Question

Is there a way to parse (in a T-SQL script) the following XML without using multiple CROSS-APPLY statements? As you can notice the Root’s children have different names ( they are completely different — don’t only differ in the last digit index ).

I would like to shred this XML into a table where every MeasuredParamName* is a column and every value represents a row. Each MeasuredParamName element will have the same number of children.

ie. (MeasuredParamName1, MeasuredParamName2, MeasuredParamName3, MeasuredParamName4)

Is there a performance concen with using cross-apply for this solution? Would a different structure for the XML be more suitable?

Ex:

<Data>
<MeasuredParamName1>
    <Value>1</Value>
    <Value>2</Value>
</MeasuredParamName1>
<MeasuredParamName2>
    <Value>3</Value>
    <Value>4</Value>
</MeasuredParamName2>
<MeasuredParamName3>
    <Value>0.5</Value>
    <Value>0.10</Value>
</MeasuredParamName3>
<MeasuredParamName4>
    <Value>20</Value>
    <Value>30</Value>
</MeasuredParamName4>   
   </Data>

Answer 1

Try this.

select  
    MeasuredParamName1,MeasuredParamName2,MeasuredParamName3,MeasuredParamName4
from
(
    select  
        x.n.value('local-name(..)', 'varchar(50)') mpname,
        x.n.value('.','varchar(20)') value,
        row_number() over (partition by x.n.value('local-name(..)', 'varchar(50)') order by x.n) as rn  
    from
        @xml.nodes('/Data//Value') x(n)
) v
pivot
(max(value) for mpname in (MeasuredParamName1,MeasuredParamName2,MeasuredParamName3,MeasuredParamName4)) p

If I were designing the XML, I would put the “row” values inside the columns.
eg

<Data>
   <Value>
       <MeasuredParam1>5</MeasuredParam1>
       <MeasuredParam2>6</MeasuredParam2>
       <MeasuredParam3>7</MeasuredParam3>
       <MeasuredParam4>8</MeasuredParam4>
   </Value>
   <Value>
       <MeasuredParam1>11</MeasuredParam1>
       <MeasuredParam2>12</MeasuredParam2>
       <MeasuredParam3>13</MeasuredParam3>
       <MeasuredParam4>14</MeasuredParam4>
   </Value>
</Data>