SQL查找n行的平均值，其中n是字段的总和

Question

我有看起来像这样的数据。

SoldToRetailer

OrderDate | Customer | Product | price | units 
------------------------------------------------- 
1-jun-2011 | customer1 | Product1 | $10 | 5 
2-jun-2011 | customer1 | Product1 | $5 | 3 
3-jun-2011 | customer1 | Product2 | $10 | 4 
4-jun-2011 | customer1 | Product1 | $4 | 4 
5-jun-2011 | customer2 | Product3 | $10 | 1 

SalesByRetailers

Customer | Product | units 
----------------------------- 
customer1 | Product1 | 5 
customer2 | Product3 | 1 

这是我所需要的。

销售（平均价格）

Customer | Product | units | Average Price 
-------------------------------------------- 
customer1 | Product1 | 5  | $3.44 
customer2 | Product3 | 1  | $10 

平均价格被定义为平均价格最近SoldToRetailer价格加起来的单位。

因此，在第一种情况下，我从6月4日和6月2日获取订单。我不需要（实际上需要）从6月1日起的订单被包括在内。

编辑：希望更好的解释。

我试图确定一个项目被卖给零售商的正确（最近的）价格。这是价格的LIFO订单。价格是通过平均最近n次订单的价格来确定的。其中n =特定产品和客户的零售总额。

在SQL伪代码看起来像这样。

Select s1.Customer, s1.product, average(s2.price) 
from SalesByRetailers s1 
join SoldToRetailer s2 
on s1.customer=s2.customer 
and s1.product=s2.product 
and (select top (count of records where s2.units = s1.units) from s2 order by OrderDate desc) 

我需要返回的是来自SoldToRetailer的记录数量，其中单位总和> = SalesByRetailer单位。

它看起来可以通过RANK或rowover分区来解决，但我很茫然。

SoldToRetailer表格是巨大的，所以性能是非常宝贵的。

运行在SQL 2008R2 感谢您的帮助

Answer 1

所以我用3种技术。首先，我创建了一个带有over by子句的表格，给我一个产品和价格的排序列表，然后编辑表格以添加运行平均值。一个OUTER APPLY子选择修复了我最后的问题。希望代码能够帮助有类似问题的其他人。

向Jeff Moden大声喊出SQLSderverCentral.com名声，以表示正在运行的平均帮助。

 SELECT d.ProductKey, 
     d.ActualDollars, 
     d.Units, 
     ROW_NUMBER() OVER(PARTITION BY ProductKey ORDER BY d.OrderDateKey DESC) AS RowNumber, 
     NULL                 AS RunningTotal, 
     CONVERT(DECIMAL(10, 4), 0)           AS RunningDollarsSum, 
     CONVERT(DECIMAL(10, 4), 0)           AS RunningAverage 
    INTO #CustomerOrdersDetails 
    FROM dbo.orders d 
    WHERE customerKey = @CustomerToSelect 

    --DB EDIT... Google "Quirky update SQL Server central. Jeff Moden's version of a 
    --Running total. Holy crap it's faster. tried trangular joins before. 
    CREATE CLUSTERED INDEX [Index1] 
    ON #CustomerOrdersDetails (ProductKey ASC, RowNumber ASC) 

    DECLARE @RunningTotal INT 
    DECLARE @PrevProductKey INT 
    DECLARE @RunningDollarsSum DECIMAL(10, 4) 

    UPDATE #CustomerOrdersDetails 
    SET @RunningTotal = RunningTotal = CASE 
              WHEN ProductKey = @PrevProductKey THEN c.Units + ISNULL(@RunningTotal, 0) 
              ELSE c.Units 
             END, 
     @RunningDollarsSum = RunningDollarsSum = CASE 
                WHEN ProductKey = @PrevProductKey THEN c.ActualDollars + ISNULL(@RunningDollarsSum, 0) 
                ELSE c.ActualDollars 
                END, 
     @PrevProductKey = ProductKey, 
     RunningAverage = @RunningDollarsSum/NULLIF(@RunningTotal, 0) 
    FROM #CustomerOrdersDetails c WITH (TABLOCKX) 
    OPTION (MAXDOP 1) 

    -- ============================================= 
    -- Update Cost fields with average price calculation 
    -- ============================================= 
    UPDATE d 
    SET DolSoldCostUSD = COALESCE(d.DolSoldCostUSD, 
            d.UnitsSold * a.RunningAverage), 
    FROM dbo.inbound d 
     OUTER APPLY (SELECT TOP 1 * 
         FROM #CustomerOrdersDetails ap 
         WHERE ap.ProductKey = d.ProductKey 
          AND d.UnitsSold + d.UnitsOnHand + d.UnitsOnOrder + d.UnitsReceived + d.UnitsReturned >= RunningTotal 
         ORDER BY RunningTotal) AS a 

Answer 2

declare @table table (customer varchar(15), product varchar(15), qty int, price decimal(6,2)) 

insert into @table (customer, product, qty, price) 
values 
('customer1', 'product1', 5, 3), 
('customer1', 'product1', 4, 4), 
('customer1', 'product1', 3, 2), 
('customer1', 'product1', 2, 13), 
('customer1', 'product1', 3, 3), 

('customer1', 'product2', 5, 1), 
('customer1', 'product2', 4, 7), 
('customer1', 'product2', 2, 5), 
('customer1', 'product2', 6, 23), 
('customer1', 'product2', 2, 1), 

('customer2', 'product1', 2, 1), 
('customer2', 'product1', 4, 4), 
('customer2', 'product1', 7, 3), 
('customer2', 'product1', 1, 12), 
('customer2', 'product1', 2, 3), 

('customer2', 'product2', 3, 2), 
('customer2', 'product2', 6, 5), 
('customer2', 'product2', 8, 4), 
('customer2', 'product2', 2, 11), 
('customer2', 'product2', 1, 2) 

select customer, product, sum(qty) as units, (sum(qty * price))/SUM(qty) as 'Average Price' from @table 
group by customer, product