SUM由两个不同的GROUP BY组成

时间：2012-03-02 12:19:58

标签： sql sql-server sql-server-2005 tsql

我从报告中得到了错误的结果。也许我错过了一些简单的东西。

该报告是一个内联表值函数，应该计算我们商店中的货物移动以及这些备件的索赔频率（在维修中更换）。

问题：商店表中的不同备件（让我们称之为SP）可以链接到“修复表”（TSP）中的相同备件。我需要SP中每个备件的货物移动以及TSP中每个不同备件的索赔计数。

这是相关部分的非常简化的摘录：

create table #tsp(id int, name varchar(20),claimed int);
create table #sp(id int, name varchar(20),fiTsp int,ordered int);

insert into #tsp values(1,'1235-6044',300);
insert into #tsp values(2,'1234-5678',400);

insert into #sp values(1,'1235-6044',1,30);
insert into #sp values(2,'1235-6044',1,40);
insert into #sp values(3,'1235-6044',1,50);
insert into #sp values(4,'1234-5678',2,60);

WITH cte AS(
    select tsp.id As TspID,tsp.name as TspName,tsp.claimed As Claimed
    ,sp.id As SpID,sp.name As SpName,sp.ordered As Ordered
    from #sp sp inner join #tsp tsp
    on sp.fiTsp=tsp.id
)
SELECT TspName, SUM(Claimed) As Claimed, Sum(Ordered) As Ordered
FROM cte
Group By TspName

drop table #tsp;
drop table #sp;

结果：

TspName       Claimed   Ordered
1234-5678       400       60
1235-6044       900       120

Ordered - 计数是正确的，但Claimed - 计数应为300而不是900，因为TspName ='1235-6044'。

我需要按Tsp.ID分组索赔计数，并按Sp.ID分组，以获取订单计数。但是在一个查询中如何？

修改：实际上TVF看起来像（请注意getOrdered和getClaimed是SVF，而我正在TSP的类别中选择外部选项）：

CREATE FUNCTION [Gambio].[rptReusedStatistics](
     @fromDate datetime
    ,@toDate datetime
    ,@fromInvoiceDate datetime
    ,@toInvoiceDate datetime
    ,@idClaimStatus varchar(50)
    ,@idSparePartCategories varchar(1000)
    ,@idSpareParts varchar(1000)
)

RETURNS TABLE AS
RETURN(
    WITH ExclusionCat AS(
        SELECT idSparePartCategory AS ID From tabSparePartCategory
        WHERE idSparePartCategory IN(- 3, - 1, 6, 172,168)
    ), Report AS(
        SELECT Cat.SparePartCategoryName AS Category
        ,TSP.SparePartDescription AS Part
        ,TSP.SparePartName AS PartNumber
        ,SP.Inventory
        ,Gambio.getGoodsIn(SP.idSparePart,@FromDate,@ToDate) GoodsIn
        ,Gambio.getOrdered(SP.idSparePart,@FromDate,@ToDate) Ordered
        --,CASE WHEN TSP.idSparePart IS NULL THEN 0 ELSE
        --  Gambio.getClaimed(TSP.idSparePart,@FromInvoiceDate,@ToInvoiceDate,@idClaimStatus,NULL)END AS Claimed
        ,CASE WHEN TSP.idSparePart IS NULL THEN 0 ELSE
            Gambio.getClaimed(TSP.idSparePart,@FromInvoiceDate,@ToInvoiceDate,@idClaimStatus,1)END AS ClaimedReused
        ,CASE WHEN TSP.idSparePart IS NULL THEN 0 ELSE
            Gambio.getCostSaving(TSP.idSparePart,@FromInvoiceDate,@ToInvoiceDate,@idClaimStatus)END AS Costsaving
        FROM  Gambio.SparePart AS SP 
        INNER JOIN tabSparePart AS TSP ON SP.fiTabSparePart = TSP.idSparePart 
        INNER JOIN tabSparePartCategory AS Cat 
        ON Cat.idSparePartCategory=TSP.fiSparePartCategory
        WHERE Cat.idSparePartCategory NOT IN(SELECT ID FROM ExclusionCat)
        AND (@idSparePartCategories IS NULL 
            OR TSP.fiSparePartCategory IN(
                SELECT Item From dbo.Split(@idSparePartCategories,',')
            )
        )
        AND (@idSpareParts IS NULL 
            OR TSP.idSparePart IN(
                SELECT Item From dbo.Split(@idSpareParts,',')
            )
        )
    )
    SELECT Category
    --, Part
    --, PartNumber
    , SUM(Inventory)As InventoryCount
    , SUM(GoodsIn) As GoodsIn
    , SUM(Ordered) As Ordered
    --, SUM(Claimed) As Claimed
    , SUM(ClaimedReused)AS ClaimedReused
    , SUM(Costsaving) As Costsaving
    , Count(*) AS PartCount
    FROM Report
    GROUP BY Category
)

