Question

在我的计算数据层中，我尝试在订单时填写客户的邮政编码，正在填充的表的子样本如下：

CustomerOrders
(
    CustomerID varchar(20),
    ...
    OrderDate date,
    ...
    CustomerPostcodeAtTimeOfOrder varchar(10)
)

此表是Customers表，Orders表和CustomerAddress表的连接，如下所示：

CustomerAddress
(
    CustomerID varchar(20),
    AddressType varchar(10),
    /*
    AddressDetails    
    */
    StartDate date,
    EndDate date,
    AddressRank int
)

可以想象，客户可能已经记录了单个日期的各种类型的地址，因此填充CustomerOrders表时的意图是加入如下：

SELECT *
FROM Customers c
LEFT JOIN Orders o
    ON o.CustomerID = c.CustomerID
OUTER APPLY
    (
        SELECT TOP 1 Postcode
        FROM CustomerAddress ca
        WHERE ca.CustomerID = c.CustomerID
        AND o.OrderDate BETWEEN ca.StartDate AND ca.EndDate
        ORDER BY AddressRank
    )

然而，通过将此连接添加到查询中我获得的性能损失意味着返回1000行从花费4秒到花费106秒。

请注意，我也在Address表上添加了非聚集索引。其定义如下：

CREATE NONCLUSTERED INDEX (IX_CustomerAddress)
ON CustomerAddress (StartDate, EndDate)
INCLUDE (AddressRank, CustomerID, Postcode)

我正在寻找有关解决此问题的最佳方法的任何建议吗？

Answer 1

我不完全确定这是否会更快地返回结果，但您可以像这样重写查询：

;WITH OrderAddress AS
(
    SELECT  o.*,
            ca.Postcode, 
            RN = ROW_NUMBER() OVER(PARTITION BY CustomerID ORDER BY AddressRank DESC)
    FROM CustomerAddress ca
    INNER JOIN Orders o
        ON ca.CustomerID = c.CustomerID
        AND o.OrderDate BETWEEN ca.StartDate AND ca.EndDate
)
SELECT *
FROM Customers c
LEFT JOIN ( SELECT *
            FROM OrderAddress
            WHERE RN = 1) o
    ON o.CustomerID = c.CustomerID;

您还应该在地址表上发布索引定义。

加入唯一ID和日期范围 - 必须返回1行

1 个答案: