们，在使用递归查询时，一定要谨慎，因为即使原表数据不多，但是三倍的访问喜爱来，代价也会很大。

SQL> explain plan for
  2
  2   select * from  test
  3    --where  bill_month='200803'
  4    start with day_number=1
  5    connect by  prior day_number=day_number-1 and prior msisdn= msisdn
  6  ;

Explained

SQL> select *  from  table( dbms_xplan.display);

PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------
-------------------------------------------------------------------------
| Id  | Operation                 |  Name       | Rows  | Bytes | Cost  |
-------------------------------------------------------------------------
|   0 | SELECT STATEMENT          |             |       |       |       |
|*  1 |  CONNECT BY WITH FILTERING|             |       |       |       |
|*  2 |   FILTER                  |             |       |       |       |
|   3 |    TABLE ACCESS FULL      | TEST        |       |       |       |
|   4 |   NESTED LOOPS            |             |       |       |       |
|   5 |    BUFFER SORT            |             |       |       |       |
|   6 |     CONNECT BY PUMP       |             |       |       |       |
|*  7 |    TABLE ACCESS FULL      | TEST        |       |       |       |
|   8 |   TABLE ACCESS FULL       | TEST        |       |       |       |
-------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
   1 - filter("TEST"."DAY_NUMBER"=1)
   2 - filter("TEST"."DAY_NUMBER"=1)

PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------
   7 - filter("TEST"."MSISDN"=NULL AND "TEST"."DAY_NUMBER"-1=NULL)
Note: rule based optimization

23 rows selected

SQL>


另外，发现了在含有其他条件的递归中，是先处理所有的递归查询，最后才用加入的条件过滤.
请看下面的例子。
和上面的执行计划对比下我们可以知道，加入条件   where  bill_month='200803' 后，实际上却是在递归完成后，最后才执行的    1 - filter("TEST"."BILL_MONTH"='200803') 。

所以，为了确保语句的性能，不要直接加入条件在start with connect by 结构中，而是要想办法将原表的数据控制住。这个可以采用子查询的办法，或者使用临时表等（最好采用临时表，将数据量从本源上控制住；因为从子查询的执行计划我们可以看到，它每次也都是访问全表，再用条件过滤，要重复三次，不是一次过滤就够了).

--直接加入条件后的执行计划
SQL> explain plan for
  2
  2   select * from  test
  3    where  bill_month='200803'
  4    start with day_number=1
  5    connect by  prior day_number=day_number-1 and prior msisdn= msisdn
  6  ;

Explained

SQL> select *  from  table( dbms_xplan.display);

PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------
--------------------------------------------------------------------------
| Id  | Operation                  |  Name       | Rows  | Bytes | Cost  |
--------------------------------------------------------------------------
|   0 | SELECT STATEMENT           |             |       |       |       |
|*  1 |  FILTER                    |             |       |       |       |
|*  2 |   CONNECT BY WITH FILTERING|             |       |       |       |
|*  3 |    FILTER                  |             |       |       |       |
|   4 |     TABLE ACCESS FULL      | TEST        |       |       |       |
|   5 |    NESTED LOOPS            |             |       |       |       |
|   6 |     BUFFER SORT            |             |       |       |       |
|   7 |      CONNECT BY PUMP       |             |       |       |       |
|*  8 |     TABLE ACCESS FULL      | TEST        |       |       |       |
|   9 |    TABLE ACCESS FULL       | TEST        |       |       |       |
--------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
   1 - filter("TEST"."BILL_MONTH"='200803')

PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------
   2 - filter("TEST"."DAY_NUMBER"=1)
   3 - filter("TEST"."DAY_NUMBER"=1)
   8 - filter("TEST"."MSISDN"=NULL AND "TEST"."DAY_NUMBER"-1=NULL)
Note: rule based optimization

25 rows selected

SQL>


--使用子查询，将过滤条件嵌在子查询中
SQL> explain plan for
  2
  2  select * from (select * from test
  3        where  bill_month='200803')
  4       start with day_number=1
  5       connect by  prior day_number=day_number-1 and prior msisdn= msisdn
  6      ;

Explained

SQL> select *  from table(dbms_xplan.display);

PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------
-------------------------------------------------------------------------
| Id  | Operation                 |  Name       | Rows  | Bytes | Cost  |
-------------------------------------------------------------------------
|   0 | SELECT STATEMENT          |             |       |       |       |
|*  1 |  CONNECT BY WITH FILTERING|             |       |       |       |
|*  2 |   FILTER                  |             |       |       |       |
|*  3 |    TABLE ACCESS FULL      | TEST        |       |       |       |
|   4 |   NESTED LOOPS            |             |       |       |       |
|   5 |    BUFFER SORT            |             |       |       |       |
|   6 |     CONNECT BY PUMP       |             |       |       |       |
|*  7 |    TABLE ACCESS FULL      | TEST        |       |       |       |
|*  8 |   TABLE ACCESS FULL       | TEST        |       |       |       |
-------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
   1 - filter("TEST"."DAY_NUMBER"=1)
   2 - filter("TEST"."DAY_NUMBER"=1)

PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------
   3 - filter("TEST"."BILL_MONTH"='200803')
   7 - filter("TEST"."BILL_MONTH"='200803' AND "TEST"."MSISDN"=NULL AND
              "TEST"."DAY_NUMBER"-1=NULL)
   8 - filter("TEST"."BILL_MONTH"='200803')
Note: rule based optimization

26 rows selected

SQL>



4. 实际中 递归查询的使用。

问题：
数据库里有字段day_number，msisdn。如何写月度连续3天有记录的手机号？表结构如下：

id   bill_month   day_number     msisdn
1      200803      1           13800000000
2      200803      1           130137.....
3      200803      2           13800000000
4      200803      3           13800000000
..............................

表中3月份连续3天有记录的纪录就是1380000000。请问如何写这样的sql？


解决方案：
SQL> create  table   test ( bill_month varchar2(20),day_number number ,msisdn varchar2(20));

Table created

SQL> insert into  test values ( '200803',1,'13800');

1 row inserted

SQL> insert into  test values ( '200803',3,'13800');

1 row inserted

SQL> insert into  test values ( '200803',2,'13800');

1 row inserted

SQL> insert into  test values ( '200803',2,'13801');

1 row inserted

SQL> insert into  test values ( '200803',4,'13804');

1 row inserted

SQL> insert into  test values ( '200803',5,'13804');

1 row inserted

SQL> commit;

Commit complete

SQL> select *  from test;

BILL_MONTH           DAY_NUMBER MSISDN
-------------------- ---------- --------------------
200803                        1 13800
200803                        3 13800
200803                        2 13800
200803                        2 13801
200803                        4 13804
200803                        5 13804

6 rows selected

SQL>
SQL> select distinct  msisdn  from test  a
  2  where  bill_month='200803'
  3  and exists
  4  ( select msisdn from  test
  5    where  bill_month='200803' and msisdn=a.msisdn
  6    start with day_number=a.day_number
  7    connect by  prior day_number=day_number-1 and prior msisdn= msisdn
  8    group by msisdn
  9    having count(*)>=3
 10    );

MSISDN
--------------------
13800


SQL> select *  from test;

BILL_MONTH           DAY_NUMBER MSISDN
-------------------- ---------- --------------------
200803                        1 13800
200803                        3 13800
200803                        2 13800
200803                        2 13801
200803                        4 13804
200803                        5 13804

6 rows selected

SQL> insert into  test values ( '200803',7,'13804');

1 row inserted

SQL> insert into  test values ( '200803',8,'13804');

1 row inserted

SQL> insert into  test values ( '200803',6,'13802');

1 row inserted

SQL> insert into  test values ( '200803',6,'13801');

1 row inserted

SQL> insert into  test values ( '200803',7,'13801');

1 row inserted

SQL> insert into  test values ( '200803',8,'13801');

1 row inserted

SQL> select *  from test;

BILL_MONTH           DAY_NUMBER MSISDN
-------------------- ---------- --------------------
200803                        1 13800
200803                        3 13800
200803                        2 13800
200803                        2 13801
200803                        4 13804
200803                        5 13804
200803                        7 13804
200803                        8 13804
200803                        6 13802
200803                        6 13801
200803                        7 13801
200803                        8 13801

12 rows selected

SQL> commit;

Commit complete

SQL>
SQL> select distinct  msisdn  from test  a
  2  where  bill_month='200803'
  3  and exists
  4  ( select msisdn from  test
  5    where  bill_month='200803' and msisdn=a.msisdn
  6    start with day_number=a.day_number
  7    connect by  prior day_number=day_number-1 and prior msisdn= msisdn
  8    group by msisdn
  9    having count(*)>=3
 10    );

MSISDN
--------------------
13800
13801