使用XmlSerializer以可扩展的方式对任意元素进行反序列化

Question

I am using XmlSerializer to deserialize XML responses from a remote API. There are a number of standard predefined classes, and I have built up classes for these. The key part of the response looks like this:

<data>
    <employee>
        <name>John Doe</name>
        <id>E1</id>
        ....
    </employee>
    <employee>
        ....
    </employee>
    ....
</data>

or

<data>
    <customer>
        <name>Yoyodyne Systems</name>
        <id>C1</id>
        ....
    </customer>
    <customer>
        ....
    </customer>
    ....
</data>

So I have Customer and Employee classes, and the Data class has properties defined like this:

    [XmlElement("customer", typeof(Customer))]
    public List<Customer> Customers { get; set; }

    [XmlElement("employee", typeof(Employee))]
    public List<Employee> Employees { get; set; }

The data element only contains one type of tag per result, and I generally know which one is expected to be returned from a given request, so I just access that property and ignore the other ones which are left empty.

But I want to make this more extensible; in particular there may be different instances of the remote application with custom classes in them, and I don't want to put them all into the base code package. I was hoping I could add a property to the Data class something like

    public List<object> ExtensionObjects { get; set; }

and then use the XmlSerializer(Type, Type[]) constructor to supply the actual types I want to deserialize. However, I don't know how to tell it to apply this to any unknown tags, or even to specify the tags. For instance, if I have a custom Office class, how can I make it that <office>...</office> tags get deserialized into it? Is there a way I can do this with XmlSerializer without having to go all the way back to parsing everything as an XDocument?

Edit:

One thing that seems promising is trying to handle the OnUnknownElement event; I can create a delegate for it and then pass it in via an XmlDeserializationEvents object to the Deserialize method call. Now I just have to figure out the best way to turn those unknown elements into instances of my extension classes.

Answer 1

I have determined one way to do this, though of course there may be better ones.

As I mentioned above, set up a handler for OnUnknownElement:

XmlDeserializationEvents events = new XmlDeserializationEvents();
events.OnUnknownElement = delegate(object sender, XmlElementEventArgs args)
{
    if (args.Element.Name == "office")
    {
        var data = (Data) args.ObjectBeingDeserialized;
        var item = new Office(args.Element);
        data.ExtensionObjects.Add(item);
    }
};
var response = (Response) serializer.Deserialize(xmlReader, events);

Then all you need do is implement the constructor to turn that XmlElement into your object. There may be a cleaner way to do it, but considering this is a simple object (just "name" and "id" properties) I'm just assuming the element is just a flat list of named elements.

public Office(XmlElement e)
{
    var valDict = e.ChildNodes.Cast<XmlNode>().Where(n => n is XmlElement)
                   .ToDictionary(x => x.Name, x => x.InnerText);

    Name = valDict["name"];
    Id = valDict["id"];
}

If there's a better way to do this that can go through the XmlSerializer again, I was unable to find it, since I couldn't convert an XmlElement to an XmlReader or anything else that the Deserialize method could take in. I suppose I could probably serialize it to text again and then reparse it, but that seems kludgy.