SUM對列組

Question

我有結構類似

tasktime - starttime,endtime,packagedetailid,packageid 
packagedetailid - packageid,productid, etc 
productime - productid and searchesperday 

，我使用這些查詢三個表如下所示

SELECT CONVERT(varchar, t.StartTime, 111) AS 'Date' 
, SUM(DATEDIFF(second, t.StartTime, t.EndTime))/60 AS DailyMinutes 
, COUNT(DISTINCT p.PackageDetailID) AS OrderCount 
, LEFT(CAST(SUM(DATEDIFF(minute, t.StartTime, t.EndTime)) 
    /(COUNT(DISTINCT p.PackageDetailID) + 0.0000001) AS varchar), 4) AS PerOrder 
, LEFT(CAST((COUNT(DISTINCT p.PackageDetailID) * 100) 
    /(e.SearchesPerDay + 0.00000001) AS varchar), 4) AS WeightedOrderCount 
FROM ProductTime AS e INNER JOIN dbo.PackageDetails AS p 
WITH (NOLOCK) ON e.ProductID = p.ProductID INNER JOIN 
TaskTime AS t WITH (NOLOCK) ON p.PackageDetailID = t.PackageDetailId 
WHERE(t.EndTime IS NOT NULL) 
AND (DATEDIFF(hour, t.StartTime, t.EndTime) < 1) 
AND (DATEDIFF(second, t.StartTime, t.EndTime) > 0) 
AND (t.UserId = '12345') 
AND (t.StartTime BETWEEN '1/1/2010' 
AND '9/13/2010') 
GROUP BY CONVERT(varchar, t.StartTime, 111), e.SearchesPerDay 
ORDER BY CONVERT(varchar, t.StartTime, 111) 

輸出爲：

**date   dailyminutes ordercount perorder weightordercount** 

2010/01/04  104    26 4.03 43.30 
2010/01/04  10     1 9.99  1.24 
2010/01/04  19     8 2.37  8.88 
2010/01/04  22     11 1.99 10.90 
2010/01/04  18     5 3.59  2.77 
2010/01/05  49     17 2.99 28.30 
2010/01/05  31     5 6.39  5.55 
2010/01/05  26     6 4.33  5.99 
2010/01/05  8     4 1.99  3.33 

但我想整理日期"Weightordercount"，以便輸出爲

**date   dailyminutes ordercount perorder weightordercount** 

2010/01/04   173   51  21.97  67.09 
2010/01/05   114   32  15.70  43.17 

我不是一個SQL專家，需要你的幫助，如果這可以通過一個SQL命令或通過Strored過程

感謝提前實現

Answer 1

（1）Don't use varchar without length - 但爲什麼你轉換爲varchar呢？

（2）你爲什麼每天按搜索進行分組？我猜這必須是一個SUM是有意義的，這裏是我會怎麼做：

;WITH x AS 
(
    SELECT 
     [Date]   = DATEDIFF(DAY, '19000101', t.StartTime), 
     DailyMinutes = SUM(DATEDIFF(SECOND, t.StartTime, t.EndTime)), 
     OrderCount  = COUNT(DISTINCT p.PackageDetailID), 
     SearchesPerDay = SUM(e.SearchesPerDay) 
    FROM 
     dbo.ProductTime AS e 
    INNER JOIN 
     dbo.PackageDetails AS p 
     ON e.ProductID = p.ProductID 
    INNER JOIN 
     dbo.TaskTime AS t 
     ON p.PackageDetailID = t.PackageDetailID 
    WHERE 
     t.EndTime IS NOT NULL 
     AND (DATEDIFF(SECOND, t.StartTime, t.EndTime) BETWEEN 1 AND 3599) 
     AND (t.UserId = '12345') 
     AND (t.StartTime >= '20100101' AND t.StartTime <= '20100913') 
    GROUP BY 
     DATEDIFF(DAY, '19000101', t.StartTime) 
) 
SELECT 
    [Date] = DATEADD(DAY, [Date], '19000101'), 
    DailyMinutes, 
    OrderCount, 
    PerOrder = CONVERT(DECIMAL(10,2), (DailyMinutes * 1.0/OrderCount)), 
    WeightedOrderCount = CONVERT(DECIMAL(10,2), (100.0 * OrderCount/SearchesPerDay)) 
FROM 
    x 
ORDER BY 
    [Date]; 

一對夫婦的其他增強功能：

（一）日期不之間使用的/時間，使用> =和<（我離開最後的< =但很可能你的意思是< 9/14或< 9/13）。 （b）不使用區域格式的日期 - '09/13/2010'不取決於語言，日期格式或區域設置的有效日期。 'YYYYMMDD'是在SQL Server中僅用於日期的最安全格式。 （c）嘗試儘可能少地執行計算 - 我將重複計算移入CTE，以便其他計算中的參考可以更簡單。你不需要CTE，你也可以使用子查詢。這也是我將兩次檢查與StartTime和EndTime之間的增量相結合的原因。 （d）您獲取小數的方法相當粗糙 - 增加小數位，轉換爲字符串，取左...

Answer 2

你的代碼是可怕的格式化您的文章，所以我打算給你一個一般的答案：

你需要使用GROUP BY條款，然後SUM每個聚合。

因此，像這樣：

select date, sum(dailyminutes), sum(ordercount), sum(perorder), sum(weightordercount) 
from (yourentirequery) a 
group by date 
order by date asc