窗口函数 - 日期之间的范围

Posted 2023-03-31

【中文标题】窗口函数 - 日期之间的范围

【英文标题】：Window Function - Range Between Dates

【发布时间】：2015-02-20 14:56:48

【问题描述】：

我收到错误消息，我的代码 (Oracle) 中缺少关键字。我的想法是创建一个 as/was 的支出 last3,6,12 视图。因此，我想知道 jan14 的 last3 花费是多少，最后 3 个月中包含的月份是：Oct13、Nov13、Dec14。

使用合并到 account_month_tb （选择 account_month_tb.acct_id、account_month_tb.month_start_dt，

--rolling spend buckets
sum(spend) over
(partition by account_month_tb.acct_id order by account_month_tb.month_start_dt
range between interval '3' month preceding and preceding month) as spend_last3,

sum(spend) over
(partition by account_month_tb.acct_id order by account_month_tb.month_start_dt
range between interval '6' month preceding and preceding month) as spend_last6,

sum(spend) over
(partition by account_month_tb.acct_id order by account_month_tb.month_start_dt
range between interval '12' month preceding and preceding month) as spend_last12
--mkdn buckets


from account_month_tb

 ) rolling_14
on (account_month_tb.acct_id = rolling_14.acct_id and
account_month_tb.month_start_dt = rolling_14.month_start_dt)
when matched then update set

account_month_tb.spend_last3 = rolling_14.spend_last3,
account_month_tb.spend_last6 = rolling_14.spend_last6,
account_month_tb.spend_last12 = rolling_14.spend_last12
;
commit;

错误信息：

命令行错误：163 列：62 错误报告 - SQL 错误：ORA-00905：缺少关键字 00905. 00000 - “缺少关键字” *原因： *行动： 承诺。

【问题讨论】：

也许显示您正在执行的整个语句以及完整的错误消息会有所帮助？如果这是一种观点，你似乎在发明新的语法。

为什么会被否决？

所以and preceding month 应该是and interval '1' month preceding？

【参考方案1】：

在windowing clause 中，preceding month 没有任何意义且无效。你的意思似乎是interval '1' month preceding。

在 CTE 中使用一些虚拟数据进行模拟：

with account_month_tb as (
  select 1 as acct_id,
    trunc(sysdate - (16 * level), 'MM') as month_start_dt,
    level as spend
  from dual
  connect by level < 10
)
select account_month_tb.acct_id, account_month_tb.month_start_dt,
  sum(spend) over (partition by account_month_tb.acct_id
    order by account_month_tb.month_start_dt
    range between interval '3' month preceding
      and interval '1' month preceding) as spend_last3
from account_month_tb;

   ACCT_ID MONTH_START_DT SPEND_LAST3
---------- -------------- -----------
         1 01-09-2014                 
         1 01-10-2014               9 
         1 01-10-2014               9 
         1 01-11-2014              24 
         1 01-12-2014              30 
         1 01-12-2014              30 
         1 01-01-2015              30 
         1 01-01-2015              30 
         1 01-02-2015              20 

SQL Fiddle 包含三个总和和更多数据。但是，您会得到很多重复，这可能不是您想要的。也许you just need distinct values。