创建滚动期间答案

【问题标题】：Creating Rolling Periods创建滚动期间
【发布时间】：2021-02-23 21:27:29
【问题描述】：

我希望每 6 个月创建一个滚动期，但我不确定执行此操作的最佳方式。我猜这可能必须递归完成？

我有一个包含类似于以下数据的付款表：

CREATE TABLE payments
    ([id] int, [payment_month] int, [payment_date] datetime, [payment_amount] int)
;
    
INSERT INTO payments
    ([id], [payment_month], [payment_date], [payment_amount])
VALUES
    (1, 201911, '2019-11-01 00:00:00', 50),
    (1, 201912, '2019-12-01 00:00:00', 50),
    (1, 202001, '2020-01-01 00:00:00', 50),
    (1, 202002, '2020-02-01 00:00:00', 50),
    (1, 202003, '2020-03-01 00:00:00', 50),
    (1, 202004, '2020-04-01 00:00:00', 50),
    (1, 202005, '2020-05-01 00:00:00', 50),
    (1, 202006, '2020-06-01 00:00:00', 25),
    (1, 202007, '2020-07-01 00:00:00', 50),
    (1, 202008, '2020-08-01 00:00:00', 50),
    (1, 202009, '2020-09-01 00:00:00', 15),
    (2, 201911, '2019-11-01 00:00:00', 50),
    (2, 201912, '2019-12-01 00:00:00', 50),
    (2, 202001, '2020-01-01 00:00:00', 25),
    (2, 202002, '2020-02-01 00:00:00', 50),
    (2, 202003, '2020-03-01 00:00:00', 45),
    (2, 202004, '2020-04-01 00:00:00', 45),
    (2, 202004, '2020-04-10 00:00:00', 20),
    (2, 202005, '2020-05-01 00:00:00', 25),
    (3, 202004, '2020-04-01 00:00:00', 50),
    (3, 202005, '2020-05-01 00:00:00', 50),
    (3, 202006, '2020-06-01 00:00:00', 50),
    (3, 202007, '2020-07-01 00:00:00', 50),
    (3, 202008, '2020-08-01 00:00:00', 50),
    (3, 202009, '2020-09-01 00:00:00', 300)
;

我还有一个日历表，我正在使用它返回如下数据：

CREATE TABLE calendar
    ([CalendarPeriod] int, [CalendarDate] datetime)
;
    
INSERT INTO calendar
    ([CalendarPeriod], [CalendarDate])
VALUES
    (202004, '2020-04-30 00:00:00'),
    (202005, '2020-05-31 00:00:00'),
    (202006, '2020-06-30 00:00:00'),
    (202007, '2020-07-31 00:00:00'),
    (202008, '2020-08-31 00:00:00'),
    (202009, '2020-09-30 00:00:00')
;

我希望得到一个输出，其中来自calendar 的每个CalendarPeriod 与其自身和前5个月配对。 CalendarPeriod作为报告期，每个月的滚动期为6个月。我的最终目标是通过payment_month 将payment_amounts 加入每个报告期，其中payment_month 等于报告期中的滚动月份之一。

编辑： 我预期的几个月和滚动期的产出如下表所示。如果需要，我可以操作表格以通过 CalendarPeriod 聚合。

+----------------+----------------+----------------+
| CalendarPeriod | rolling_period | rolling_amount |
+----------------+----------------+----------------+
| 202004         | 201911         | 100            |
+----------------+----------------+----------------+
| 202004         | 201912         | 100            |
+----------------+----------------+----------------+
| 202004         | 202001         | 75             |
+----------------+----------------+----------------+
| 202004         | 202002         | 100            |
+----------------+----------------+----------------+
| 202004         | 202003         | 95             |
+----------------+----------------+----------------+
| 202004         | 202004         | 165            |
+----------------+----------------+----------------+
| 202005         | 201912         | 100            |
+----------------+----------------+----------------+
| 202005         | 202001         | 75             |
+----------------+----------------+----------------+
| 202005         | 202002         | 100            |
+----------------+----------------+----------------+
| 202005         | 202003         | 95             |
+----------------+----------------+----------------+
| 202005         | 202004         | 165            |
+----------------+----------------+----------------+
| 202005         | 202005         | 125            |
+----------------+----------------+----------------+
| 202006         | 202001         | 75             |
+----------------+----------------+----------------+
| 202006         | 202002         | 100            |
+----------------+----------------+----------------+
| 202006         | 202003         | 95             |
+----------------+----------------+----------------+
| 202006         | 202004         | 165            |
+----------------+----------------+----------------+
| 202006         | 202005         | 125            |
+----------------+----------------+----------------+
| 202006         | 202006         | 75             |
+----------------+----------------+----------------+
| 202007         | 202002         | 100            |
+----------------+----------------+----------------+
| 202007         | 202003         | 95             |
+----------------+----------------+----------------+
| 202007         | 202004         | 165            |
+----------------+----------------+----------------+
| 202007         | 202005         | 125            |
+----------------+----------------+----------------+
| 202007         | 202006         | 75             |
+----------------+----------------+----------------+
| 202007         | 202007         | 100            |
+----------------+----------------+----------------+
| 202008         | 202003         | 95             |
+----------------+----------------+----------------+
| 202008         | 202004         | 165            |
+----------------+----------------+----------------+
| 202008         | 202005         | 125            |
+----------------+----------------+----------------+
| 202008         | 202006         | 75             |
+----------------+----------------+----------------+
| 202008         | 202007         | 100            |
+----------------+----------------+----------------+
| 202008         | 202008         | 100            |
+----------------+----------------+----------------+
| 202009         | 202004         | 165            |
+----------------+----------------+----------------+
| 202009         | 202005         | 125            |
+----------------+----------------+----------------+
| 202009         | 202006         | 75             |
+----------------+----------------+----------------+
| 202009         | 202007         | 100            |
+----------------+----------------+----------------+
| 202009         | 202008         | 100            |
+----------------+----------------+----------------+
| 202009         | 202009         | 315            |
+----------------+----------------+----------------+

-- Alternate:
+----------------+----------------+
| CalendarPeriod | rolling_amount |
+----------------+----------------+
| 202004         | 635            |
+----------------+----------------+
| 202005         | 660            |
+----------------+----------------+
| 202006         | 635            |
+----------------+----------------+
| 202007         | 660            |
+----------------+----------------+
| 202008         | 660            |
+----------------+----------------+
| 202009         | 880            |
+----------------+----------------+

提前感谢您的帮助。

【问题讨论】：

您是否尝试过在calendar 和calendar 之间使用join 来匹配6 个月的窗口？
我还没有尝试过，尽管我确信会有办法做到这一点。如果我能避免这么多的连接就更好了，这是一个更大的存储过程的一部分

标签： sql sql-server datetime sum aggregate-functions

【解决方案1】：

据我了解您的问题，您希望每个时期的过去 6 个月的付款金额滚动。

我不认为你真的需要你展示的中间数据集来实现这个目标。您可以只使用窗口函数：

select c.calendarperiod,
    sum(p.payment_amount) month_amount,
    sum(sum(p.payment_amount)) over(
        order by c.calendarperiod
        rows between 5 preceding and current row
    ) rolling_6_month_amount
from calendar c
left join payments p 
    on  p.payment_date >= datefromparts(year(c.calendardate), month(c.calendardate), 1)
    and p.payment_date <  dateadd(day, 1, c.calendardate)
group by c.calendarperiod

查询从日历开始，并带上每个月对应的付款——我调整了日期间隔，以防当月最后一天有付款。然后，我们按时期汇总。最后，我们可以使用窗口函数来回顾过去 5 个时期，并对相应的付款求和。

【讨论】：

这很接近，但我认为总数并不像预期的那样匹配。例如，202004 应该是 635。我想考虑在为CalendarPeriod 支付的任何滚动周期内付款。 202004 应该包含（包括）201911 和 202004 之间的付款数据 -- @GMB 更新
@tlk27：请edit your question 显示您想要的样本数据结果。
@tlk27：感谢您的编辑。问题在于日历表，它从 2020 年 4 月开始，而数据从 2019 年 11 月开始。如果您扩展日历表，那么您将获得所需的结果。看到这个 db fiddle：dbfiddle.uk/…
感谢您的时间和帮助，非常感谢。我不知道rows between x preceding and current 功能