How to remove DataContractSerializer verbosity

Question

We are attempting to serialize an object tree. And while we have been successful. I was hoping to find a way to simplify down the generated xml.

The objects look something like the following:

public class RuleSet<T>
{
    public IEnumerable<IState<T>> States { get; set; }
    public IEnumerable<ICondition<T>> Conditions { get; set; }
}

public class State<T> : IState<T>
{
    public string Id { get; set; }
    public List<ITransition<T>> Transitions { get; set; }
}

public class Transition<T> : ITransition<T>
{
    public ICondition<T> Condition { get; set; }
    public IState<T> Next { get; set; }
}

public class Condition<T> : ICondition<T>
{
    public string Id { get; set; }
    public string Name { get; set; }    
}

We are using a really simple serialization code at the moment:

public void blah()
{
    var condition1 = new Condition<object>() {
        Id = "C1", AttributeName = "Foo", ExpectedValue = "Bar"
    };
    var condition2 = new Condition<object>() {
        Id = "C2", AttributeName = "Bar", ExpectedValue = "Foo"
    };

    var state1Transitions = new List<ITransition<object>>();
    var state2Transitions = new List<ITransition<object>>();
    var state3Transitions = new List<ITransition<object>>();

    var state = new State<object> {
        Id = "S1", Transitions = state1Transitions
    };
    var state2 = new State<object> {
        Id = "S2", Transitions = state2Transitions
    };
    var state3 = new State<object> {
        Id = "S3", Transitions = state3Transitions
    };

    state1Transitions.Add(new Transition<object> {
        Condition = condition1, Next = state2
    });
    state1Transitions.Add(new Transition<object> {
        Condition = condition2, Next = state3
    });
    state2Transitions.Add(new Transition<object> {
        Condition = condition2, Next = state3
    });
    var ruleSet = new RuleSet<object> {
        States = new List<IState<object>> {state, state2, state3}, 
        Conditions = new List<ICondition<object>>{condition1, condition2}
    };

    var stream1 = new MemoryStream();

    var serializer = new DataContractSerializer(typeof(RuleSet<object>),
        new List<Type> { 
            typeof(State<object>), 
            typeof(Transition<object>),
        typeof(AttributeEqualTo<object>)
    });

    serializer.WriteObject(stream1, ruleSet);
    stream1.Position = 0;

    var xml = new StreamReader(stream1).ReadToEnd();
    Console.WriteLine(xml);
}

When the XML is generated the output for each level is complete instead of only containing the reference to the object. Basically for each Transition<T> we get a complete object definition for each state and condition even if they are defined elsewhere.

Is there a way to get those to simply be references?

Answer 1

It all comes down to how you create your DataContractSerializer.

You'll want to call the overload of the constructor that allows you to indicate that you will preserve object references, with the following signature:

public DataContractSerializer(
    Type type,
    IEnumerable<Type> knownTypes,
    int maxItemsInObjectGraph,
    bool ignoreExtensionDataObject,
    bool preserveObjectReferences,
    IDataContractSurrogate dataContractSurrogate
)

You can pass default values for most of the parameters. In your case, the call to the DataContractSerializer constructor will look like this:

var serializer = new DataContractSerializer(typeof(RuleSet<object>),
    new [] {
        typeof(State<object>),
        typeof(Transition<object>),
        typeof(AttributeEqualTo<object>)
    },
    Int32.MaxValue,
    false,
    /* This is the important flag to set. */
    true,
    null
);

Note, from the preserveObjectReferences parameter documentation, it uses non-standard XML (emphasis mine):

preserveObjectReferences

Type: System.Boolean

true to use non-standard XML constructs to preserve object reference data; otherwise, false.

If you need other code outside of .NET to interpret this, then untangling the references might prove to be difficult (but shouldn't be impossible).

However, it prevents the object graph from replicating itself over and reduce the size of your XML (possibly considerably, given how deep your references go).