ROWNUM()应用案例-实现一个拉链表

Posted 2020-12-14 踏叶乘风

有以下数据表：事件标识表未这个股票进入(1)或者移除(2)某个股票池。我们需要对数据进行去重，就是连续多次的进入，或者连续多次的移除，都只取第一次的日期。

 

 

如图中，黄色的记录是不需要的。(这个表命名为demo)

 

 

 

首先第一步

 select SCODE,POOLID,EFFECTIVE_DAY,ADJUSTMODE
         ,row_number() over (partition by SCODE,POOLID,ADJUSTMODE order by EFFECTIVE_DAY) as rn
         ,row_number() over (partition by SCODE,POOLID order by EFFECTIVE_DAY) as rn2
  from  demo t 

得到如下结果(我将进入的标注了红色):

 

 

 观察数据的规律，然后我们用RN2-RN1。再按SCODE+POOLID+ADJUSTMODE分组中，按日期进行排序。

SQL:

select x.*,rn2-rn,row_number() over (partition by scode,poolid,adjustmode,rn2-rn order by rn) as rn3
from (      
  select SCODE,POOLID,EFFECTIVE_DAY,ADJUSTMODE
         ,row_number() over (partition by SCODE,POOLID,ADJUSTMODE order by EFFECTIVE_DAY) as rn
         ,row_number() over (partition by SCODE,POOLID order by EFFECTIVE_DAY) as rn2
  from  demo t 
  ) x
order by 1,2,3

 

 

 

然后发现RN3=1的就是我们需要的记录.

 

完整的代码如下：

with demo as (
  select \'00001\' as SCODE,1 POOLID,date\'2010-01-01\' as EFFECTIVE_DAY,1 as ADJUSTMODE from dual
      union all
      select \'00001\' as SCODE,1 POOLID,date\'2010-01-03\' as EFFECTIVE_DAY,1 as ADJUSTMODE from dual
      union all
      select \'00001\' as SCODE,1 POOLID,date\'2010-01-05\' as EFFECTIVE_DAY,2 as ADJUSTMODE from dual
      union all
      select \'00001\' as SCODE,1 POOLID,date\'2010-01-07\' as EFFECTIVE_DAY,2 as ADJUSTMODE from dual
      union all
      select \'00001\' as SCODE,1 POOLID,date\'2010-02-01\' as EFFECTIVE_DAY,1 as ADJUSTMODE from dual
      union all
      select \'00001\' as SCODE,1 POOLID,date\'2010-02-03\' as EFFECTIVE_DAY,1 as ADJUSTMODE from dual
      union all
      select \'00001\' as SCODE,1 POOLID,date\'2010-02-07\' as EFFECTIVE_DAY,2 as ADJUSTMODE from dual
      union all
      select \'00002\' as SCODE,1 POOLID,date\'2010-01-01\' as EFFECTIVE_DAY,1 as ADJUSTMODE from dual
      union all
      select \'00002\' as SCODE,1 POOLID,date\'2010-01-05\' as EFFECTIVE_DAY,2 as ADJUSTMODE from dual
      union all
      select \'00002\' as SCODE,1 POOLID,date\'2010-01-07\' as EFFECTIVE_DAY,2 as ADJUSTMODE from dual
      union all
      select \'00002\' as SCODE,1 POOLID,date\'2010-02-01\' as EFFECTIVE_DAY,1 as ADJUSTMODE from dual
      union all
      select \'00002\' as SCODE,1 POOLID,date\'2010-02-03\' as EFFECTIVE_DAY,1 as ADJUSTMODE from dual
      union all
      select \'00002\' as SCODE,1 POOLID,date\'2010-02-07\' as EFFECTIVE_DAY,2 as ADJUSTMODE from dual
)  
,data as (
select x.*,rn2-rn,row_number() over (partition by scode,poolid,adjustmode,rn2-rn order by rn) as rn3
from (      
  select SCODE,POOLID,EFFECTIVE_DAY,ADJUSTMODE
         ,row_number() over (partition by SCODE,POOLID,ADJUSTMODE order by EFFECTIVE_DAY) as rn
         ,row_number() over (partition by SCODE,POOLID order by EFFECTIVE_DAY) as rn2
  from  demo t 
  ) x
)
select * from data where rn3=1
order by 1,2,3