Deserialising JSON to derived types in Asp.Net Web API

Question

I'm calling a method of my WebAPI sending a JSON that I would like to match (or bind) with a model.

In the controller I have a method like:

public Result Post([ModelBinder(typeof(CustomModelBinder))]MyClass model); 

'MyClass', which is given as a parameter is an abstract class. I would like that at, depending of the type of json passed, the correct inherited class is instantiated.

To achieve it, I'm trying to implement a custom binder. The problem is that (I don't know if it's very basic but I can't find anything) I don't know how to retrieve the raw JSON (or better, some kind of serialization) that comes in the request.

I see:

• actionContext.Request.Content

But all methods are exposed as async. I don't know who this fits with passing the generate model to the controller method...

Answer 1

You don't need a custom model binder. Nor do you need to muck about with the request pipeline.

Take a look at this other SO: How to implement custom JsonConverter in JSON.NET to deserialize a List of base class objects?.

I used this as the basis for my own solution to the same problem.

Starting off with the JsonCreationConverter<T> referenced in that SO (slightly modified to fix issues with serialization of types in responses):

public abstract class JsonCreationConverter<T> : JsonConverter {     /// <summary>     /// this is very important, otherwise serialization breaks!     /// </summary>     public override bool CanWrite     {         get         {             return false;         }     }     /// <summary>      /// Create an instance of objectType, based properties in the JSON object      /// </summary>      /// <param name="objectType">type of object expected</param>      /// <param name="jObject">contents of JSON object that will be      /// deserialized</param>      /// <returns></returns>      protected abstract T Create(Type objectType, JObject jObject);      public override bool CanConvert(Type objectType)     {         return typeof(T).IsAssignableFrom(objectType);     }      public override object ReadJson(JsonReader reader, Type objectType,       object existingValue, JsonSerializer serializer)     {         if (reader.TokenType == JsonToken.Null)             return null;         // Load JObject from stream          JObject jObject = JObject.Load(reader);          // Create target object based on JObject          T target = Create(objectType, jObject);          // Populate the object properties          serializer.Populate(jObject.CreateReader(), target);          return target;     }      public override void WriteJson(JsonWriter writer, object value,        JsonSerializer serializer)     {         throw new NotImplementedException();     } }  

And now you can annotate your type with the JsonConverterAttribute, pointing Json.Net to a custom converter:

[JsonConverter(typeof(MyCustomConverter))] public abstract class BaseClass{   private class MyCustomConverter : JsonCreationConverter<BaseClass>   {      protected override BaseClass Create(Type objectType,         Newtonsoft.Json.Linq.JObject jObject)      {        //TODO: read the raw JSON object through jObject to identify the type        //e.g. here I'm reading a 'typename' property:         if("DerivedType".Equals(jObject.Value<string>("typename")))        {          return new DerivedClass();        }        return new DefaultClass();         //now the base class' code will populate the returned object.      }   } }  public class DerivedClass : BaseClass {   public string DerivedProperty { get; set; } }  public class DefaultClass : BaseClass {   public string DefaultProperty { get; set; } } 

Now you can use the base type as a parameter:

public Result Post(BaseClass arg) {  } 

And if we were to post:

{ typename: 'DerivedType', DerivedProperty: 'hello' } 

Then arg would be an instance of the DerivedClass, but if we posted:

{ DefaultProperty: 'world' } 

Then you'd get an instance of the DefaultClass.

EDIT - Why I prefer this method to TypeNameHandling.Auto/All

I do believe that using the TypeNameHandling.Auto/All espoused by JotaBe is not always the ideal solution. It might well be in this case - but personally I won't do it unless:

• My API is only ever going to be used by me or my team
• I don't care about having a dual XML-compatible endpoint

When Json.Net TypeNameHandling.Auto or All are used, your web server will start sending out type names in the format MyNamespace.MyType, MyAssemblyName.

