Oracle SQL：滞后于一系列值

Question

我们如何在值的范围内使用分析函数 LAG。如果分区中没有较早 in_date 的记录，则它应该 return null 否则，前一个 in_date

的年月

示例：

CID  IN_DATE        
1    2020-05-29
1    2020-06-10
1    2020-06-21
1    2020-07-08
1    2020-08-11
1    2020-10-01
2    2020-05-05
2    2020-05-03
2    2020-06-01
2    2020-06-02
2    2020-06-03

预期输出，

CID  IN_DATE        LAG
1    2020-05-29     null
1    2020-06-10     2020-05
1    2020-06-21     2020-05
1    2020-07-08     2020-06
1    2020-08-11     2020-07
1    2020-10-01     2020-08
2    2020-05-05     null 
2    2020-05-03     null
2    2020-06-01     2020-05
2    2020-06-02     2020-05
2    2020-06-03     2020-05
2    2020-07-03     2020-06
2    2020-08-13     2020-07

我当前的查询使用 LAG returns 以下

with data as (
select 1  CID,  TO_DATE('2020-05-29','YYYY-MM-DD')  IN_DATE from dual union all
select 1,  TO_DATE('2020-06-10','YYYY-MM-DD') from dual union all
select 1,  TO_DATE('2020-06-21','YYYY-MM-DD')  from dual union all
select 1,  TO_DATE('2020-07-08','YYYY-MM-DD') from dual union all
select 1,  TO_DATE('2020-08-11','YYYY-MM-DD') from dual union all
select 2,  TO_DATE('2020-05-05','YYYY-MM-DD') from dual union all
select 2,  TO_DATE('2020-05-03','YYYY-MM-DD') from dual union all
select 2,  TO_DATE('2020-06-01','YYYY-MM-DD') from dual union all
select 2,  TO_DATE('2020-06-02','YYYY-MM-DD') from dual union all
select 2,  TO_DATE('2020-06-03','YYYY-MM-DD') from dual 
    ) 
    select CID, 
           to_char(TO_DATE(IN_DATE), 'YYYY-MM-DD') IN_DATE, 
           LAG(to_char(TO_DATE(IN_DATE), 'YYYY-MM')) OVER (PARTITION BY CID 
                                                           ORDER BY to_char(TO_DATE(IN_DATE), 'YYYY-MM') ) LAG 
    from data

当前结果

    CID IN_DATE     LAG
1   1   2020-05-29  NULL
2   1   2020-06-10  2020-05
3   1   2020-06-21  2020-06
4   1   2020-07-08  2020-06
5   1   2020-08-11  2020-07
6   2   2020-05-05  NULL
7   2   2020-05-03  2020-05
8   2   2020-06-01  2020-05
9   2   2020-06-02  2020-06
10  2   2020-06-03  2020-06

看起来，LAG 不允许我们使用分区超范围是否有替代方法

Answer 1

记录 7 的滞后字段不为空，因为滞后函数正在查找记录 6。

我可以通过两种方式来澄清您的问题：

如果分区中没有较早 in_date
的记录，您想 return null
如果 in_date 来自当前记录 in_date 之前一个月的分区中没有记录，你想 return null =10=]

Answer 2

如果您想要本月之前的上个月，则不需要 LAG()。我会建议：

with data as (
    select 1  CID,  TO_DATE('2020-05-29','YYYY-MM-DD')  IN_DATE from dual union all
    select 1,  TO_DATE('2020-06-10','YYYY-MM-DD') from dual union all
    select 1,  TO_DATE('2020-06-21','YYYY-MM-DD')  from dual union all
    select 1,  TO_DATE('2020-07-08','YYYY-MM-DD') from dual union all
    select 1,  TO_DATE('2020-08-11','YYYY-MM-DD') from dual union all
    select 2,  TO_DATE('2020-05-05','YYYY-MM-DD') from dual union all
    select 2,  TO_DATE('2020-05-03','YYYY-MM-DD') from dual union all
    select 2,  TO_DATE('2020-06-01','YYYY-MM-DD') from dual union all
    select 2,  TO_DATE('2020-06-02','YYYY-MM-DD') from dual union all
    select 2,  TO_DATE('2020-06-03','YYYY-MM-DD') from dual 
    ) 
select CID, 
       to_char(TO_DATE(IN_DATE), 'YYYY-MM-DD') IN_DATE, 
       TO_CHAR(MAX(IN_DATE) OVER (PARTITION BY CID 
                                  ORDER BY TRUNC(IN_DATE, 'MON')
                                  RANGE BETWEEN UNBOUNDED PRECEDING AND INTERVAL '1' MONTH PRECEDING
                                 ),
               'YYYY-MM') as LAG 
from data;

Here 是一个 db<>fiddle.

Oracle SQL：滞后于一系列值

Oracle SQL: LAG over a range of values

sql

oracle

plsql

analytic-functions

oracle12c