For example, suppose I have some mobile apps that uses some user data, the data class:

class UserData{
  public:
  Address address;
  bool isVerified=false;
};

class Address{
};

The UserData may be loaded from the server JSON, so I need rapidJson, and also write some methods to assign the data from rapidJson to the object.

Instead of using active record pattern that writes the json-to-object methods (eg:fromJson()) to the object itself, I create some parallel classes that resembles the structure of UserData:

class UserDataSerializer{
public:
  static void fromJson(rapidJson::Value value,UserData& userData){
      AddressSerializer::fromJson(value["address"],userData.address);

      userData.isVerified=value["isVerified"].GetBool();
  }
};

class AddressSerializer{
  static void fromJson(rapidJson::Value value,Address& address){
  }
};

So that only the pages that loads UserData depend on the UserDataSerializer class, while the other pages that accesses user data only depend on UserData class. When UserData adds a new object property, I also need a new serializer.

My question is, is it a code smell?