After reading the MSDN reference, I still have questions about when to use the KnownType attribute. I understand that the attribute communicates type information to the serializer, but when is this needed? Is it appropriate when the class being serialized has references of a base class type, and there are up-cast derivative classes that could be set to those references?
Moreover, are there any drawbacks to overusing the attribute? For instance, in the previous example, if the serialized class was marked with KnownType(baseClass) even though there was an explicit reference to that type?
[KnownType]
is needed to tell it about subtypes. The disadvantage of not using it is that the following won't work:
[DataContract]
class Foo {}
[DataContract]
class Bar : Foo {}
with a method on the WCF interface that returns:
public Foo GetFoo() { return new Bar(); }
Without the attribute, the serializer (especially for mex/proxy-generated types) won't know about Bar
, and it will fail. With the attribute:
[DataContract, KnownType(typeof(Bar))]
class Foo {}
it will work. This only applies to DataContractSerializer
- with NetDataContractSerializer
you get the type data in a different way.