The current behavior might be intended, but it seems that the JSON utility class does not properly handle ObjectIds when serializing and parsing. I would expect that creating a DBObject (and saving to generate an ObjectId), serializing it, and then parsing would result in a new DBObject instance with the same _id as the original. That is, I would expect that parse(serialize(o)).id to equal o.id. This is not the case.

The following code illustrates the problem:

DBCollection c = getMeACollection();

// clear the collection
c.remove(new BasicDBObject());

DBObject o = new BasicDBObject();
c.save(o);
String json = JSON.serialize(o);
o = (DBObject)JSON.parse(json);
c.save(o);

results in two documents in the collection (I would expect one):

{ "_id" : ObjectId("4bd62964037394994a295cd3") } { "_id" : "4bd62964037394994a295cd3" }

• • Interestingly, the ObjectId class makes an attempt at dealing with this (see ObjectId.equals and ObjectId.massageToObjectId), but it does not seem to be handled thoroughly enough to prevent the type of behavior I illustrate in my example.