oracle 下实现累计求和

row_number() OVER (PARTITION BY COL1 ORDER BY COL2) 表示根据COL1分组，在分组内部根据 COL2排序，而此函数计算的值就表示每组内部排序后的顺序编号（组内连续的唯一的). 
 

与rownum的区别在于：使用rownum进行排序的时候是先对结果集加入伪列rownum然后再进行排序，而此函数在包含排序从句后是先排序再计算行号码．
 

row_number()和rownum差不多，功能更强一点（可以在各个分组内从1开时排序）． 

rank()是跳跃排序，有两个第二名时接下来就是第四名（同样是在各个分组内）． 

dense_rank()l是连续排序，有两个第二名时仍然跟着第三名。相比之下row_number是没有重复值的 ．
lag（arg1,arg2,arg3): 
 

 arg1是从其他行返回的表达式
arg2是希望检索的当前行分区的偏移量。是一个正的偏移量，时一个往回检索以前的行的数目。
arg3是在arg2表示的数目超出了分组的范围时返回的值。

在oracle开发过程中经常会遇到累加、减、乘、除的问题。
下面对这个做个小的总结

--创建测试表
create table aspen(parent_id number,part_id number,qnty number);
insert into aspen values(1,1,2);
insert into aspen values(1,2,3);
insert into aspen values(1,3,2);
insert into aspen values(1,4,5);
insert into aspen values(2,2,3);
insert into aspen values(2,3,5);
insert into aspen values(2,4,7);

--实现累加
select parent_id,part_id,qnty,
sum(qnty) over(partition by parent_id order by part_id) running_prod
from aspen

 

 PARENT_ID    PART_ID       QNTY RUNNING_PROD
---------- ---------- ---------- ------------
         1          1          2            2
         1          2          3            5
         1          3          2            7
         1          4          5           12
         2          2          3            3
         2          3          5            8
         2          4          7           15

--实现累减
select parent_id,part_id,qnty,
 sum(decode(rn,1,qnty,-qnty)) over(partition by parent_id order by part_id) running_prod
from (select parent_id,part_id,qnty,
      row_number() over(partition by parent_id order by part_id) rn
      from aspen)

 PARENT_ID    PART_ID       QNTY RUNNING_PROD
---------- ---------- ---------- ------------
         1          1          2            2
         1          2          3           -1
         1          3          2           -3
         1          4          5           -8
         2          2          3            3
         2          3          5           -2
         2          4          7           -9

--实现累乘
select parent_id,part_id,
exp(sum(ln(qnty)) over(partition by parent_id order by part_id)) running_prod
from aspen

PARENT_ID    PART_ID RUNNING_PROD
--------- ---------- ------------
        1          1            2
        1          2            6
        1          3           12
        1          4           60
        2          2            3
        2          3           15
        2          4          105

--实现累除
select parent_id,part_id,
exp(sum(decode(rn,1,ln(qnty),-ln(qnty))) over(partition by parent_id order by part_id)) running_prod
from (select parent_id,part_id,qnty,
      row_number() over(partition by parent_id order by part_id) rn
      from aspen )

 PARENT_ID    PART_ID RUNNING_PROD
---------- ---------- ------------
         1          1            2
         1          2   .666666667
         1          3   .333333333
         1          4   .066666667
         2          2            3
         2          3           .6
         2          4   .085714286

row_number() OVER (PARTITION BY COL1 ORDER BY COL2) 和sum(qnty) over(partition by parent_id order by part_id) 可实现累加，其余类似。