Why use xsd.exe tool to create class for XSD schema and then use serializer to create XML file, when you can use XDocument.
Should I be always using XDocument and manually create the XML document by nesting XElement elements, or is there something I am not seeing that makes using XSD schema and XmlSerializer better.
XDocument and XmlSerializer/xsd.exe are useful for different sorts of things.
When you're dealing with a large XML document, you should really have an XSD schema. This is an agreement between all parties on what the structure of the XML document should be, what properties / children / data types, it has etc.
If you have an XSD, then you might as well use it to generate C# classes. You could hand-write XDocument code, but then you need to keep referring back to the XSD to see what properties / children you should be accessing, and what their data types are. If the XSD changes, you have to go and manually search through your XDocument code looking for things which need to be changed.
If instead you generate C# classes, then the properties you can access, and their data types, are right there in the type system: you don't need to refer to the XSD. If the XSD changes, you can re-generate your C# classes and get compiler errors in places you need to change.
If you have an XSD, you should also validate that XML documents you receive match that XSD, see e.g. this answer.
XmlSerializer is often also more efficient than XDocument.
However, sometimes you need to work with someone else's XML document, and either there's no XSD or the generated C# classes are far more complex than you need. Or sometimes you're just doing something very simple and it's not worth writing an XSD. Or sometimes the XML does something really annoying which doesn't map well to C# classes. In those cases, using XDocument is often the better choice.
(The one thing you should avoid at all costs is XmlDocument -- that does the same thing as XDocument, but much worse.)
XmlSerializeris more efficient in that it doesn't need to create a DOM - it is based onXmlReader/XmlWriterIIRC - and it is usually easier to get correctly working deserialization code by not writing it yourself - deserialization is deceptively complex in the edges; but if the DOM-based approach works fine for your data, and you don't need to formally define it in a separate schema: fine, have fun! Do what works!XDocument, you need to keep referring back to the structure of the XML document when you access children / properties, so you know what they're called and what type they have. Withxsd.exe, you get generated C# types which already have all of the right properties / children on them, with the right types -- you don't need to keep looking that information up, as it's right there in the type system. It also means that if your schema is changed, you get a compile-time error, rather than having to find all the places whereXDocumentis used to reference a now-removed propertyxsd.exeagain when you change the XSD, otherwise the generated C# classes won't match the updated XSD. You can separately validate that an XML document which you receive matches an XSD, see e.g. stackoverflow.com/a/752774/1086121XDocumentis often the better choice.