Question

I am working on an application in which, i have to read XML files that have a different set of nodes each time, although only a certain number of nodes appear in all the files, the combination in which they appear keep on changing, the XML files are generated by another system which i cannot control, I am looking into Linq to XML and XML serialization, but i guess serialization is not a choice since it needs pre-built classes to create objects.

Example XML data

<Employee>
  <PersonalInfo>
    <FirstName>Vamsi</FirstName>
    <LastName>Krishna</LastName>
  </PersonalInfo>
  <EmploymentInfo>
    <Department>
      <Id>101</Id>
      <Position>SD</Position>
    <Department>
  <EmploymentInfo>
</Employee>

Another Format

<Employee>
  <PersonalInfo>
    <FirstName>Vamsi</FirstName>
    <LastName>Krishna</LastName>
  </PersonalInfo>      
</Employee>

You can observe that EmploymentInfo node is completely missing in the second example, there are many number of combinations in which the XML data can be presented to the application, I have to read the XML file validate it insert into an SQL Server database through my C# code.

Answer 1

I’d say it depends.

If you just want to communicate with another system in a strongly-typed way, and you can expect the XML schemas to not be changing very frequently, you might be OK with XML serialization. Just encapsulate the deserialization into a separate component and write different versions of them (yes, you’ll need to be able to determine the schema version that is currently used). I mean, each version would have it’s own set of classes that are targeted by the serializer.

But if you really cannot infer a system out of the schemas used by the external app and need some intelligent parser, you’d better use XPath or Linq to XML or some other XML-level APIs to manually handle the XML-s.

BTW, both of your samples are pretty easy for the XMLSerializer. In the second case it will just set EmploymentInfo to null.