优化 TSQL 除了

Question

如果我可以包含表 MasterStaging（临时表）中的 ID，我可以优化连接：

• 在EXCEPT 中包含ID 会扭曲结果，因为MasterStaging 中的ID 总是与StatusComparison 不同
• MasterStaging.ID 无关紧要，只是一个自动编号，不代表客户 ID
• SQL 旨在显示缺少的客户，无论 ID 是什么。
• CustomerAccountNo 不是唯一编号

如何在JOIN包含获取ID top优化？

这就是我想要的：

与下面的 SQL 相同，但我将 JOIN 仅使用 ID

) x ON e.ID = x.ID  

这是我目前所拥有的：

UPDATE ecl.MasterStaging 
SET NewAccount = 1
    FROM ecl.MasterStaging e WITH (NOLOCK)
    JOIN (
          SELECT  
              ISNULL(Usable, 0) AS Usable ,
              ISNULL(TypeRC, 0) AS TypeRC ,
              ISNULL(CustomerNumber, 0) AS CustomerNumber ,
              ISNULL(CustomerAccountNo, 0) AS CustomerAccountNo ,
              ISNULL(LoadProfileClass, 0) AS LoadProfileClass ,
              ISNULL(MeterNo, 0) AS MeterNo ,
              ISNULL(PrimaryPhoneNumber, 0) AS PrimaryPhoneNumber ,
              ISNULL(CustomerName1, 0) AS CustomerName1 ,
              ISNULL(ServiceAddress1, 0) AS ServiceAddress1 ,
              ISNULL(ServiceCity, 0) AS ServiceCity ,
              ISNULL(ServiceState, 0) AS ServiceState ,
              ISNULL(ServiceZip, 0) AS ServiceZip ,
              ISNULL(BillingAddress1, 0) AS BillingAddress1 ,
              ISNULL(BillingCity, 0) AS BillingCity ,
              ISNULL(BillingState, 0) AS BillingState ,
              ISNULL(substring(BillingZip, 1, 5), 0) as BillingZip ,
              ISNULL(substring(BillingZip4, 7, 4), 0) as BillingZip4
          FROM    
              ecl.MasterStaging WITH (NOLOCK)           

          EXCEPT

           SELECT  Usable ,
            TypeRC ,
            CustomerNumber ,
            CustomerAccountNo ,
            LoadProfileClass ,
            MeterNo ,
            PrimaryPhoneNumber ,
            CustomerName1 ,
            ServiceAddress1 ,                   
            ServiceCity ,
            ServiceState ,
            ServiceZip ,
            BillingAddress1 ,                   
            BillingCity ,
            BillingState ,
            BillingZip ,
            BillingZip4 
    FROM    ecl.StatusComparison  WITH (NOLOCK)
    WHERE   [Status] <> 'D'
            ) x
    ON 
            ISNULL(e.Usable,0) = x.Usable AND 
            ISNULL(e.TypeRC,0) = x.TypeRC AND 
            ISNULL(e.CustomerNumber,0) = x.CustomerNumber AND 
            ISNULL(e.CustomerAccountNo,0) = x.CustomerAccountNo AND
            ISNULL(e.LoadProfileClass,0) = x.LoadProfileClass AND
            ISNULL(e.MeterNo,0) = x.MeterNo AND
            ISNULL(e.PrimaryPhoneNumber,0) = x.PrimaryPhoneNumber AND
            ISNULL(e.CustomerName1,0) = x.CustomerName1 AND
            ISNULL(e.ServiceAddress1,0) = x.ServiceAddress1 AND 
            ISNULL(e.ServiceCity,0) = x.ServiceCity AND
            ISNULL(e.ServiceState,0) = x.ServiceState AND
            ISNULL(e.ServiceZip,0) = x.ServiceZip AND
            ISNULL(e.BillingAddress1,0) = x.BillingAddress1 AND
            ISNULL(e.BillingCity,0) = x.BillingCity AND
            ISNULL(e.BillingState,0) = x.BillingState AND
            ISNULL(e.BillingZip,0) = x.BillingZip AND
            ISNULL(e.BillingZip4,0) = x.BillingZip4

Answer 1

一个想法，但这可能并不理想：您可以在两个表上创建一个计算和持久的列，该列将计算所有列的校验和，并且您可以加入它。不过有碰撞的风险：

（我没有测试代码，如果有错误请告诉我）：

ALTER TABLE ecl.MasterStaging
ADD hash AS CHECKSUM(Usable ,
            TypeRC ,
            CustomerNumber ,
            CustomerAccountNo ,
            LoadProfileClass ,
            MeterNo ,
            PrimaryPhoneNumber ,
            CustomerName1 ,
            ServiceAddress1 ,                   
            ServiceCity ,
            ServiceState ,
            ServiceZip ,
            BillingAddress1 ,                   
            BillingCity ,
            BillingState ,
            BillingZip ,
            BillingZip4 ) PERSISTED

缺点：

• 更多存储空间
• 在插入和更新期间进行更多计算
• 您需要更改表的结构
• 如果您需要在 LOB 上进行比较，它将不起作用