解决方案：

感谢Aliostad，我通过先分组然后加入（实际TVF，减少到最低限度）来解决它：

WITH Report AS(
        SELECT Cat.SparePartCategoryName AS Category
        ,TSP.SparePartDescription AS Part
        ,TSP.SparePartName AS PartNumber
        ,SP.Inventory
        ,SP.GoodsIn
        ,SP.Ordered
        ,Gambio.getClaimed(TSP.idSparePart,@FromInvoiceDate,@ToInvoiceDate,@idClaimStatus,1) AS ClaimedReused
        ,Gambio.getCostSaving(TSP.idSparePart,@FromInvoiceDate,@ToInvoiceDate,@idClaimStatus) AS Costsaving
        FROM  (
            SELECT GSP.fiTabSparePart 
            ,SUM(GSP.Inventory)AS Inventory
            ,SUM(Gambio.getGoodsIn(GSP.idSparePart,@FromDate,@ToDate))AS GoodsIn
            ,SUM(Gambio.getOrdered(GSP.idSparePart,@FromDate,@ToDate))AS Ordered
            FROM Gambio.SparePart GSP
            GROUP BY GSP.fiTabSparePart
        )As SP
        INNER JOIN tabSparePart TSP ON  SP.fiTabSparePart = TSP.idSparePart
        INNER JOIN tabSparePartCategory AS Cat 
        ON Cat.idSparePartCategory=TSP.fiSparePartCategory
    )
    SELECT Category
    , SUM(Inventory)As InventoryCount
    , SUM(GoodsIn) As GoodsIn
    , SUM(Ordered) As Ordered
    , SUM(ClaimedReused)AS ClaimedReused
    , SUM(Costsaving) As Costsaving
    , Count(*) AS PartCount
    FROM Report
    GROUP BY Category

5 个答案:

答案 0 :(得分：3)

您先加入，然后进行GROUPing by。您需要先将其反转，然后再使用GROUP BY，然后再加入。

所以在我的子查询中，我首先分组，然后加入：

select 
    claimed,
    ordered 
from 
     #tsp 
inner JOIN
      (select 
             fitsp, 
             SUM(ordered) as ordered 
       from 
              #sp
        group by 
              fitsp) as SUMS           
on 
     SUMS.fiTsp = id;

答案 1 :(得分：3)

我认为你只需要选择Claimed并将其添加到Group By中以获得你想要的东西。

WITH cte AS(
select tsp.id As TspID,tsp.name as TspName,tsp.claimed As Claimed
,sp.id As SpID,sp.name As SpName,sp.ordered As Ordered
from #sp sp inner join #tsp tsp
on sp.fiTsp=tsp.id )

SELECT TspName, Claimed, Sum(Ordered) As Ordered
FROM cte
Group By TspName, Claimed

答案 2 :(得分：1)

您的cte是tsp和sp之间的内部联接，这意味着您查询的数据如下所示：

SpID   Ordered   TspID   TspName     Claimed
1      30        1       1235-6044   300
2      40        1       1235-6044   300
3      50        1       1235-6044   300
4      60        2       1234-5678   400

请注意TspID，TspName和Claimed如何重复。按TspName分组意味着数据分为两组，一组用于1235-6044，另一组用于1234-5678。第一组有3行用于运行聚合函数，第二组只有一行。这就是为什么你的sum(Claimed)会让你获得300 * 3 = 900。

正如Aliostad建议的那样，您应首先按TspID分组，然后执行Ordered的总和，然后加入tsp。

答案 3 :(得分：1)

无需加入，只需进行选择：

create table #tsp(id int, name varchar(20),claimed int);
create table #sp(id int, name varchar(20),fiTsp int,ordered int);

insert into #tsp values(1,'1235-6044',300);
insert into #tsp values(2,'1234-5678',400);

insert into #sp values(1,'1235-6044',1,30);
insert into #sp values(2,'1235-6044',1,40);
insert into #sp values(3,'1235-6044',1,50);
insert into #sp values(4,'1234-5678',2,60);

WITH cte AS(
    select tsp.id As TspID,tsp.name as TspName,tsp.claimed As Claimed
    ,sp.id As SpID,sp.name As SpName,sp.ordered As Ordered
    from #sp sp inner join #tsp tsp
    on sp.fiTsp=tsp.id
)

SELECT id, name, SUM(claimed) as Claimed, (SELECT SUM(ordered) FROM #sp WHERE #sp.fiTsp = #tsp.id GROUP BY #sp.fiTsp) AS Ordered
FROM #tsp
GROUP BY id, name

drop table #tsp;
drop table #sp;

产地：

id  name        Claimed Ordered 
1   1235-6044   300 120 
2   1234-5678   400 60

- 编辑 -

根据附加信息，我可以尝试按照示例分割CTE以形成数据。我完全承认，Aliostad的方法可能会产生更清晰的查询，但这是使用子选择的尝试（完全失明）：

CREATE FUNCTION [Gambio].[rptReusedStatistics](
     @fromDate datetime
    ,@toDate datetime
    ,@fromInvoiceDate datetime
    ,@toInvoiceDate datetime
    ,@idClaimStatus varchar(50)
    ,@idSparePartCategories varchar(1000)
    ,@idSpareParts varchar(1000)
)

RETURNS TABLE AS
RETURN(
    WITH ExclusionCat AS (
        SELECT idSparePartCategory AS ID From tabSparePartCategory
        WHERE idSparePartCategory IN(- 3, - 1, 6, 172,168)
    ), ReportSP AS (
        SELECT fiTabSparePart
        ,Inventory
        ,Gambio.getGoodsIn(idSparePart,@FromDate,@ToDate) GoodsIn
        ,Gambio.getOrdered(idSparePart,@FromDate,@ToDate) Ordered
        FROM  Gambio.SparePart
    ), ReportTSP AS (
        SELECT TSP.idSparePart
        ,Cat.SparePartCategoryName AS Category
        ,TSP.SparePartDescription AS Part
        ,TSP.SparePartName AS PartNumber
        ,CASE WHEN TSP.idSparePart IS NULL THEN 0 ELSE
            Gambio.getClaimed(TSP.idSparePart,@FromInvoiceDate,@ToInvoiceDate,@idClaimStatus,1)END AS ClaimedReused
        ,CASE WHEN TSP.idSparePart IS NULL THEN 0 ELSE
            Gambio.getCostSaving(TSP.idSparePart,@FromInvoiceDate,@ToInvoiceDate,@idClaimStatus)END AS Costsaving
        FROM  tabSparePart AS TSP 
        INNER JOIN tabSparePartCategory AS Cat 
        ON Cat.idSparePartCategory=TSP.fiSparePartCategory
        WHERE Cat.idSparePartCategory NOT IN(SELECT ID FROM ExclusionCat)
        AND (@idSparePartCategories IS NULL 
            OR TSP.fiSparePartCategory IN(
                SELECT Item From dbo.Split(@idSparePartCategories,',')
            )
        )
        AND (@idSpareParts IS NULL 
            OR TSP.idSparePart IN(
                SELECT Item From dbo.Split(@idSpareParts,',')
            )
        )
    )       
    SELECT Category
    --, Part
    --, PartNumber
    , (SELECT SUM(Inventory) FROM ReportSP WHERE ReportSP.fiTabSparePart = idSparePart GROUP BY fiTabSparePart) AS Inventory
    , (SELECT SUM(GoodsIn) FROM ReportSP WHERE ReportSP.fiTabSparePart = idSparePart GROUP BY fiTabSparePart) AS GoodsIn
    , (SELECT SUM(Ordered) FROM ReportSP WHERE ReportSP.fiTabSparePart = idSparePart GROUP BY fiTabSparePart) AS Ordered
    , Claimed
    , ClaimedReused
    , Costsaving
    , Count(*) AS PartCount
    FROM ReportTSP
    GROUP BY Category
)

如果没有更好地理解整个架构，很难涵盖所有可能性，但这是否有效（我怀疑PartCount对于所有实例都是1）希望它会给你一些关于替代方法的新思路。 / p>

答案 4 :(得分：0)

SELECT
tsp.name
,max(tsp.claimed) as claimed
,sum(sp.ordered) as ordered 
from #sp sp 
inner join #tsp tsp 
on sp.fiTsp=tsp.id
GROUP BY tsp.name
相关问题
最新问题