SQL Server 2008 中使用 XML.Modify() 进行查询的性能优化

Question

I am generating an XML using FOR XML in SQL Server 2008.

The XML output generated is of 5.18 MB and the process up to generation of XML takes just 5 seconds.

Now, I am adding attributes in this XML using XML.modify. After modification the size increases to 5.25 MB. But the file takes 17 minutes for this remaining processing (adding the attributes).

I am keeping the XML data in an XML type variable,there are around 6000 rows in this XML document and i am adding the attributes in approx half of the rows. These attributes are added under different node with some condition. The maximum time in 17 minutes approx(16 minutes) are consumed while adding these attributes.

I would request, if some one can tell me the best practices and ways to improving the performance.

XML Generated before adding the Styles

        <?xml version="1.0"?>
        <?mso-application progid="Excel.Sheet"?>
        <Workbook xmlns:x="urn:schemas-microsoft-com:office:excel" xmlns:ss="urn:schemas-microsoft-com:office:spreadsheet" xmlns="urn:schemas-microsoft-com:office:spreadsheet">
          <Worksheet xmlns:x="urn:schemas-microsoft-com:office:excel" xmlns:ss="urn:schemas-microsoft-com:office:spreadsheet" xmlns="urn:schemas-microsoft-com:office:spreadsheet" ss:Name="INDEX">
            <Table xmlns:x="urn:schemas-microsoft-com:office:excel" xmlns:ss="urn:schemas-microsoft-com:office:spreadsheet" xmlns="urn:schemas-microsoft-com:office:spreadsheet">
              <Column></Column>
              <Row xmlns:x="urn:schemas-microsoft-com:office:excel" xmlns:ss="urn:schemas-microsoft-com:office:spreadsheet" xmlns="urn:schemas-microsoft-com:office:spreadsheet">
                <Cell xmlns:x="urn:schemas-microsoft-com:office:excel" xmlns:ss="urn:schemas-microsoft-com:office:spreadsheet" xmlns="urn:schemas-microsoft-com:office:spreadsheet">
                  <Data ss:Type="String">#</Data>
                </Cell>
                <Cell xmlns:x="urn:schemas-microsoft-com:office:excel" xmlns:ss="urn:schemas-microsoft-com:office:spreadsheet" xmlns="urn:schemas-microsoft-com:office:spreadsheet">
                  <Data ss:Type="String">Study Name</Data>
                </Cell>

XML after adding attributes

        <?xml version="1.0"?>
        <?mso-application progid="Excel.Sheet"?>
        <Workbook xmlns="urn:schemas-microsoft-com:office:spreadsheet" xmlns:x="urn:schemas-microsoft-com:office:excel" xmlns:ss="urn:schemas-microsoft-com:office:spreadsheet">
          <Styles xmlns="urn:schemas-microsoft-com:office:spreadsheet" xmlns:x="urn:schemas-microsoft-com:office:excel" xmlns:ss="urn:schemas-microsoft-com:office:spreadsheet">

          <Worksheet xmlns="urn:schemas-microsoft-com:office:spreadsheet" xmlns:x="urn:schemas-microsoft-com:office:excel" xmlns:ss="urn:schemas-microsoft-com:office:spreadsheet" ss:Name="INDEX">
            <Table xmlns="urn:schemas-microsoft-com:office:spreadsheet" xmlns:x="urn:schemas-microsoft-com:office:excel" xmlns:ss="urn:schemas-microsoft-com:office:spreadsheet" ss:DefaultRowHeight="50">
              <Row xmlns="urn:schemas-microsoft-com:office:spreadsheet" xmlns:x="urn:schemas-microsoft-com:office:excel" xmlns:ss="urn:schemas-microsoft-com:office:spreadsheet" ss:StyleID="s66">
                <Cell xmlns="urn:schemas-microsoft-com:office:spreadsheet" xmlns:x="urn:schemas-microsoft-com:office:excel" xmlns:ss="urn:schemas-microsoft-com:office:spreadsheet" ss:MergeAcross="5">
                  <Data ss:Type="String">#</Data>
                </Cell>
                <Cell xmlns="urn:schemas-microsoft-com:office:spreadsheet" xmlns:x="urn:schemas-microsoft-com:office:excel" xmlns:ss="urn:schemas-microsoft-com:office:spreadsheet">
                  <Data ss:Type="String">Study Name</Data>
                </Cell>

Code to generate Basic XML

CREATE TABLE TempData (
    [ID] [INT] NOT NULL IDENTITY(1, 1) PRIMARY KEY
    ,[Col1] [NVARCHAR](MAX)
    ,[Col2] [NVARCHAR](MAX)
    ,[Col3] [NVARCHAR](MAX)
    ,[WORKSHEET_ID] INT 
    ,[FORM_OID] [NVARCHAR](MAX)
    ,[Col4] [NVARCHAR](MAX)
    ,[Col5] [NVARCHAR](MAX)
    ,[Col6] [NVARCHAR](MAX)
    ,[Col7] [NVARCHAR](MAX)
    ,[Col8] [NVARCHAR](MAX)
    ,[Col9] [NVARCHAR](MAX)
    ,[RANKING] INT
    );

SELECT  @xml = (SELECT (
          SELECT  
            CASE WHEN a2.[FORM_OID]='INDEX' THEN a2.[FORM_OID]
            ELSE  SUBSTRING(CONVERT(varchar(10),(a2.[WORKSHEET_ID]-1))+'_'+ a2.[FORM_OID],0,31) 
            END  '@ss:Name',(
                  SELECT '' as [Column],(
                           SELECT   
                               (SELECT 
                                      'String' as [Data/@ss:Type],
                                      [Col1] as [Data]
                                      FOR XML PATH('Cell'), TYPE),
                                      (SELECT 
                                     'String' as [Data/@ss:Type],
                                      [Col2] as [Data]  
                                      FOR XML PATH('Cell'), TYPE),
                                      (SELECT 
                                     'String' as [Data/@ss:Type],
                                      [Col3] as [Data]  
                                      FOR XML PATH('Cell'), TYPE),
                                      (SELECT 
                                     'String' as [Data/@ss:Type],
                                      [Col4] as [Data]  
                                      FOR XML PATH('Cell'), TYPE),
                                      (SELECT 
                                     'String' as [Data/@ss:Type],
                                      [Col5] as [Data]  
                                      FOR XML PATH('Cell'), TYPE),
                                      (SELECT 
                                     'String' as [Data/@ss:Type],
                                      [Col6] as [Data]  
                                      FOR XML PATH('Cell'), TYPE),
                                      (SELECT 
                                     'String' as [Data/@ss:Type],
                                      [Col7] as [Data]  
                                      FOR XML PATH('Cell'), TYPE),
                                      (SELECT 
                                     'String' as [Data/@ss:Type],
                                      [Col8] as [Data]  
                                      FOR XML PATH('Cell'), TYPE),
                                      (SELECT 
                                     'String' as [Data/@ss:Type],
                                      [Col9] as [Data]  
                                      FOR XML PATH('Cell'), TYPE),
                                      (SELECT 
                                     'String' as [Data/@ss:Type],
                                      [RANKING] as [Data]  
                                      FOR XML PATH('Cell'), TYPE)
                               FROM TempData  a1
                               WHERE a1.[WORKSHEET_ID]=a2.[WORKSHEET_ID]
                               GROUP BY [ID],[Col1],[Col2],[Col3],[FORM_OID],[Col4],[Col5],[Col6],[Col7],[Col8],[Col9],[RANKING] --form oid
                               order by [ID]
                         FOR XML PATH('Row'), TYPE
                         )
                  FOR XML PATH('Table'), TYPE
                         )
           FROM TempData  a2
           group by [WORKSHEET_ID],[FORM_OID]
    FOR XML PATH('Worksheet'), TYPE)          
    FOR XML PATH('Workbook'), TYPE )

Code to add Style (One of the many attributes i am adding)

SET @xml.modify('declare default element namespace  "urn:schemas-microsoft-com:office:spreadsheet";
                    declare namespace ss="urn:schemas-microsoft-com:office:spreadsheet" ;
                    declare namespace x="urn:schemas-microsoft-com:office:excel";
                    insert  attribute ss:MergeAcross{"5"}           
                    into (/Workbook/Worksheet[position()=sql:variable("@worksheets")][1]/Table/Row[1]/Cell[1])[1]')

Answer 1

This modified query will at least reduce the number of bytes drastically by avoiding so many repeated namespace declarations. If you cannot go with my suggestion (read below), this should speed up things anyway. 3000 RBAR modify calls should be faster with about 10% of the size...

