Question

我有一个订单表，每行都有：

订单ID
订单签出日期
订单交易价值

我希望按日期（yyyy-mm-dd）对该期间的交易值进行分组，但我也想返回该期间的运行总额。如果我针对这几个月的数据（到11月的13天）运行此查询，那么我期望看到返回的结果集

13行，每月的每一天一次，
当月某天的交易值
到目前为止该月的运行总计

我可以使用分组依据获取交易值分组依据日期，但没有运行总计。如果我在分区上使用，则每笔交易我将获得1行，但当天的交易价值之和。

当前代码如下：

SELECT 
    left(convert(varchar(30),OrderCheckOutDate,120),10), 
    SUM (OrderTotalFixed) 
    OVER (Partition By left(convert(varchar(30),OrderCheckOutDate,120),10))
FROM 
    tblOrder
Where
    UserExternalUserLevel = 0 
And
    OrderCheckOutDate > '2019-11-01'

好像查询正在按毫秒进行分区，但我认为转换将处理此问题？

Answer 1

要按日期分组，您不需要OVER (Partition By)并将该列放在“分组依据”下

尝试一下：

;WITH TotalByDay AS (
SELECT 
    convert(DATE,OrderCheckOutDate) OrderCheckOutDate ,
    SUM (OrderTotalFixed) OrderTotalFixed 
FROM 
    tblOrder
Where
    UserExternalUserLevel = 0 
And
    OrderCheckOutDate > '2019-11-01'
GROUP BY  convert(DATE,OrderCheckOutDate)
)

SELECT 
   OrderCheckOutDate,
   (SELECT SUM(TotalByDay) FROM TotalByDay innerT where innerT.OrderCheckOutDate <= T.OrderCheckOutDate) RunningTotal
FROM TotalByDay t
GROUP BY t.OrderCheckOutDate

Answer 2

与其他示例相比，这里的版本缩小了代码的大小。它也应该比某些方法快一点，因为没有子查询会强制多次读取表。使用运行总计和group by语句时要记住的技巧是，您不将window函数应用于原始列，而是应用于原始列的汇总（因此我将SUM嵌套在SUM中）。

SELECT 
    CAST(OrderCheckOutDate AS date) AS OrderCheckOutDate, 
    SUM (OrderTotalFixed) AS DayTotal,
    SUM (SUM (OrderTotalFixed)) OVER (ORDER BY CAST(OrderCheckOutDate AS date)) AS RunningTotal
FROM 
    tblOrder
Where
    UserExternalUserLevel = 0 
And
    OrderCheckOutDate > '2019-11-01'
GROUP BY CAST(OrderCheckOutDate AS date)

您可以在dbfiddle

上看到此功能

例如，如果您的实际用例更加复杂，则您希望每个月重置运行总额，可以通过将PARTITION添加到窗口函数中来处理。

按月分区将如下所示：

SELECT 
    CAST(OrderCheckOutDate AS date) AS OrderCheckOutDate,
    EOMONTH(CAST(OrderCheckOutDate AS date)) AS MonthEnd,
    SUM (OrderTotalFixed) AS DayTotal,
     SUM (SUM (OrderTotalFixed)) OVER (PARTITION BY EOMONTH(CAST(OrderCheckOutDate AS date)) ORDER BY CAST(OrderCheckOutDate AS date) ) AS RunningTotal
FROM 
    tblOrder
Where
    UserExternalUserLevel = 0 
And
    OrderCheckOutDate > '2019-11-01'
GROUP BY CAST(OrderCheckOutDate AS date)

再一次看到它与dbfiddle一起起作用

Answer 3

这是我做过的类似事情，希望您可以根据自己的餐桌/需求进行调整。内部查询（TOTALS_Q）仅按月获取销售和计数。外部查询也会获取累计值。

累加的重要部分是“未绑定的行和当前行之间的行”

SELECT 
  SaleMonth, Month_Order_Count, MonthTotal,   
  SUM(Month_Order_Count) OVER(ORDER BY SaleMonth ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) AS Cumulative_Order_Count,
  SUM(MonthTotal) OVER(ORDER BY SaleMonth ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) AS Cumulative_Order_Value
FROM 
(
  SELECT SUB_Q.SaleMonth, Count( DISTINCT SUB_Q.OrderID) as Month_Order_Count, SUM(Value) as MonthTotal  FROM
  (
    SELECT CONCAT( DatePart( year, OH.OrderDate), '-', FORMAT( OH.OrderDate, 'MM')) as SaleMonth,
    OI.OrderID,
    OI.Value
  FROM ORDERHEADER OH 
  INNER JOIN OrderItem OI ON OI.OrderID = OH.OrderID
  INNER JOIN ECodeOrder ECO ON ECO.OrderId = OH.ORDERID
  INNER JOIN Product P ON P.ProductId = OI.ProductId AND P.Denomination >= 10000
  ) SUB_Q
  GROUP BY SUB_Q.SaleMonth
)TOTALS_Q
ORDER BY SaleMonth

按日期总计加上运行总计

3 个答案: