为单个记录提供结果的条件聚合查询

Question

我有如下表格

订购

OrderID	UserID
1	7
2	1
3	13
4	7

出租物品

RentalItemsID	ItemID	ReturnTypeID	OrderID
1	9	5	1
2	3	4	2
3	4	5	2
4	5	5	2
5	10	4	3
6	11	4	4
7	14	5	4

ReturnTypeID 5：缺陷返回

ReturnTypeID 4：正常返回

select [order].UserID,
    (
        select avg(case when ReturnTypeID = 5 then 1.0 else 0 end) as DefectedRate
        from RentalItems 
        where RentalItems.OrderID = [Order].OrderID 
    ) as DefectedReturnRate
from [Order] 
where OrderID in (select OrderID from RentalItems where ReturnTypeID=5)

当前输出：计算为：DefetedReturnByUserPerOrder/AllReturnsByUserPerOrder

UserID	DefectReturnRate	Showing Calculation, (Not incl. in output)
7	1.00	Order1: 1/1 - Contained 1 item and it was defected returned
1	0.66	Order2: 2/3 - Contained 3 items and 2 were defected returned
7	0.50	Order4: 1/2 - Contained 2 items and 1 was defected returned

如您所见，它计算的是每个订单的 DefetecedReturnRate，而不是用户在所有订单中租用的物品总数

所需的输出： TotalDefetedReturnByUserForALLOrders/AllReturnsByUserForAllOrders

UserID	DefectReturnRate	Showing Calculation, (Not incl. in output)
7	0.66	Order1&4: 2/3 - Total Rented items by User 7 were 3 and 2 were defected Returns
1	0.66	Order2: 2/3 - Contained 3 items and 2 were defected returned

我认为这将是我的最后一个问题，因此我们将不胜感激任何帮助。我可以提供 DML+DDL 数据，但我不确定如何提供。我试图生成脚本，但它只生成表而不是数据。但表格和数据与我在上面的表格中输入的完全一样。

Answer 1

你可以这样做：

SELECT UserID, AVG(case when ReturnTypeID = 5 then 1.0 else 0 end) as DefectedRate
FROM orders 
JOIN dbo.RentalItems AS ri
   ON ri.OrderID = orders.OrderID
WHERE ri.OrderID in (select OrderID from RentalItems where ReturnTypeID = 5)
GROUP BY UserID

Answer 2

您也可以稍微不同地编写它，而无需两次点击 RentalItems。

这种方式产生 5 次逻辑读取而不是 12 次逻辑读取，因此在现实世界中性能会更高，如果这对您的情况很重要。

select distinct userid, Returned
from (
    select o.userid,  Sum (case when ReturnTypeId=5 then 1.0 end) over( partition by userid) / Count(*) over(partition by userid)  Returned
    from Orders o
    join RentalItems ri on ri.orderid=o.orderid
)s
where returned is not null