Deserialize object property with StringReader vs XmlNodeReader

Question

Why does XmlSerializer populate my object property with an XmlNode array when deserializing an empty typed element using XmlNodeReader instead of an empty string like it does when using StringReader (or XmlTextReader)?

The second assertion in the following code sample fails:

var doc = new XmlDocument();
doc.Load(new StringReader(@"
    <Test xmlns:xsi=""http://www.w3.org/2001/XMLSchema-instance""
          xmlns:xsd=""http://www.w3.org/2001/XMLSchema"">
      <Value xsi:type=""xsd:string"" />
    </Test>"));
var ser = new XmlSerializer(typeof (Test));

var reader1 = new StringReader(doc.InnerXml);
var obj1 = (Test) ser.Deserialize(reader1);
Debug.Assert(obj1.Value is string);

var reader2 = new XmlNodeReader(doc.FirstChild);
var obj2 = (Test) ser.Deserialize(reader2);
Debug.Assert(obj2.Value is string);

public class Test
{
    public object Value { get; set; }
}

I'm guessing it has something to do with the null internal NamespaceManager property but I'm not sure how to work around this mysterious limitation. How can I reliably deserialize a subset of my parsed XML document without converting it back into a string and re-parsing?

Answer 1

It looks like this is a very old XmlNodeReader bug that Microsoft have no intention of fixing. (Archived Microsoft Connect link here). I found a workaround on Lev Gimelfarb's blog here that adds namespaces to the reader's NameTable as prefixes are looked up.

public class ProperXmlNodeReader : XmlNodeReader
{
    public ProperXmlNodeReader(XmlNode node) : base(node)
    {
    }

    public override string LookupNamespace(string prefix)
    {
        return NameTable.Add(base.LookupNamespace(prefix));
    }
}