How to update xml in table having multiple same elements in xml, from data by joining another table

Question

I have table which is having XML column. Now this XML column have multiple tags like an array. Each of these tags have an attribute() which i want to update from a table by joining it.

Table_A is as following

Id    XML
1     "<Root><Object ObjId = "1" Text = "A"><Object Id = "2" Text = "B"></Root>"
2     "<Root><Object ObjId = "1" Text = "M"><Object Id = "12" Text = "N"></Root>"

Each of this XML is as following

<Root>
   <Object Id = "1" Text = "A">
   <Object Id = "2" Text = "B">
   <Object Id = "3" Text = "C">
   <Object Id = "4" Text = "D">
   <Object Id = "5" Text = "E">
</Root>

Table_B i want to join to update records in XML:

Table_A_Id    ObjId      Value
1             1          "Q"
1             2          "R"
2             1          "S"
2             12         "T"

I couldn't find the solution to create a dynamic path in XML.modify method. Following is the query i have started, which might explain what i am trying to do.

UPDATE TblA SET XML.modify(replace value of (/Root/Object) with Table_B.Value)
FROM TABLE_A TblA
CROSS APPLY TblA.XML.nodes('/Root/Object') AS xmlObjs(obj)
INNER JOIN Table_B TblB ON TblB.Table_A_Id = TblA.Id AND TblB.ObjId = xmlObjs.obj.value('@ObjId[1]','varchar(MAX)')

Answer 1

What you seem to know is: .modify() will not allow more than one change per call. You would have to use a CURSOR or WHILE loop in order to update each occurance one after the other.

Therefore I'd suggest this approach:

First we create mockup tables to simulate your issue:

DECLARE @tblA TABLE(Id INT, [XML] XML)
INSERT INTO @tblA VALUES
(1,N'<Root>
       <Object Id = "1" Text = "A"/>
       <Object Id = "2" Text = "B"/>
       <Object Id = "5" Text = "E"/>
   </Root>')
,(2,N'<Root>
       <Object Id = "1" Text = "F"/>
       <Object Id = "2" Text = "G"/>
       <Object Id = "12" Text = "J"/>
       <Object Id = "13" Text = "J"/>
   </Root>')

DECLARE @tblB TABLE(Table_A_Id INT,[ObjId] INT,[Value] VARCHAR(10));
INSERT INTO @tblB VALUES
 (1,1 ,'Q')
,(1,2 ,'R')
,(2,1 ,'S')
,(2,12,'T');

--The query

WITH cte AS
(
    SELECT tA.Id
          ,tA.[XML]
          ,(
            SELECT A.obj.value('@Id','int') AS [@Id]
                  ,COALESCE(tB.[Value],A.obj.value('@Text','varchar(max)')) AS [@Text]
            FROM tA.[XML].nodes('/Root/Object') A(obj)
            LEFT JOIN @tblB tB ON tB.Table_A_Id=tA.Id AND tB.[ObjId]=A.obj.value('@Id','int')
            FOR XML PATH('Object'),ROOT('Root'),TYPE
           ) AS NewXml
    FROM @tblA tA
)
UPDATE cte SET cte.[Xml]=NewXml;

--Check the result

SELECT * FROM @tblA;

The idea in short:

• We use an updateable CTE to read table A's data into a list
• We use a correlated sub-query to create the modified XML on the fly.
• We write the new XML over the existing one.

UPDATE

Next time please try to avoid a chameleon question... Your comment let this change into something entirely different... Next time please close one question, if it is answered as is and start a new question in case you find, that your initial question did not really cover your needs...

Try this:

DECLARE @tblA TABLE(Id INT, [XML] XML)
INSERT INTO @tblA VALUES
(1,N'<Root> 
       <AnotherTag val="abc"/> 
       <AnotherTag val="DEF"/> 
       <Object Id =  "1" Text = "F" OtherProperty = "123" /> 
       <Object Id =  "2" Text = "G" SampleProperty = "Anything" DataProperty="Sample Data" /> 
       <Object Id = "12" Text = "I" OtherProperty = "123"/> 
       <Object Id = "13" Text = "J" DataProperty = "Sample"/> 
     </Root>')
,(2,N'<Root>
       <Object Id = "1" Text = "F"/>
       <Object Id = "2" Text = "G"/>
       <Object Id = "12" Text = "I"/>
       <Object Id = "13" Text = "J"/>
   </Root>')

DECLARE @tblB TABLE(Table_A_Id INT,[ObjId] INT,[Value] VARCHAR(10));
INSERT INTO @tblB VALUES
 (1,1 ,'Q')
,(1,2 ,'R')
,(2,1 ,'S')
,(2,12,'T');

WITH cte AS
(
    SELECT ta.Id
          ,ta.[XML]
          ,Combined.[CombXml].query('<Root>
                                     {
                                        for $elmt in /combined/embedded/Root/*
                                        let $bVal := /combined/b_data[ObjId[1] = $elmt/@Id]/Value/text()
                                        return
                                        if(local-name($elmt) eq "Object") then 
                                            <Object Id="{$elmt/@Id}" Text="{if(empty($bVal)) then $elmt/@Text else $bVal}">
                                            {$elmt/@*[local-name() != "Id" and local-name() != "Text"]}
                                            </Object>
                                        else
                                            $elmt
                                     }
                                     </Root>') AS NewXml
    FROM @tblA ta
    OUTER APPLY(SELECT (SELECT [ObjId],[Value] 
                        FROM @tblB tb 
                        WHERE tb.Table_A_Id=ta.Id 
                        FOR XML PATH('b_data'),TYPE) AS [*]
                       ,ta.[XML] AS [embedded]
                       FOR XML PATH(''),ROOT('combined'),TYPE) Combined([CombXml])
)
UPDATE cte SET [XML] = NewXml;

SELECT * FROM @tblA;

The idea behind:

We need to get the side data into the XML in order to use XQuery-FLWOR.

The APPLY will create a XML like this:

<combined>
  <b_data>
    <ObjId>1</ObjId>
    <Value>Q</Value>
  </b_data>
  <b_data>
    <ObjId>2</ObjId>
    <Value>R</Value>
  </b_data>
  <embedded>
    <Root>
      <AnotherTag val="abc" />
      <AnotherTag val="DEF" />
      <Object Id="1" Text="F" OtherProperty="123" />
      <Object Id="2" Text="G" SampleProperty="Anything" DataProperty="Sample Data" />
      <Object Id="12" Text="J" OtherProperty="123" />
      <Object Id="13" Text="J" DataProperty="Sample" />
    </Root>
  </embedded>
</combined>

Against this combined XML we can run a FLWOR query using .query().

• This query will run through all nodes below <Root>
• the corresponding <Value> of your side data (in the XML its <b_data>) is assigned to $bVal.
• Now we check, if the current element's name is <Object>.
• If so, we write the attributes Id and Text directly and add all other attributes without looking into them.
• If not, we just return $elmt as is.

The result for the Id=1 looks like this afterwards:

<Root>
  <AnotherTag val="abc" />
  <AnotherTag val="DEF" />
  <Object Id="1" Text="Q" OtherProperty="123" />
  <Object Id="2" Text="R" SampleProperty="Anything" DataProperty="Sample Data" />
  <Object Id="12" Text="I" OtherProperty="123" />
  <Object Id="13" Text="J" DataProperty="Sample" />
</Root>

You can see, that "Q" and "R" are changed where Id is 1 or 2 - according to the side data.

I must admit, that this gets complicated...

Depending on your real data and the complexity of your XML a looping approach with .modify() might be better...