Temp Table 是最佳解决方案吗？

Question

我正在构建一个存储过程，它从几个不同的 table 中提取大量不同的数据，我需要一种方法从销售 table 中提取销售信息，然后执行各种对该销售数据的总结。在下面的示例中，我使用临时 table 完成了此操作，但有人建议可能有更好的方法。有没有更有效的方法来完成我在这里所做的事情？

 SELECT * INTO #TempSales FROM [Sales] WHERE ClientId = @ClientId
    SELECT 
        [Customer].[CustomerID],
        [Customer].[AccountBalAmountOpen],
        [Customer].[AccountAgeAmountDays0],
        [Customer].[AccountAgeAmountDays30],
        [Customer].[AccountAgeAmountDays60],
        [Customer].[AccountAgeAmountDays90],
        [Customer].[AccountAgeAmountDaysOver90],
        (SELECT SUM(SalesAmount) FROM #TempSales WHERE CustomerId = [Customer].[CustomerID] AND Origin = 385) AS ServiceLifeTimeSales,
        (SELECT SUM(SalesAmount) FROM #TempSales WHERE CustomerId = [Customer].[CustomerID] AND Origin = 385 AND MONTH(SaleDate) = MONTH(GETDATE()) AND YEAR(SaleDate) = YEAR(GETDATE())) AS ServiceMonthToDateSales,
        (SELECT SUM(SalesAmount) FROM #TempSales WHERE CustomerId = [Customer].[CustomerID] AND Origin = 385 AND YEAR(SaleDate) = YEAR(GETDATE())) AS ServiceYearToDateSales,
        (SELECT SUM(SalesAmount) FROM #TempSales WHERE CustomerId = [Customer].[CustomerID] AND Origin = 385 AND YEAR(SaleDate) = (YEAR(GETDATE()) - 1)) AS ServicePreviousYearSales,
    (SELECT SUM(SalesAmount) FROM #TempSales WHERE CustomerId = [Customer].[CustomerID] AND Origin = 460) AS PartsLifeTimeSales,
        (SELECT SUM(SalesAmount) FROM #TempSales WHERE CustomerId = [Customer].[CustomerID] AND Origin = 460 AND MONTH(SaleDate) = MONTH(GETDATE()) AND YEAR(SaleDate) = YEAR(GETDATE())) AS PartsMonthToDateSales,
        (SELECT SUM(SalesAmount) FROM #TempSales WHERE CustomerId = [Customer].[CustomerID] AND Origin = 460 AND YEAR(SaleDate) = YEAR(GETDATE())) AS PartsYearToDateSales,
        (SELECT SUM(SalesAmount) FROM #TempSales WHERE CustomerId = [Customer].[CustomerID] AND Origin = 460 AND YEAR(SaleDate) = (YEAR(GETDATE()) - 1)) AS PartsPreviousYearSales,
        [Orders].[CustomerId] AS ParentCustomerId,
        [Orders].[OrderId],
        [Orders].[OrderStatus],
        [Orders].[UnitId],
        [Orders].[FleetId],
        [Orders].[CreatedDate] AS OrderCreatedDate,
        [Orders].[OrderType],
        [OrderParts].[OrderId] AS ParentOrderId,
        [OrderParts].[PartId],
        [OrderParts].[PartDescription],
        [OrderParts].[QuantityShip],
        [OrderParts].[QuantityBackOrder],
        [OrderParts].[CreatedDate] AS PartCreatedDate
      FROM [Customer]
      LEFT JOIN [Orders]
      ON [Orders].[CustomerId] = [Customer].[CustomerID]
      LEFT JOIN [OrderParts]
      ON [OrderParts].[OrderId] = [Orders].[OrderId]
      WHERE [Customer].[ClientID] = @ClientId
DROP TABLE #TempSales

Answer 1

虽然临时 tables 可用于存储将在其他查询中重用的复杂查询的中间结果，但我认为您在创建临时 table 如下：

SELECT * INTO #TempSales FROM [Sales] WHERE ClientId = @ClientId

您没有使用任何东西加入 table，也没有进行计算或使用聚合函数，因此如果 table 被正确索引（至少 ClientId、CustomerId ，Origin 和 SalesDate)，您的查询应该执行良好甚至更快，因为使用 SELECT INTO 创建的 tables 没有任何索引，除非您手动添加它们。

但是您可能正在使用临时 table 来提高速度，因为您遇到这样的情况：

AND YEAR(SaleDate) = YEAR(GETDATE())
AND MONTH(SaleDate) = MONTH(GETDATE()) AND YEAR(SaleDate) = YEAR(GETDATE())

which make the query non-sargable 因为您将筛选列包含在函数中。由于这个原因，查询优化器无法正确使用 SaleDate 上的任何现有索引，并且由于时间 table 的行数较少，因此与 [=21] 相比，对其执行完整扫描所需的时间更少=] table.

您可以使用不对筛选列应用任何函数的条件来修复此问题（尽管您可以对常量值使用函数）：

declare @thisYearStart as datetime, @nextYearStart as datetime,
  @thisMonthStart as datetime, @nextMonthStart as datetime

set @thisYearStart=DATEFROMPARTS(YEAR(GETDATE()),1,1)
set @nextYearStart=DATEADD(year,1,@thisYearStart)
set @thisMonthStart=DATEFROMPARTS(YEAR(GETDATE()),MONTH(GETDATE()),1)
set @nextMonthStart=DATEADD(month,1,@thisMonthStart)

或者如果您使用的 SQL 版本早于 SQL Server 2012：

set @thisYearStart=CAST(CAST(YEAR(GETDATE()) as char(4))+'0101' AS datetime) -- ISO format
set @nextYearStart=DATEADD(year,1,@thisYearStart)
set @thisMonthStart=DATEADD(month,MONTH(GETDATE())-1,@thisYearStart)
set @nextMonthStart=DATEADD(month,1,@thisMonthStart)

然后只需使用：

AND SaleDate>=@thisYearStart AND SaleDate<@nextYearStart
AND SaleDate>=@thisMonthStart AND SaleDate<@nextMonthStart

请注意，您不需要使用变量，您可以直接在 WHERE 条件下使用 DATEFROMPARTS(...)，但由于您已经使用它们创建了一个存储过程，因此它们将使查询更具可读性。

Temp Table 是最佳解决方案吗？

Is a Temp Table the best solution?

sql

sql-server

stored-procedures

temp-tables