Question

目前我们导入一个表Import_File，它有一个Options列，该列具有分隔值。我们需要将这些分隔值加载到不同的表中。

目前我们一次只做一行，因为行数可以是100k +

有没有办法加快下面的代码？

Declare @InvId uniqueidentifier
Declare @xml xml
Declare CurrFeatureList Cursor For
Select 
    import.InventoryId,
     N'<root><r><![CDATA[' + replace( import.OPTIONS ,',',']]></r><r><![CDATA[') + ']]></r></root>'
From                Import_File import with (nolock)
Where
    import.options IS NOT NULL
    And ISNULL(import.IsFeatureProcessed,0) = 0
    And LEN(ISNULL(import.OPTIONS,''))>10
    And import.InventoryId Is Not Null

OPEN CurrFeatureList

FETCH NEXT FROM CurrFeatureList
INTO @InvId, @xml

Print 'Inventory Import #10000'
Print GetDate()
WHILE @@FETCH_STATUS = 0
BEGIN
    BEGIN TRY

        Insert Into Import_File_Feature
        (
            FeatureId,
            InventoryId,
            FeatureText,
            FeatureGroup,
            FeatureCategory,
            FeatureIsAvailable,
            FeatureIsStandard
        )
        Select 
            NEWID(),
            @InvId,
            t.value('.','varchar(250)'),
            '',
            '',
            1,
            1
        From @xml.nodes('//root/r') as a(t)

        FETCH NEXT FROM CurrFeatureList
        INTO @InvId, @xml
    END TRY
    BEGIN CATCH
        Print 'Error    ' 
        Print @InvId 
        Print   ERROR_NUMBER() 
        Print   ERROR_SEVERITY()
        Print   ERROR_STATE()
        Print   ERROR_PROCEDURE()
        Print   ERROR_LINE()
        Print   ERROR_MESSAGE()

        FETCH NEXT FROM CurrFeatureList
        INTO @InvId, @xml
    END CATCH
END

Close CurrFeatureList
Deallocate CurrFeatureList
GO

Answer 1

我见过光标的唯一原因＆amp;用于此类事情的TRY / CATCH块用于在开发新的ETL过程时识别和分析不良记录。如果那不是你正在做的事情那么光标不是必需的，会让你慢下来。

让我们回顾一下你在做什么：

您将数据拖入游标（CurrFeatureList）并将Import_File.Options转换为XML字段，以便稍后使用XML NODES方法“拆分”您的字符串。
开出光标
对于每个InventoryId，您：一个。将Import_File.Options拆分为多行湾将InventoryId和关联的行插入Import_File_Feature C。如果出现错误则将其打印出来并转到下一条记录

你应该做什么

请注意我是如何使用XML节点拆分此字符串的：

DECLARE @x varchar(100) = 'abc,cde,fff';

SELECT item = xxx.value('(text())[1]', 'varchar(100)')
FROM (VALUES (CAST(('<r>'+REPLACE(@x,',','</r><r>') +'</r>') AS xml))) x(xx)
CROSS APPLY xx.nodes('r') xxx(xxx);

<强>结果

item
-----
abc
cde
fff

选项1

将初始连接和后续XML / XML节点拆分器逻辑组合到一个语句中并执行插入：

WITH 
yourData AS
(
  Select 
      import.InventoryId,
       x = N'<r><![CDATA[' + replace( import.OPTIONS ,',',']]></r><r><![CDATA[') + ']]></r>'
  From Import_File import with (nolock)
  Where
  import.options IS NOT NULL
  And ISNULL(import.IsFeatureProcessed,0) = 0
  And LEN(ISNULL(import.OPTIONS,''))>10
  And import.InventoryId Is Not Null
),
split AS
(
  SELECT InventoryId, item = i.value('.', 'varchar(8000)')
  FROM yourData
  CROSS APPLY x.nodes('r') s(i)  
)
Insert Into Import_File_Feature
(
  FeatureId,
  InventoryId,
  FeatureText,
  FeatureGroup,
  FeatureCategory,
  FeatureIsAvailable,
  FeatureIsStandard
 )
Select 
   newid(),
   import.InventoryId,
   item, -- this is the split out item from import.Options
    '',
    '',
    1,
    1
FROM split;

选项2

获取DelimitedSplit8K的副本并使用它来进行拆分。

WITH split AS
(
  Select 
      import.InventoryId,
      import.OPTIONS
  From Import_File import with (nolock)
  CROSS APPLY dbo.DelimitedSplit8K(import.OPTIONS, ',')
  Where
  import.options IS NOT NULL
  And ISNULL(import.IsFeatureProcessed,0) = 0
  And LEN(ISNULL(import.OPTIONS,''))>10
  And import.InventoryId Is Not Null
)
Insert Into Import_File_Feature
(
  FeatureId,
  InventoryId,
  FeatureText,
  FeatureGroup,
  FeatureCategory,
  FeatureIsAvailable,
  FeatureIsStandard
 )
Select 
   newid(),
   import.InventoryId,
   item,
    '',
    '',
    1,
    1
FROM split

请注意，因为我没有任何表定义或示例数据，所以我无法测试上面的代码。

MSSQL批量字符串到XML插入

1 个答案: