Why is XmlSerializer's Deserialize() spitting out a child object which is a XmlNode[]?

问题

I'm using XmlSerializer to serialize and then deserialize a simple object. When I deserialize the object to my surprise I find a child object was not properly deserialized but instead turned into XmlNode[].

Here is very nearly the structure I've got:

// This line I put in here as a way of sneaking into the XML the 
// root node's C# namespace, since it's not the same as the 
// deserializing code and the deserializing code seemed unable to 
// deserialize properly without knowing the Type (see my code below).
// So I basically just use this fake construct to get the namespace 
// and make a Type of it to feed the XmlSerializer() instantiation.
[XmlRoot(Namespace = "http://foo.com/CSharpNamespace/Foo.Bar")]   

// This is because QueuedFile can be given to the Argument array.
[XmlInclude(typeof(QueuedFile))]
// This class is Foo.Bar.CommandAndArguments
public class CommandAndArguments {
    public String Command;
    public object[] Arguments;
}

// I don't think this matters to XmlSerialize, but just in case...
[Serializable()] 

// I added this line just thinking maybe it would help, but it doesn't
// do anything.  I tried it without the XmlType first, and that
// didn't work.
[XmlType("Foo.Baz.Bat.QueuedFile")]

// This class is Foo.Baz.Bat.QueuedFile (in a different c# 
// namespace than CommandAndArguments and the deserializing code)
public QueuedFile {
    public String FileName;
    public String DirectoryName;
}

And the code which deserializes it looks like:

public static object DeserializeXml(String objectToDeserialize)
        {
            String rootNodeName = "";
            String rootNodeNamespace = "";

            using (XmlReader xmlReader = XmlReader.Create(new StringReader(objectToDeserialize)))
            {
                if (xmlReader.MoveToContent() == XmlNodeType.Element)
                {
                    rootNodeName = xmlReader.Name;
                    rootNodeNamespace = xmlReader.NamespaceURI;

                    if (rootNodeNamespace.StartsWith("http://foo.com/CSharpNamespace/"))
                    {
                        rootNodeName = rootNodeNamespace.Substring("http://foo.com/CSharpNamespace/".Length) + "." +
                                       rootNodeName;
                    }
                }
            }

            //MessageBox.Show(rootNodeName);
            try
            {
                Type t = DetermineTypeFromName(rootNodeName);

                if (t == null)
                {
                    throw new Exception("Could not determine type of serialized string.  Type listed as: "+rootNodeName);                    
                }

                var s = new XmlSerializer(t);
                return s.Deserialize(new StringReader(objectToDeserialize));

                // object o = new object();
                // MethodInfo castMethod = o.GetType().GetMethod("Cast").MakeGenericMethod(t);
                // return castMethod.Invoke(null, new object[] { s.Deserialize(new StringReader(objectToDeserialize)) });
            }
            catch (InvalidOperationException)
            {
                return null;
            }
        }

And here is the XML when the CommandAndArguments is serialized:

<?xml version="1.0" encoding="utf-16"?>
<CommandAndArguments xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:xsd="http://www.w3.org/2001/XMLSchema" xmlns="http://foo.com/CSharpNamespace/Foo.Bar">
  <Command>I am a command</Command>
  <Arguments>
    <anyType xsi:type="Foo.Baz.Bat.QueuedFile">
      <FileName xmlns="">HelloWorld.txt</FileName>
      <DirectoryName xmlns="">C:\foo\bar</DirectoryName>
    </anyType>
  </Arguments>
</CommandAndArguments>

But when I deserialize I am given a CommandAndArguments object where Arguments is XmlNode[] with the first item being the attribute giving the QueuedFile as the type and the other indices being elements of the properties. But why wasn't the QueuedFile object recreated?

I suspect this might somehow have do with C# namespaces and the engine doing the deserializing not being able to find or work with QueuedFile... But I don't see why since when I forgot the XmlInclude() it made sure to tell me it didn't expect QueuedFile and now that I've added the XmlInclude() I get no error, just an incomplete deserialization.

Help? I've read everything I can find to read and Googled everything I know to Google and am stuck. I certainly have a lot to learn about XML serialization but I'm not sure how I'm failing at something which should be pretty simple (I actually did something almost exactly like this before without any problem, the only difference then was that everything was in the same C# namespace).

回答1:

Are you able to change the XML format or is it fixed? I don't know what the problem you are having is, but I use the DataContractSerializer classes extensively with no problems.

http://msdn.microsoft.com/en-us/library/system.runtime.serialization.datacontractserializer.aspx

public static void WriteObject(string fileName)
        {
            Console.WriteLine(
                "Creating a Person object and serializing it.");
            Person p1 = new Person("Zighetti", "Barbara", 101);
            FileStream writer = new FileStream(fileName, FileMode.Create);
            DataContractSerializer ser =
                new DataContractSerializer(typeof(Person));
            ser.WriteObject(writer, p1);
            writer.Close();
        }

        public static void ReadObject(string fileName)
        {
            Console.WriteLine("Deserializing an instance of the object.");
            FileStream fs = new FileStream(fileName,
            FileMode.Open);
            XmlDictionaryReader reader =
                XmlDictionaryReader.CreateTextReader(fs, new XmlDictionaryReaderQuotas());
            DataContractSerializer ser = new DataContractSerializer(typeof(Person));

            // Deserialize the data and read it from the instance.
            Person deserializedPerson =
                (Person)ser.ReadObject(reader, true);
            reader.Close();
            fs.Close();
            Console.WriteLine(String.Format("{0} {1}, ID: {2}",
            deserializedPerson.FirstName, deserializedPerson.LastName,
            deserializedPerson.ID));
        }

回答2:

To anyone coming along with a similar problem, depending on your situation you're probably better off with NetDataContractSerializer. It is an alternative to DataContractSerializer which records the .Net types in the XML making deserialization a breeze, since it knows exactly what types are involved and thus you do not need to tell it what type the root object is with the deserialize command. And it can produce output in XML or binary form (I prefer XML for easier debugging).

Here is some sample code for easily serializing and deserializing an object to and from a string:

private static object Deserialize(string xml)
{
    object toReturn = null;

    using (Stream stream = new MemoryStream())
    {
        byte[] data = System.Text.Encoding.UTF8.GetBytes(xml);
        stream.Write(data, 0, data.Length);
        stream.Position = 0;
        var netDataContractSerializer = new NetDataContractSerializer();

        toReturn = netDataContractSerializer.Deserialize(stream);
    }

    return toReturn;
}

private static string Serialize(object obj)
{
    using (var memoryStream = new MemoryStream())
    using (var reader = new StreamReader(memoryStream))
    {
        var netDataContractSerializer = new NetDataContractSerializer();
        netDataContractSerializer.Serialize(memoryStream, obj);

        memoryStream.Position = 0;
        return reader.ReadToEnd();
    }
}

Easy as pie!

来源：https://stackoverflow.com/questions/13506196/why-is-xmlserializers-deserialize-spitting-out-a-child-object-which-is-a-xmln

标签

.net

serialization

xmlserializer