Question

我已根据建议修改了代码，但现在性能已经下降。欢迎任何建议

select *
    from (
    SELECT 
         d.ID, 
           d.HeaderId,
        CASE WHEN h.MyType = 'C' THEN
            RANK() OVER (PARTITION BY  l.Work ORDER BY l.Address1 DESC) 
        ELSE
            RANK() OVER (PARTITION BY  l.Home ORDER BY l.Address1 DESC)
        END 
        AS 'RANK',


        CASE WHEN h.MyType = 'C' THEN
            COUNT(l.Work) OVER (PARTITION BY  l.Work)
        ELSE
            COUNT(l.Home) OVER (PARTITION BY  l.Home)
        END  
        AS 'MAXCOUNT'

    FROM schema1.Details AS d 
        JOIN schema1.BatchHeader AS h 
          ON d.HeaderId = h.ID
        JOIN schema2.Details AS l
          ON d.LeadBatchDetailId = l.Id
        LEFT JOIN LDCs AS ldcElec 
          ON l.LDC_Elec = schema3.Code
        LEFT JOIN LDCs AS ldcGas 
          ON l.ldc_gas = ldcGas.Code
        LEFT JOIN schema2.Accounts ag
          ON (l.Work = ag.Phone AND 'G' = ag.Business AND h.MyType = 'C')
          OR (l.Home = ag.Phone AND 'G' = ag.Business AND h.MyType = 'R')
        LEFT JOIN schema2.Accounts ae 
          ON (l.Work = ae.Phone AND 'E' = ae.Business AND h.MyType = 'C')
          OR (l.Home = ae.Phone AND 'E' = ae.Business AND h.MyType = 'R')
        WHERE d.HeaderId = @Id) 
    ) a     
WHERE [RANK] = [MAXCOUNT] 
ORDER BY LdcGasName, LdcElecName

Answer 1

我不相信SQL会这样做，支持在排名函数中嵌入的AND型逻辑。您可能需要将其分解为更多子查询。一些极端的伪代码：

为“C”
全外连接选择“R”
使用子查询，外部查询执行coalesce和所有外连接

但是你必须处理COUNT ... OVER ...，而且这个数量不依赖于h.MyType = C或R，这让我想知道MAXCOUNT是否会最终使得排名值等于是如此受限制。

这一切都假设我正在读它......但这是一个复杂的查询，如果不了解底层表和业务逻辑就很难理解。

使用Partition Plus COUNT提高SELECT CASE的性能

1 个答案: