提高 Table 值函数性能

Question

你能帮我改进一下table值函数吗:

    Create Function [dbo].[xxx](@InvoiceId numeric(18 ,0) ,@InvoiceType int ,@ReceiptAccOwner int)
    Returns Table
    As
    Return(


    Select * From(

            Select N'InvoiceReceiptNo' + Cast(ROW_NUMBER() over(order by ir.ReceiptRecID) as nvarchar(10)) As row

                ,ir.StockHoldingInvoiceID As InvoiceId

                ,ir.ReceiptNo As ReceiptValue
                ,ir.ReceiptAccOwner
            from InvoiceReceipt As ir       
            Where  ir.ReceiptAccOwner = @ReceiptAccOwner
            And (
                   (ir.StockHoldingInvoiceID = @InvoiceId And @InvoiceType in(1 ,6 ,8) )
                 or(ir.THCInvoiceID = @InvoiceId And @InvoiceType = 2)
                 or(ir.EvaluatInvoiceID = @InvoiceId And @InvoiceType = 3)
                 or(ir.StripInvoiceID = @InvoiceId And @InvoiceType = 4)
                 or(ir.EvaluationExportInvoiceID = @InvoiceId And @InvoiceType = 5)
                 or(ir.ServiceInvoiceID = @InvoiceId And @InvoiceType = 7)
                 or(ir.MovementInvoiceId = @InvoiceId And @InvoiceType = 10)
                 or(ir.InterchangeReplicaInvoice = @InvoiceId And @InvoiceType = 11)
                 or(ir.MovementInvoiceId = @InvoiceId And @InvoiceType = 12)
                 or(ir.BLConsigneeChangeRequestInvoiceID = @InvoiceId And @InvoiceType = 15)
                 )

            UNION All

            Select N'InvoiceReceiptPrice' + Cast(ROW_NUMBER() over(order by ir.ReceiptRecID) as nvarchar(10)) As row

                ,ir.StockHoldingInvoiceID

                ,Cast(IsNull(ir.ReceiptPrice ,0) as nvarchar(20)) As ReceiptValue
                ,ir.ReceiptAccOwner
            from InvoiceReceipt As ir
            Where  ir.ReceiptAccOwner = @ReceiptAccOwner
            And (
                   (ir.StockHoldingInvoiceID = @InvoiceId And @InvoiceType in(1 ,6 ,8) )
                 or(ir.THCInvoiceID = @InvoiceId And @InvoiceType = 2)
                 or(ir.EvaluatInvoiceID = @InvoiceId And @InvoiceType = 3)
                 or(ir.StripInvoiceID = @InvoiceId And @InvoiceType = 4)
                 or(ir.EvaluationExportInvoiceID = @InvoiceId And @InvoiceType = 5)
                 or(ir.ServiceInvoiceID = @InvoiceId And @InvoiceType = 7)
                 or(ir.MovementInvoiceId = @InvoiceId And @InvoiceType = 10)
                 or(ir.InterchangeReplicaInvoice = @InvoiceId And @InvoiceType = 11)
                 or(ir.MovementInvoiceId = @InvoiceId And @InvoiceType = 12)
                 or(ir.BLConsigneeChangeRequestInvoiceID = @InvoiceId And @InvoiceType = 15)
                 )

    ) As tb
    pivot(
             Max(ReceiptValue) --,Sum(ReceiptPrice)
             For row in ([InvoiceReceiptNo1] ,[InvoiceReceiptNo2] ,[InvoiceReceiptNo3] ,[InvoiceReceiptNo4] ,[InvoiceReceiptNo5] ,[InvoiceReceiptNo6] 
                        ,[InvoiceReceiptPrice1] ,[InvoiceReceiptPrice2] ,[InvoiceReceiptPrice3] ,[InvoiceReceiptPrice4] ,[InvoiceReceiptPrice5] ,[InvoiceReceiptPrice6])
         )   piv


    )

我在存储过程中使用此函数，在 select 语句中使用外部应用，此功能应该 return 每张发票只有一条记录。

XML 执行计划： Select * 来自 [dbo].[fn_GetInvoiceReceiptInfo](149735 , 1 ,1) https://anotepad.com/notes/wh34ey

Answer 1

答案取决于您拥有的索引。除非您特别详细说明您拥有哪些索引以及使用了哪些索引，否则很难想出更好的解决方案。

例如，您有一个聚合 window 函数，可以通过更改聚类索引或为您的特定查询添加覆盖索引来使其更快。但是，只有在这确实是瓶颈时才有意义。

执行计划是怎么说的？

您可能想尝试的另一件事是在 WHERE 子句中重写所有不带 OR 的查询，并对所有查询使用 UNION ALL。这将允许数据库并行化您的查询。

同样，您可以尝试这些方法，但它们是否有效取决于索引和查询计划瓶颈。

Answer 2

好的，查看您提供的执行计划，您正在执行 Key Lookups 这意味着您可以调整索引以减少此处的执行时间。

目前，您在 table InvoiceReceipt 到 return 上使用索引 Relationship_82_FK 这两个值 ReceiptRecID 和 StockHoldingInvoiceID.现在，由于此索引不包含您需要的所有信息 SQL 必须返回并对照您的聚簇索引查找 return 附加信息。尝试更改索引 Relationship_82_FK 以将 ReceiptAccHolder 和 ReceiptPrice 作为包含字段。

提高 Table 值函数性能

Improve Table Value Function Performance

tsql

sql-server

query-performance