GSON: how to prevent StackOverflowError while keeping circular references?

Question

I have a structure with circular references. And for debug purposes, I want to dump it. Basically as any format, but I chose JSON.

Since it can be any class, I chose

Answer 1

Simply make the fields transient (as in private transient int field = 4;). GSON understands that.

Edit

No need for a built-in annotation; Gson lets you plug in your own strategies for excluding fields and classes. They cannot be based on a path or nesting level, but annotations and names are fine.

If I wanted to skip fields that are named "lastName" on class "my.model.Person", I could write an exclusion strategy like this:

class MyExclusionStrategy implements ExclusionStrategy {

    public boolean shouldSkipField(FieldAttributes fa) {                
        String className = fa.getDeclaringClass().getName();
        String fieldName = fa.getName();
        return 
            className.equals("my.model.Person")
                && fieldName.equals("lastName");
    }

    @Override
    public boolean shouldSkipClass(Class<?> type) {
        // never skips any class
        return false;
    }
}

I could also make my own annotation:

@Retention(RetentionPolicy.RUNTIME)
public @interface GsonRepellent {

}

And rewrite the shouldSkipField method as:

public boolean shouldSkipField(FieldAttributes fa) {
    return fa.getAnnotation(GsonRepellent.class) != null;
}

This would enable me to do things like:

public class Person {
    @GsonRepellent
    private String lastName = "Troscianko";

    // ...

To use a custom ExclusionStrategy, build Gson object using the builder:

Gson g = new GsonBuilder()
       .setExclusionStrategies(new MyOwnExclusionStrategy())
       .create();

Answer 2

I know this question has a few years now, but I'd like to contribute with my solution. Although @fdreger's answer is completely valid in case you want to exclude a field always, it doesn't work if you want to exclude it just in certain cases, avoiding this way the recursion. The way I approached the problem is:

1. I write my own JsonSerializer. In it, I define a static variable to control de number of times an object of this same class is serialize and depending on the value, the object can be serialized or not.

    import com.fasterxml.jackson.core.JsonGenerator;
    import com.fasterxml.jackson.core.JsonProcessingException;
    import com.fasterxml.jackson.databind.JsonSerializer;
    import com.fasterxml.jackson.databind.SerializerProvider;
    import java.io.IOException;
   
    public class UserJsonSerializer extends JsonSerializer<User> {
   
        private static final ThreadLocal<Integer> depth = new ThreadLocal<Integer>() {
           @Override
           protected Integer initialValue() {
               return 0;
           }
        };
   
        @Override
        public void serialize(User user, JsonGenerator generator, SerializerProvider provider) throws IOException, JsonProcessingException {
           // Here, we limit the number of instances to return. In this case, just 1.
   
           depth.set(depth.get() + 1);
           if(depth.get() >= 1) {
               generator.writeNull();
           } else {
               generator.writeObject(user);
           }
        }
   
        public static void clear() {
           depth.remove();
        }
   }
   

2. Bind the UserJsonSerializer to the class you want to control

   public class SomeClass implements Serializable {
       @JsonSerialize(using = UserJsonSerializer.class)
       private User user;
   
      //...others fields...
   }
   

3. Don't forget to call UserJsonSerializer#clear() method to reinitialize the counter everytime you're going to parse a new entity.

I hope this helps.