I have said in comments that I think this is a security concern. Mention was made of this in some documentation I read from Microsoft. It's not mentioned any more, it seems, however I still feel it's a valid concern. I don't ever want to expose namespace-qualified type names and assembly names to the outside world. It's increasing my attack surface. So, yes, I can not have Object properties/parameters my API types, but who's to say the rest of my site is completely hole-free? Who's to say a future endpoint doesn't expose the ability to exploit type names? Why take that chance just because it's easier?

Also - if you are writing a 'proper' API, i.e. specifically for consumption by third-parties and not just for yourself, and you're using Web API, then you're most likely looking to leverage the JSON/XML content-type handling (as a minimum). See how far you get trying to write documentation that's easy to consume, which refers to all your API types differently for XML and JSON formats.

By overriding how JSON.Net understands the type names, you can bring the two into line, making the choice between XML/JSON for your caller purely based on taste, rather than because the type names are easier to remember in one or the other.

Answer 2

You don't need to implement it by yourself. JSON.NET has native support for it.

You have to specify the desired TypeNameHandling option for the JSON formatter, like this (in global.asax application start event):

JsonSerializerSettings serializerSettings = GlobalConfiguration.Configuration    .Formatters.JsonFormatter.SerializerSettings; serializerSettings.TypeNameHandling = TypeNameHandling.Auto; 

If you specify Auto, like in the above sample, the parameter will be deserialized to the type specified in the $type property of the object. If the $type property is missing, it will be deserialized to the parameter's type. So you only have to specify the type when you're passing a parameter of a derived type. (This is the most flexible option).

For example, if you pass this parameter to a Web API action:

var param = {     $type: 'MyNamespace.MyType, MyAssemblyName', // .NET fully qualified name     ... // object properties }; 

The parameter will be deserialized to an object of MyNamespace.MyType class.

This also works fro sub-properties, i.e., you can have an object like this, which specifies that an inner property is of a given type

var param = {     myTypedProperty: {       $type: `...`       ... }; 

Here you can see a sample on JSON.NET documentation of TypeNameHandling.Auto.

This works at least since JSON.NET 4 release.

NOTE

You don't need to decorate anything with attirbutes, or do any other customization. It will work without any changes in your Web API code.

IMPORTANT NOTE

The $type must be the first property of the JSON serialized object. If not, it will be ignored.

COMPARISON TO CUSTOM JsonConverter/JsonConverterAttribute

I'm comparing the native solution to this answer.

To implement the JsonConverter/JsonConverterAttribute:

• you need to implement a custom JsonConverter, and a custom JsonConverterAttribute
• you need to use attributes to mark the parameters
• you need to know beforehand the possible types expected for the parameter
• you need to implement, or change the implementation of your JsonConverter whenever your types or properties change
• there is a code smell of magic strings, to indicate the expected property names
• you are not implementing something generic that can be used with any type
• you're reinventing the wheel

In the author of the answer there's a comment regarding security. Unless you do something wrong (like accepting a too generic type for your parameter, like Object) there is no risk of getting an instance of the wrong type: JSON.NET native solution only instantiates an object of the parameter's type, or a type derived from it (if not, you get null).

And these are the advantages of JSON.NET native solution:

• you don't need to implement anything (you only have to configure the TypeNameHandling once in your app)
• you don't need to use attributes in your action parameters
• you don't need to know the possible parameter types beforehand: you simply need to know the base type, and specify it in the parameter (it could be an abstract type, to make polymorphism more obvious)
• the solution works for most cases (1) without changing anything
• this solution is widely tested, and optimized
• you don't need magic strings
• the implementation is generic, and will accept any derived type

(1): if you want to receive parameter values that don't inherit from the same base type, this will not work, but I see no point on doing so

So I can't find any disadvantages, and find many advantages on JSON.NET solution.

WHY USING CUSTOM JsonConverter/JsonConverterAttribute

This is a good working solution that allows customization, that can be modified or extended to adapt it to your particular case.

If you want to do something that the native solution cannot do, like customizing the type names, or inferring the type of the parameter based on available property names, then do use this solution adapted to your own case. The other one cannot be customized, and will not work for your needs.