Sql Server 選擇查詢：按月表的總和和最后輸入

Question

我有一張這樣的桌子：

-------------------------------------------------
Id  | Date                    | Paid  | Remaining |
--------------------------------------------------
 1  | 2020-02-02 12:57:59.950 |500.00 |NULL       |
 2  | 2020-02-02 12:57:59.950 |20.00  |NULL       |
 3  | 2020-02-19 12:10:59.950 |300.00 |NULL       |
 4  | 2020-02-20 12:10:59.950 |100.00 |150.00     |
 5  | 2020-02-27 12:10:59.950 |100.00 |70.00      |
 6  | 2020-03-02 12:57:59.950 |500.00 |50.00      |
 7  | 2020-03-19 12:10:59.950 |300.00 |NULL       |
 8  | 2020-03-20 12:10:59.950 |100.00 |20.00      |
 9  | 2020-03-27 12:10:59.950 |60.00  |60.00      |

我想做一個選擇查詢，按月返回每個日期組：

• 支付總額（Sum）
• 按月最后剩余

---

| Date        | TotalPaid   | LastRemaining |
----------------------------------------
| 2020-02-01  |1020.00      | 70.00
| 2020-03-01  |960.00       | 60.00

Answer 1

這是一種方法：

select yyyymm, sum(paid),
       max(case when seqnum = 1 then remaining end) as last_remaining
from (select t.*, yyyymm,
             row_number() over (partition by v.yyyymm order by date desc) as seqnum             
      from t cross apply
           (values (datefromparts(year(date), month(date), 1))
           ) v(yyyymm)
     ) t
group by yyyymm
order by yyyymm;

您實際上也可以在沒有子查詢的情況下執行此操作，但它更混亂：

select distinct v.yyyymm,
       sum(t.paid) over (partition by v.yyyymm) as paid,
       first_value(t.remaining) over (partition by v.yyyymm order by t.date desc) as last_remaining
from t cross apply
      (values (datefromparts(year(date), month(date), 1))
      ) v(yyyymm);

這使用帶有窗口函數的select distinct作為“first()”/“last()”聚合函數的最后一個的變通方法。