Why?

Sooner or later we would need to serialize/deserialize our lovely discriminated unions and OneOf doesn't come with built-in serialization/deserialization. OneOf.Serialization was created mainly to solve this challange by giving convention-based serialization/deserialization when using the powerful tool OneOf.

NOTE: OneOf.Serialization has direct dependecy on Newtonsoft.Json.

How to use it?

1. Take your discriminated union and annotate it with JsonConvert attribute, giving OneOfJsonConverter<T> as an argument (where T is your discriminated union)

[JsonConvert(typeof(OneOfJsonConverter<Status>))]
public class Status : OneOfBase<Idle, Running, Completed> {
    
    // Recommended way of seamless consuming/creating cases
    public Status(Idle idle) : base(0, idle) {}
    public Status(Running running) : base(1, null, idle) {}
    public Status(Completed completed) : base(2, null, null, completed)

    public static implicit operator Status(Idle value) => value == null? null : new Status(value);
    public static implicit operator Status(Running value) => value == null? null : new Status(value);
    public static implicit operator Status(Completed value) => value == null? null : new Status(value);
}

2. Take all cases of your discriminated union and inherit OneOfCase marker class

public class Idle : OneOfCase {}

public class Running : OneOfCase 
{
    public DateTime StartedAt { get; }

    public Running(DateTime startedAt) {
        StartedAt = startedAt;
    }
}

public class Completed : OneOfCase 
{
    public DateTime CompletedAt { get; }

    public Running(DateTime completedAt) {
        CompletedAt = completedAt;
    }
}

3. Serialize!

    Status status = new Completed(DateTime.Now);
    Console.WriteLine(JsonConvert.SerializeObject(status));
    // output:
    // {\r\n    \"CompletedAt\": \"2020-11-17T21:29:52.0664389+01:00\",\r\n    \"Value\": \"Completed\"\r\n  }