.NET XmlSerializer and multiple references to the same object

Question

My repository has List, List and List where an Enrolment has Enrolment.Student and Enrol

Answer 1

You should/can use Reference Tracking with the datacontract serializer:

//deserilaize:
using(MemoryStream memStmBack = new MemoryStream()) {
  var serializerForth = new DataContractSerializer(
    typeof(YourType),
    null,
    0x7FFF /*maxItemsInObjectGraph*/ ,
    false /*ignoreExtensionDataObject*/ ,
    true /*preserveObjectReferences*/ ,
    null /*dataContractSurrogate*/ );

  byte[] data = System.Text.Encoding.UTF8.GetBytes(xml);
  memStmBack.Write(data, 0, data.Length);
  memStmBack.Position = 0;
  var lsBack = (YourType) serializerForth.ReadObject(memStmBack);

}
//serialize...
using(MemoryStream memStm = new MemoryStream()) {
    var serializer = new DataContractSerializer(
      typeof(YourType),
      knownTypes,
      0x7FFF /*maxItemsInObjectGraph*/ ,
      false /*ignoreExtensionDataObject*/ ,
      true /*preserveObjectReferences*/ ,
      null /*dataContractSurrogate*/ );

    serializer.WriteObject(memStm, yourType);

    memStm.Seek(0, SeekOrigin.Begin);

    using(var streamReader = new StreamReader(memStm)) {
        result = streamReader.ReadToEnd();

Or use

[Serializable]
[DataContract(IsReference = true)]

Answer 2

You can implement interface IXmlSerializable to Enrolment and in WriteXml method generate student and course XML which will contains only keys e.g.:

<Student Id="5"/>
<Course Id="6"/>

and in ReadXml method you can load references from this. You must also set XmlIgnore attribute to Student and Course property.

Answer 3

There is no solution for this issue using the XML Serializer. It does not have a concept of identity that it might use to remove duplication.

The best you can do is to serialize the pool of objects separately from their references. You could then recreate your lists after deserialization.

BTW, are you aware that the XmlSerializer is not specific to C#?

Answer 4

How does this sound as a solution:

1. XMLIgnore each secondary reference ie Enrolment.Student & Enrolment.Course
2. create a property for each secondary reference that is used to serialize/deserialize a foreign key for that reference instead - Prefix with XML_FK. eg XML_FK_Student & XML_FK_Course
3. Create a method XML_FinalizeDeserialization that is called after deserialization to load the references using those foreign key properties.

Answer 5

Oh the pains of serialization :-> ...

There was never a generic solution for this, I guess that's why MS stripped it out of the Silverlight framework.

I never rely on any automatic serialization mechanisms of the .net framework. For my own models and repositories, I usually know or can easily programmatically determine which properties are simple scalar ones (numbers/strings/etc) and which are links to other objects (as well as which are lists of either).

There are basically 2 scenarios:

1: We want to serialize/transfer only the flat information of objects. In that case I transfer only the respective IDs for properties that link to other objects. The receiver can then make subsequent queries to get all other objects they need.

2: We want to transfer as much information as possible, i.e. deeper nested XML with several levels, mostly for some reporting functionality displaying everything directly using merely some CSS on the XML. In that case, it is actually desired that objects that are the same will be resolved multiple times into the XML tree.

Sometimes I need to tweak the first scenario a little bit in order to avoid too many subsequent query calls, but usually I get along very well. I.e. I have built into our code base that we can specify which additional objects we want to resolve when, and/or it's configured somewhere.