Gson Type Adapter vs. Custom Deseralizer

Question

The example below shows a class (Club) that contains a collection of an abstract class (Member). I'm confused as to whether I need a TypeAdapter or JsonDeserializer to make the Deserialization work correctly. Serialization works just fine without any help, but Deserialization is throwing exceptions. To illustrate I've built the following "clone" test. If anyone could show a working example I would be very grateful.

First Club Class

package gson.test;
import java.util.ArrayList;

import com.google.gson.Gson;

public class Club {
    public static void main(String[] args) {
        // Setup a Club with 2 members
        Club myClub = new Club();
        myClub.addMember(new Silver());
        myClub.addMember(new Gold());

        // Serialize to JSON
        Gson gson = new Gson();
        String myJsonClub = gson.toJson(myClub); 
        System.out.println(myJsonClub);

        // De-Serialize to Club
        Club myNewClub = gson.fromJson(myJsonClub, Club.class);
        System.out.println(myClub.equals(myNewClub) ? "Cloned!" : "Failed");
    }

    private String title = "MyClub";
    private ArrayList<Member> members = new ArrayList<Member>();

    public boolean equals(Club that) {
        if (!this.title.equals(that.title)) return false;
        for (int i=0; i<this.members.size(); i++) {
            if (! this.getMember(i).equals(that.getMember(i))) return false;
        }
        return true;
    }
    public void addMember(Member newMember) { members.add(newMember); }
    public Member getMember(int i) { return members.get(i); }
}

Now the Abstract Base Class Member

package gson.test;
public abstract class Member {
    private int type;
    private String name = "";

    public int getType() { return type; }
    public void setType(int type) { this.type = type; }
    public boolean equals(Member that) {return this.name.equals(that.name);}
}

And two concrete sub-classes of Member (Gold and Silver)

package gson.test;
public class Gold extends Member {
    private String goldData = "SomeGoldData";
    public Gold() {
        super();
        this.setType(2);
    }
    public boolean equals(Gold that) {
        return (super.equals(that) && this.goldData.equals(that.goldData)); 
    }
}

package gson.test;
public class Silver extends Member {
    private String silverData = "SomeSilverData";
    public Silver() {
        super();
        this.setType(1);
    }
    public boolean equals(Silver that) { 
        return (super.equals(that) && this.silverData.equals(that.silverData)); 
    }
}

And finally the output

    {"title":"MyClub","members":[{"silverData":"SomeSilverData","type":1,"name":""},{"goldData":"SomeGoldData","type":2,"name":""}]}
    Exception in thread "main" java.lang.RuntimeException: Failed to invoke public gson.test.Member() with no args
        at com.google.gson.internal.ConstructorConstructor$3.construct(ConstructorConstructor.java:107)
        at com.google.gson.internal.bind.ReflectiveTypeAdapterFactory$Adapter.read(ReflectiveTypeAdapterFactory.java:186)
...

Answer 1

You can do both. Which one you pick depends really on potential performance impact, and how much code are willing to write.

Deserializers are more expensive. That is because the input to deserializer is a json tree, and GSon will have to create a full JsonElement subtree for the element that matches your class, before it can pass it to your deserializer. If your model has a lot of nesting, that cost increases. For plain objects, it will be negligible.

It seems that you will know which class to create based on the value of type property that will be included in target object. Your deserializer will need to

• look into the passed JsonElement object, read the type property, determine the type
• call context.deserialize() with the class and the same element that was passed to you
• throw an error if type was missing or invalid

Your type adapter will have to be more complex. The input to the type adapter is a stream, not an element/subtree. You can load the next value entirely from the stream, parse it, and then do exactly what deserializer did, which doesn't make sense and you can just use the deserializer instead. Alternatively, you can read the stream, see what properties there are, save them into local variables, until you get to the type property (you can't predict its location), then finish reading the remainder of the properties, and create your final Gold/Silver objects based on type, and all the properties read and saved.