Question

我正在努力解决查询性能问题。我将“平面”XML存储为表格中的XML。我正在尝试将该XML的所有元素和属性提取到某个平面表中。问题是我要解析大约200列/属性。这就是它的样子：

SELECT
      x.i.value('./@Id', 'nvarchar(max)') as [Id],
      x.i.value('(./Elem1/text())[1]', 'nvarchar(max)') as [Elem1],
      x.i.value('(./Elem2/text())[1]', 'nvarchar(max)') as [Elem2],
      x.i.value('./@Attr1', 'nvarchar(max)') as [Attr1],
      x.i.value('./@Attr2', 'nvarchar(max)') as [Attr2],
      x.i.value('./@Attr3', 'nvarchar(max)') as [Attr3],
      x.i.value('./@Attr4', 'nvarchar(max)') as [Attr4],
      x.i.value('./@Attr5', 'nvarchar(max)') as [Attr5],
      x.i.value('./@Attr6', 'nvarchar(max)') as [Attr6],
      x.i.value('./@Attr7', 'nvarchar(max)') as [Attr7],
      x.i.value('(./Elem3/text())[1]', 'nvarchar(max)') as [Elem3],
      x.i.value('(./Elem4/text())[1]', 'nvarchar(max)') as [Elem4],
      x.i.value('(./Elem5/text())[1]', 'nvarchar(max)') as [Elem5],
      x.i.value('(./Elem6/text())[1]', 'nvarchar(max)') as [Elem6],
      ...
      x.i.value('(./Elem200/text())[1]', 'nvarchar(max)') as [Elem200]
   FROM [XmlLoad].[DataTable] as t
   CROSS APPLY t.[Xml].nodes('./Log') as x(i);

XML列的示例内容是：

<Log Id="1" Attr1="a" Attr2="a" Attr3="a" Attr4="a" Attr5="a" Attr6="a" Attr7="a" >
    <Elem1>value</Elem1>
    <Elem2>value</Elem2>
    <Elem3>value</Elem3>
    <Elem4>value</Elem4>
    <Elem5>value</Elem5>
    <Elem6>value</Elem6>
    ...
    <Elem200>value</Elem200>
</Log>

SQL服务器做的是为每个元素创建一个嵌套循环，这给出了疯狂的执行计划。当我提取10个元素时，这不是问题，但对于200，这非常慢。

我尝试过主要和次要的XML索引，性能变得更糟，我也尝试将解析更改为：

x.i.value('(./Log/Elem1/text())[1]', 'nvarchar(max)') as [Elem1],

没有CROSS APPLY但执行计划是相同的。执行计划是...... part of execution plan

完整计划看起来像这样:) full execution plan

有什么方法可以改善吗？我应该以某种方式重建查询吗？

SQL Server XQuery性能（嵌套循环）

0 个答案: