當前記錄和下一條記錄之間的列總和

Question

稍微難以解釋，我的SQL Server不是最好的，但任何事情都可以在這裏完成。

首先，創建一些表：

CREATE TABLE [dbo].[Quarterly](
[QuarterDate] [datetime] NOT NULL, 
[SomeText] [nvarchar](50) NULL, 
CONSTRAINT [PK_Quarterly] PRIMARY KEY CLUSTERED 
(
[QuarterDate] ASC 
) 
GO 

CREATE TABLE [dbo].[TmpDegreeDays](
[Date] [datetime] NOT NULL, 
[Value] [nvarchar](50) NOT NULL, 
CONSTRAINT [PK_TmpDegreeDays] PRIMARY KEY CLUSTERED 
(
[Date] ASC 
) 
GO 

然後插入一些數據：

INSERT [dbo].[Quarterly] ([QuarterDate], [SomeText]) VALUES (CAST(0x00009CF100000000 AS DateTime), N'Blah') 
INSERT [dbo].[Quarterly] ([QuarterDate], [SomeText]) VALUES (CAST(0x00009D4B00000000 AS DateTime), N'Fools') 
INSERT [dbo].[Quarterly] ([QuarterDate], [SomeText]) VALUES (CAST(0x00009DA600000000 AS DateTime), N'Later') 
INSERT [dbo].[Quarterly] ([QuarterDate], [SomeText]) VALUES (CAST(0x00009E0400000000 AS DateTime), N'Something') 
INSERT [dbo].[Quarterly] ([QuarterDate], [SomeText]) VALUES (CAST(0x00009E5E00000000 AS DateTime), N'New year') 
INSERT [dbo].[Quarterly] ([QuarterDate], [SomeText]) VALUES (CAST(0x00009EC300000000 AS DateTime), N'In april') 

然後從2010-01-01插入日期範圍至（含）2012-03-10成表TmpDegreeDays

最後：

我想計算對於當前QuarterDate與Quarterly結果集中下一條記錄之間的Quarterly表中的每條記錄，請使用[值]的和SUM。

喜歡的東西：最終輸出的

DECLARE @startDate datetime, @endDate datetime 
SET @startDate = '2010-01-01' 
SET @endDate = '2010-12-31' 

SELECT q.QuarterDate, q.SomeText, CustomSum = 
(SELECT SUM(CAST([Value] AS float)) 
FROM TmpDegreeDays 
WHERE [date] >= q.QuarterDate AND *Current QuarterDate* < *Some query here to get next row QuarterDate*) 

FROM Quarterly q 

WHERE q.QuarterDate BETWEEN @startDate AND @endDate 

例子我要找：

2010-01-01 Sum of [Value] between 2010-01-01 and 2010-03-31 
2010-04-01 Sum of [Value] between 2010-04-01 and 2010-06-30 
2010-07-01 Sum of [Value] between 2010-07-01 and 2010-09-31 
2010-10-03 Sum of [Value] between 2010-10-03 and 2010-10-03 

這是否有道理？

Answer 1

你可以嘗試這樣的事情 - 一個使用CTE（可在SQL Server 及更高版本）發現的日期每個季度，然後總結度值：

;WITH QuarterlyDates AS 
(  
    SELECT 
     QuarterDate, 
     QuarterEndDate = DATEADD(MILLISECOND, -3, DATEADD(MONTH, 3, QuarterDate)) 
    FROM [dbo].[Quarterly] 
) 
SELECT 
    qd.QuarterDate, qd.QuarterEndDate, 
    DegreeSum = (SELECT SUM(DegreeValue) 
       FROM [dbo].[TmpDegreeDays] 
       WHERE [Date] BETWEEN qd.QuarterDate AND qd.QuarterEndDate) 
FROM 
    QuarterlyDates qd 

奇怪功能確定QuarterEndDate的根源在於SQL Server中的DATETIME的精度爲3.33ms。因此，給定月份的最後日期是該月份的最後一天，時間爲23:59:59.997（不是.999）。因此，我需要在本季度的開始日期前增加三個月，然後從下一季度的開始日期減去3毫秒以獲取本季度的最後一個毫秒。確定，爲了從表格中得到一個季度的開始和結束日期，需要兩個嵌套的CTE以確定「下一個季度的結束日期減去3毫秒」本季度 - 這樣的事情：

;WITH QuarterStarts AS 
(
    SELECT 
     QuarterDate, 
     QNumber = ROW_NUMBER() OVER(ORDER BY QuarterDate) -- ordering number 
    FROM [dbo].[Quarterly] 
), 
Quarters AS 
(
    SELECT 
     QuarterStartDate = q1.QuarterDate, 
     QuarterEndDate = DATEADD(MILLISECOND, -3, q2.QuarterDate) 
    FROM 
     QuarterStarts q1 
    INNER JOIN 
     QuarterStarts q2 ON q2.QNumber = q1.QNumber + 1 
) 
SELECT 
    q.QuarterDate, q.QuarterEndDate, 
    DegreeSum = (SELECT SUM(DegreeValue) 
       FROM [dbo].[TmpDegreeDays] 
       WHERE [Date] BETWEEN q.QuarterDate AND q.QuarterEndDate) 
FROM 
    Quarters q