I hope you can use this for your full query:

CREATE TABLE TempData (
    [ID] [INT] NOT NULL IDENTITY(1, 1) PRIMARY KEY
    ,[Col1] [NVARCHAR](MAX)
    ,[Col2] [NVARCHAR](MAX)
    ,[Col3] [NVARCHAR](MAX)
    ,[WORKSHEET_ID] INT 
    ,[FORM_OID] [NVARCHAR](MAX)
    ,[Col4] [NVARCHAR](MAX)
    ,[Col5] [NVARCHAR](MAX)
    ,[Col6] [NVARCHAR](MAX)
    ,[Col7] [NVARCHAR](MAX)
    ,[Col8] [NVARCHAR](MAX)
    ,[Col9] [NVARCHAR](MAX)
    ,[RANKING] INT
    );
INSERT INTO TempData(Col1,Col2,Col3,WORKSHEET_ID,FORM_OID,Col4,Col5,Col6,Col7,Col8,Col9,RANKING) 
VALUES('test1','test2','test3',1,'formOID','test4','test5','test6','test7','test8','test9',100);

DECLARE @xml XML;
WITH XMLNAMESPACES('urn:schemas-microsoft-com:office:spreadsheet' AS ss
                  ,'urn:schemas-microsoft-com:office:excel' AS x
                  ,DEFAULT 'urn:schemas-microsoft-com:office:spreadsheet')
SELECT  @xml = 
(
    SELECT  
        CASE WHEN a2.[FORM_OID]='INDEX' THEN a2.[FORM_OID]
        ELSE  SUBSTRING(CONVERT(varchar(10),(a2.[WORKSHEET_ID]-1))+'_'+ a2.[FORM_OID],0,31) 
        END  '@ss:Name',
        (
            SELECT '' as [Column],
            (
                SELECT   
                    'String' as [Cell/Data/@ss:Type],
                    [Col1] as [Cell/Data],
                    '',
                    'String' as [Cell/Data/@ss:Type],
                    [Col2] as [Cell/Data],
                    '',
                    'String' as [Cell/Data/@ss:Type],
                    [Col3] as [Cell/Data],
                    '',
                    5 AS [Cell/@ss:MergeAcross],       --Read below!
                    'String' as [Cell/Data/@ss:Type],
                    [Col4] as [Cell/Data],
                    '',
                    'String' as [Cell/Data/@ss:Type],
                    [Col5] as [Cell/Data],
                    '',
                    'String' as [Cell/Data/@ss:Type],
                    [Col7] as [Cell/Data],
                    '',
                    'String' as [Cell/Data/@ss:Type],
                    [Col8] as [Cell/Data],
                    '',
                    'String' as [Cell/Data/@ss:Type],
                    [Col9] as [Cell/Data],
                    '',
                    'String' as [Cell/Data/@ss:Type],
                    [RANKING] as [Cell/Data]

                    FROM TempData  a1
                    WHERE a1.[WORKSHEET_ID]=a2.[WORKSHEET_ID]
                    GROUP BY [ID],[Col1],[Col2],[Col3],[FORM_OID],[Col4],[Col5],[Col6],[Col7],[Col8],[Col9],[RANKING] --form oid
                    order by [ID]
                FOR XML PATH('Row'), TYPE
            ) 
            FOR XML PATH('Table'), TYPE
        )
    FROM TempData  a2
    group by [WORKSHEET_ID],[FORM_OID]
    FOR XML PATH('Worksheet'), ROOT('Workbook')
);

SELECT @xml;

--CleanUp
--DROP TABLE TempData;

If you look closely to the "Col4" you'll see, that I introduced the MergeAcross-Attribut directly in the call. What do you think about adding more columns to your temp table or define a second temptable for styles if they are 1:n and add them "on the fly"?