使用 SQL 聚合基于不同月份的年度数据

Question

我有一个“金额”表，其中包含客户的每月付款。每个客户的每笔付款都有一行。我想从他们第一次支付的月份开始每年汇总他们的付款。例如，在下表中，对于用户 ID 132，我想将他从 2019 年第 9 个月到 2020 年第 8 个月（一整年）的付款汇总为一行，然后从 2020 年第 9 个月到下一个汇总为另一行行。

基本上，我希望根据他们加入的月份将每年的用户数量作为行。我不确定如何使用 SQL 聚合这些数据，希望能在此提供帮助。

示例表（如果更简单，我可以将年月列合并为原始数据本身的日期列）>

+--------+------+-------+--------+
| userID | year | month | amount |
+--------+------+-------+--------+
| 132    | 2019 | 9     | 836    |
+--------+------+-------+--------+
| 132    | 2019 | 10    | 702    |
+--------+------+-------+--------+
| 132    | 2019 | 11    | 161    |
+--------+------+-------+--------+
| 132    | 2019 | 12    | 955    |
+--------+------+-------+--------+
| 132    | 2020 | 1     | 969    |
+--------+------+-------+--------+
| 132    | 2020 | 2     | 977    |
+--------+------+-------+--------+
| 132    | 2020 | 3     | 986    |
+--------+------+-------+--------+
| 132    | 2020 | 4     | 639    |
+--------+------+-------+--------+
| 132    | 2020 | 5     | 411    |
+--------+------+-------+--------+
| 132    | 2020 | 6     | 302    |
+--------+------+-------+--------+
| 132    | 2020 | 7     | 929    |
+--------+------+-------+--------+
| 132    | 2020 | 8     | 884    |
+--------+------+-------+--------+
| 132    | 2020 | 9     | 644    |
+--------+------+-------+--------+
| 132    | 2020 | 10    | 640    |
+--------+------+-------+--------+
| 132    | 2020 | 11    | 121    |
+--------+------+-------+--------+
| 132    | 2020 | 12    | 980    |
+--------+------+-------+--------+
| 1447   | 2020 | 11    | 356    |
+--------+------+-------+--------+
| 1447   | 2020 | 12    | 351    |
+--------+------+-------+--------+

样本输出（这里的年周期列只是为了表明自用户加入以来总数属于哪一年）。>

+--------+------------+----------------------+
| userID | Year Cycle | Current Total Amount |
+--------+------------+----------------------+
| 132    | 1          | 8751                 |
+--------+------------+----------------------+
| 132    | 2          | 2385                 |
+--------+------------+----------------------+
| 1447   | 1          | 707                  |
+--------+------------+----------------------+

Answer 1

参考资料 -

1. Common Table Expressions (CTE), https://www.sqlshack.com/sql-server-common-table-expressions-cte/
2. SELECT - OVER Clause

使用CTE 如下所示获得您想要的结果。

获取日期列以获取具有第一个 CTE AmountWithDate 的每个用户的最短日期。

然后使用之前的CTE AmountWithDate 创建另一个CTE AmountWithYearDifference 以从用户的第一笔交易中获取years。必须使用over(partition by ...)

然后使用AmountWithYearDifference 与GROUP BY 与userid & dt 和SUM(Amount) 以获得所需的输出。

;WITH AmountWithDate AS (
    SELECT *, DATEFROMPARTS(year, month, 1) AS dt 
    FROM Amounts
)
, AmountWithYearDifference AS (
    SELECT *, DATEDIFF(MONTH, (min(dt) over(partition by userid)), dt) / 12 AS years
    FROM AmountWithDate
)
SELECT userid, 
        min(dt) AS dt, DATEPART(MONTH, MIN(dt)) as month, 
        DATEPART(YEAR, MIN(dt)) as year, 
        sum(Amount)
FROM AmountWithYearDifference
GROUP BY userid, years

---

编辑如果您已经拥有date 列，则无需先拥有CTE，您可以直接使用第二个CTE。我假设您的date 列名是dt。

;WITH AmountWithYearDifference AS (
    SELECT *, DATEDIFF(MONTH, (min(dt) over(partition by userid)), dt) / 12 AS years
    FROM Amounts
)
SELECT userid, 
        min(dt) AS dt, DATEPART(MONTH, MIN(dt)) as month, 
        DATEPART(YEAR, MIN(dt)) as year, 
        sum(Amount)
FROM AmountWithYearDifference
GROUP BY userid, years

如果您不熟悉CTE 并希望使用inner sql query，那么您可以编写如下查询。只需在inner query 中从CTE 写入query，如下所示。

SELECT userid, 
        min(dt) AS dt, DATEPART(MONTH, MIN(dt)) as month, 
        DATEPART(YEAR, MIN(dt)) as year, 
        sum(Amount)
FROM (
    SELECT *, DATEDIFF(MONTH, (min(dt) over(partition by userid)), dt) / 12 AS years
    FROM Amounts
)
GROUP BY userid, years

Answer 2

您可以使用row_number()为每个用户生成一个序列一个数字，然后将每12个分组为1个循环

select userId, cycle, sum(amount)
from
(
    select *, 
           cycle = (row_number() over (partition by userId 
                                           order by year, month) - 1) / 12 + 1
    from   Amounts
) t
group by userId, cycle

db<>fiddle demo