The issue of serializing/transforming model objects is not new, heck I’ve been doing this for quite some time:

• RMI (ejb/corba)
• XML (jms, soap, etc..)
• JSON 

• annotate your attributes with @MessageAttr
• have your class implement IModelObject

Here is a simple example:

<public class User implements IModelObject, Serializable {
private static final long serialVersionUID = 1L;
@MessageAttr
public String name;

public void setUsername(String username) {
this.username = username;
}
public void setPassword(String password) {
this.password = password;
}
...
}

Unit test it

Now to test the rendering of our object to JSON with a simple junit test…. As you can see, we leverage the Appcelerator framework to serialize our objects to JSON

public class ForumTest extends TestCase {
 
public void testSimple() {
User user = createUser();
Message message = new Message();
JSONMessageDataObject data = new JSONMessageDataObject();
message.setData(data);
message.getData().put("user", user);
message.getData().put("count", 1);
String messagestr = data.toDataString();
assertEquals(messagestr,"\"user\":{\"password\":
\"pwd\",\"threads\":0,\"fullName\"
:\"antewew\",\"username\":\"azuercher\",\"state\":
\"mystate\",\"email\":\"email\",
\"posts\":0,\"id\":0},\"count\":1}");
}
private User createUser() {
User user = new User();
user.setEmail("email");
user.setFullName("ante wew");
user.setId(new Long(0));
user.setPassword("pwd");
user.setPosts(new Long(0));
user.setState("mystate");
user.setThreads(new Long(0));
user.setUsername("azuercher");
return user;
}
}

Dealing with recursion

Whats always a bit of a tangle is understanding how to deal with the recursive/circular relationship.
If you take a look at JSONObject you will see 2 overridden methods for createBean

public static JSONObject createBean(IModelObject object,MessageAttr
    parentAtt, String context,String[] parentSuppres,int level, int maxlevels)
public static JSONObject createBean(IModelObject object)

The latter is obviously a bit more simple, but the former is where the power is. In the MessageAttr annotation, you can provide the suppress attribute which is a comma separated list of aggregates (using bean.name notation to not serialize). This is used for attributes in your IModelObject implementation where the association is with another IModelObject. The following model is for a forum object model where the following aggregate hierarchy exists:

* Forum
** ForumThread
*** Post

In the snippet below, I’ve omitted the getter/setter methods for the aggregates for simplicity.

public class User implements IModelObject, Serializable {
@MessageAttr (suppress="user,thread.lastPost")
public Post lastPost;
}
public class Post implements IModelObject, Serializable {
@MessageAttr(suppress="lastPost,forum.lastPost")
public Forumthread thread;
 
@MessageAttr (suppress="lastPost")
public User user;
}
public class Forumthread implements IModelObject, Serializable {
@MessageAttr (suppress="lastPost")
public Forum forum;
@MessageAttr (suppress="thread,user.lastPost")
public Post lastPost;
}
public class Forum implements IModelObject, Serializable {
@MessageAttr (suppress="thread.forum,thread.lastPost,user.lastPost")
public Post lastPost;
}

Rolling your own serialization

Assuming you know what your JSON string is going to look like, you can use our RawMessageDataList and RawMessageDataObject to serialze your objects. This is pretty useful if you already are rendering JSON in an existing framework and don’t want to have to transform to and back again. The snippet below shows with static strings just so that you get the idea:

IMessageDataList people = new
RawMessageDataList(
"[{'name':'joe','age':22},{'name':'jane','age':33}]");
IMessageDataList dog = new RawMessageDataObject("{'breed':'doberman','weight':78}");
Message message = new Message();
message.setData(new JSONMessageDataObject());
message.getData().put("people", people);
message.getData().put("dog", dog);

Using Coarse Grained objects

This is probably the simplest/prettiest way to implement your services assuming that you aren’t interested in using fine grained classes that are associated with most of today’s persistence frameworks. Here is how you would create a single compound JSON object:

IMessageDataObject dog =
MessageUtils.createMessageDataObject();
obj.put("breed","doberman");
obj.put("wieght",78);Message message = new Message();
message.setData(new JSONMessageDataObject());
message.getData().put("dog", dog);

and now with a collection:

IMessageDataList&lt;IMessageDataObject&gt; people=
MessageUtils.createMessageDataObjectList();
IMessageDataObject joe =
MessageUtils.createMessageDataObject();
joe.put("name","joe");
joe.put("age",22);
IMessageDataObject jane =
MessageUtils.createMessageDataObject();
joe.put("name","jane");
joe.put("age",33);
people.put(joe);
people.put(jane);
Message message = new Message();
message.setData(new JSONMessageDataObject());
message.getData().put("people", people);

Summary

As you can see there are quite a bit of alternatives for you based on what your needs are to accommodate your service implementations. I’ve personally used all of the above as I’ve implemented:

• Model Objects: using Hibernate for persistence
• Custom Serialization: for integrating with pre-rendered objects (commons-monitoring)
• Coarse Grained: in implementing a dashboard/event driven solution