sql - Microsoft SQL Server 查询改进

Question

是否有机会提高以下查询的速度:

select distinct
    a.InvoiceAcc,
    (select top 1 b.CustomerName 
     from dbo.tblsales as b 
     where b.InvoiceAcc = a.InvoiceAcc),
    (select sum(b.SalesValue) 
     from dbo.tblsales as b 
     where b.InvoiceAcc = a.InvoiceAcc and b.Month = '2016-1') as [01],
    (select sum(b.SalesValue) 
     from dbo.tblsales as b 
     where b.InvoiceAcc = a.InvoiceAcc and b.Month = '2016-2') as [02],
    (select sum(b.SalesValue) 
     from dbo.tblsales as b 
     where b.InvoiceAcc = a.InvoiceAcc and b.Month = '2016-3') as [03],
    (select sum(b.SalesValue) 
     from dbo.tblsales as b 
     where b.InvoiceAcc = a.InvoiceAcc and b.Month = '2016-4') as [04],
    (select sum(b.SalesValue) 
     from dbo.tblsales as b 
     where b.InvoiceAcc = a.InvoiceAcc and b.Month = '2016-5') as [05],
    (select sum(b.SalesValue) from dbo.tblsales as b where b.InvoiceAcc = a.InvoiceAcc and b.Month = '2016-6') as [06],
    (select sum(b.SalesValue) from dbo.tblsales as b where b.InvoiceAcc = a.InvoiceAcc and b.Month = '2016-7') as [07],
    (select sum(b.SalesValue) from dbo.tblsales as b where b.InvoiceAcc = a.InvoiceAcc and b.Month = '2016-8') as [08],
    (select sum(b.SalesValue) from dbo.tblsales as b where b.InvoiceAcc = a.InvoiceAcc and b.Month = '2016-9') as [09],
    (select sum(b.SalesValue) from dbo.tblsales as b where b.InvoiceAcc = a.InvoiceAcc and b.Month = '2016-10') as [10],
    (select sum(b.SalesValue) from dbo.tblsales as b where b.InvoiceAcc = a.InvoiceAcc and b.Month = '2016-11') as [11],
    (select sum(b.SalesValue) from dbo.tblsales as b where b.InvoiceAcc = a.InvoiceAcc and b.Month = '2016-12') as [12]
from
    dbo.tblsales as a 
group by 
    a.InvoiceAcc, CustomerName

参数月份发送为 2016 且 InvoiceAcc(非必需)时，SQL Server 中的 View 会是什么样？

非常感谢。

Answer 1

根据您的示例，您想要这个

select 
  InvoiceAcc,
  CustomerName,
  sum(CASE WHEN [Month] = '2016-1' then salesvalue else 0 end) as [01],
  sum(CASE WHEN [Month] = '2016-2' then salesvalue else 0 end) as [02],
  sum(CASE WHEN [Month] = '2016-3' then salesvalue else 0 end) as [03],
  sum(CASE WHEN [Month] = '2016-4' then salesvalue else 0 end) as [04],
  sum(CASE WHEN [Month] = '2016-5' then salesvalue else 0 end) as [05],
  sum(CASE WHEN [Month] = '2016-6' then salesvalue else 0 end) as [06],
  sum(CASE WHEN [Month] = '2016-7' then salesvalue else 0 end) as [07],
  sum(CASE WHEN [Month] = '2016-8' then salesvalue else 0 end) as [08],
  sum(CASE WHEN [Month] = '2016-9' then salesvalue else 0 end) as [09],
  sum(CASE WHEN [Month] = '2016-10' then salesvalue else 0 end) as [10],
  sum(CASE WHEN [Month] = '2016-11' then salesvalue else 0 end) as [11],
  sum(CASE WHEN [Month] = '2016-12' then salesvalue else 0 end) as [12]
from dbo.tblsales 
group by InvoiceAcc, CustomerName

我确实发现测试(例如Month = '2016-1')很奇怪——我希望有一个日期字段和这样的测试MONTH(invoiceDate) = 1

我还认为，在 sql 中使用 unique 通常表明查询效果不佳 - 来自设计良好的模型的正确设计的查询很少需要 unique。

---

如果您想删除其他名称，最好的方法是:

select 
  InvoiceAcc,
  CustomerName,
  sum(CASE WHEN [Month] = '2016-1' then salesvalue else 0 end) as [01],
  sum(CASE WHEN [Month] = '2016-2' then salesvalue else 0 end) as [02],
  sum(CASE WHEN [Month] = '2016-3' then salesvalue else 0 end) as [03],
  sum(CASE WHEN [Month] = '2016-4' then salesvalue else 0 end) as [04],
  sum(CASE WHEN [Month] = '2016-5' then salesvalue else 0 end) as [05],
  sum(CASE WHEN [Month] = '2016-6' then salesvalue else 0 end) as [06],
  sum(CASE WHEN [Month] = '2016-7' then salesvalue else 0 end) as [07],
  sum(CASE WHEN [Month] = '2016-8' then salesvalue else 0 end) as [08],
  sum(CASE WHEN [Month] = '2016-9' then salesvalue else 0 end) as [09],
  sum(CASE WHEN [Month] = '2016-10' then salesvalue else 0 end) as [10],
  sum(CASE WHEN [Month] = '2016-11' then salesvalue else 0 end) as [11],
  sum(CASE WHEN [Month] = '2016-12' then salesvalue else 0 end) as [12]
from (
  SELECT InvoiceAcc, CustomerName, Month, salesvalue,
         ROW_NUMBER() OVER (PARTITION BY InvoiceAcc ORDER BY CustomerName) AS RN
         FROM tblsales
) x
WHERE RN=1
group by InvoiceAcc, CustomerName