CSC/ECE 517 Fall 2009/wiki2 18 cc

Introduction and Objective

Immutable classes are used to ensure the object state never gets changed. On the other hand the concept of serialization is that applications can send and receive objects across the network; this is a step further than just keeping the object alive (persistence), which is the ability to call objects remotely, however private instances may be exposed by performing serialization.

Now the problem arises when serializing an immutable class and the immutability must be preserved between applications. There is a nice description of this problem in the stackoverflow.com Forum

The Serialization Proxy Pattern is capable to handle this kind of situations; henceforth the main objectives are to explain:

1. It's idiomatic usage.

2. How it is designed and gives details of its implementation.

3. How the issue described above is handled.

When a Class is Immutable?

If a class state cannot be modified after being created then it is called immutable, on the contrary a mutable class can be modified after being created. There are also partially immutable classes, in which some of their attributes can be defined as immutable.

The key advantage is that they cannot be corrupted since the state cannot be changed.

To learn more about Immutable classes refer to the Java.sun.com Immutable Tutorial

The following is an example to illustrate immutable property:

 String  course_id=  "ECE517" ;
 course_id.toLowerCase();

In this case the course_id didn’t change to lower case, it remains as "ECE517" , although the output is “ece517" , instead the output is a new string that can be assigned to another variable, therefore and for the moment the values are still immutable, however the following code can change the value.

 course_id =  course_id.toLowerCase();

How to Make a Class Immutable?

Java Example

To make a class immutable we need to keep the class state safe by any means, in other words, assignments should never occur, the following code is an example of immutable class in JAVA

  final class Immutable_ece517Class 
  { 
     // To make the variables private use the term final
     final string nombre; 
     // Constructor where we can provide the constant value 
     public Immutable_ece517Class (string paramNombre) 
     { 
        nombre = paramNombre; 
     } 
     // Define the methods to return variable 
     public string getNombre
     { 
        return nombre; 
     } 
  }

The trick here is that final before the variable restricts to write to private variables, the method after that is only to read the variable value without writing it.

Ruby Example

To implement immutable class in Ruby, the programmer must use the Struct.immutable instead of using Struct.new, the following is the code example implementation.

  def Struct.immutable(*args)  # Factory method for immutable classes
    Struct.new(*args).class_eval do # Define struct, then modify it
      undef []=                     # Un-define general mutation
      args.each do |sym|            # For each field of the struct
        mutator = :"#{sym.to_s}="   # Symbol for setter method
        remove_method mutator       # Un-define that method
      end
      self                          # Return the class
    end
  end

It is recommended to visit Wiki/Immutable Code, in there you can find imutable implementations in other languages.

What is Serialization?

It is known as the process from a source application to convert an object into a sequence of bits and to store or sent them through the network, in this way the same sequence can be retrieved by another end application that reconstructs or clones the original object in order to use it, this capability to store entire objects instead of simple parameters make possible to have high level distributed applications, without serialization the object would persist only while it is being executed (transient) and the application interaction would not be possible, e.g. remote procedure calls. Figure 1 Shows the Serialization/De-serialization process.

        Figure 1. The process of Serialization/De-serialization 

There is an option when an object is serialized, the state can also recorded in the byte stream, when it is retrieved its last state is also recovered, this feature allow the possibility to save the object state in memory and retrieve for its use later by the same application, the details to implement this feature are discussed in Proxy Serialization section, although there is also an application named as persistent that does the same, serialization can be also used as a persistent procedure to keep alive classes in local memory storage after being executed.

It is also possible to serialize complex data structures like applets and complete GUI (Graphical User Interface), but the scope of this wiki will be on class serialization only.

To have an insight on Serialization we recommend to visit wiki/Serialization

How to Implement Serialization of Classes?

Java Serialization

The next step would be in how to implement serialization of classes. In this case Java.io package offers the “writeObject()” to serialize and for un-serialize it uses “readObject( )” both are methods from ObjectOuputStream and ObjectOuputStream respectively, for an object to be serialized must have implemented the java.io.Serializable interface, the reason behind this is security, some private classes should not be exposed when de-serialize them, this is why a class must explicitly declare as serialize.

Below is the serialization code implementation in JAVA: Part taken from wiki/Object_serialization

  import java.io.*;
  /* The object to serialize. */
  class ObjectToSerialize implements Serializable {
      static private final long serialVersionUID = 42L;
      private String firstAttribute;
      private int secondAttribute;
      public ObjectToSerialize(String firstAttribute, int secondAttribute) {
          this.firstAttribute = firstAttribute;
          this.secondAttribute = secondAttribute;
      }
      @Override
      public String toString() {
          return firstAttribute + ", " + secondAttribute;
      }
  }
   public static void main(String[] args) {
       ObjectToSerialize original = new ObjectToSerialize("some text", 123);
       System.out.println(original);
       try {
           saveObject(original, "object.ser");
           ObjectToSerialize loaded = (ObjectToSerialize) loadObject("object.ser");
           System.out.println(loaded);
       } catch (Exception e) {
           e.printStackTrace();
       }
   }
  }

Ruby Serialization

Ruby uses the methods of dump and load from the Marshal module to perform serialization, however not all objects can be serialized, the exeptions are Bindings, Procedures and Singleton objects,

Below is the code implementation in Ruby.

  class Klass
    def initialize(str)
      @str = str
    end
    def sayHello
      @str
    end
  end
  
  o = Klass.new("hello\n")
  data = Marshal.dump(o)
  obj = Marshal.load(data)
  obj.sayHello   »  "hello\n"

The Main Issue when De-serialize an Immutable Class

The weakness in serialization is that the details of the hidden private instances may be exposed in the other end application which is not good, however in order to make remote call the end application must understand the de-serialized objects, this is why serialization formats have been developed to keep proprietary software, one clear example is that a password and login should never be serialized, it is too risky if this information is exposed in the form of a serialized variable class.

The main issue is that hidden private instances (immutable classes) can be exposed when de-serialize in the other end application and the idea is that they must remain safe and unchanged across applications. The solution proposed is the use of the Proxy Serialization Pattern.

You can find an alternate description of the problem or main issue written very well in a blog by "Bob Carpenter" Alias

What is a Proxy Pattern?

The Proxy Pattern, which is one of the most used design patterns, defines a proxy object that interacts between source and sink applications within a network, one common use of it is the ability to call a remote tasks and to share information trough a network.

To to learn more on this topic please visit Wiki/Proxy Pattern

There are two most common types of proxies:

The first one is virtual proxy which creates the proxy object that interacts with the same application or other application within the same computer (stored in local memory), its purpose is to keep an object state alive after its execution so that it can be used later on.

The second one is remote proxy which creates the proxy object to be used in a different address space, commonly in a remote object on a distant host. It creates the illusion of interacting with the object or task directly, however there is a proxy between them that makes the interaction possible.

So, What is Proxy Serialization Pattern?

In general and summarized, the proxy serialization means that a class is serialized along with the method “writeReplace()” (Java), the result of this procedure is the so called “proxy serialization”, on the other end side the class is de-serialized along with the method “readResolve()” (Java) to reconstruct the class state (private properties).

The process to make a Proxy Serialization in more detail is listed below:

1. Before class serialization takes place, Java checks if the class contains the method “Object writeReplace()”.

2. If it does then the return value of “Object writeReplace()” is serialized together with the class (called Proxy Serialization), other wise only perform normal serialization on the class.

3. The proxy serialized value is used to keep the state of the original object to write (e.g. immutability), so that in the other end application the state can be recovered.

4. When de-serializing, Java checks if the class has the implementation of the method “Object readResolve()” if it does then it completes the de-serialization process and then the return value of the “readResolve()” is added to the de-serialization, which contains the state of the original object to write (e.g. immutability)

        Figure 2. The process of Proxy Serialization/De-serialization 

The key advantage of Proxy Serialization process is that proxy is implemented as a private static nested class, in figure 2 shows that its state is serialized (control of the read/write process) as well as the class itself, on the other end side the state is de-serialized (control of the read/write process) along with the class, in this way the class state is kept in the cloned class, this would keep the properties for example of an immutable implementation class.

Proxy Serialize Implementation in JAVA

The following code shows an example in Java; the first one is a typical immutable class, the second one is the serialization proxy implemented.

Typical Immutable Class

  public class Tot_price {
      final int mPrice;
      final int mTax;
      public int price() { return mPrice; }
      public int tax() { return mTax; }
      public int total() {
      return price() + tax();
  }

Typical Immutable with Class Proxy Serialization

  public class Tot_price implements Serializable {
      private final mPrice;
      private final mTax;
      public Tot_price(double price, double tax) { mPrice = price; mTax=tax; }
      ...
      private Object writeReplace() {
          return new SerializationProxy(this);
      }
      
      private static class SerializationProxy
          implements Serializable {
          int mPrice; int mTax;
          public SerializationProxy() { }
          public SerializationProxy(Tot_price cost) {
              mPrice = cost.price();  mTax = cost.tax();
          }
          Object readResolve() {
              return new Tot_price(mPrice,mTax);
          }
      }
  }

Proxy Serialization Identification

A random serial ID number is added whenever a class is Proxy Serialized by default. However, it is highly recommended to add it within the code in order to preserve forward compatibility, otherwise if the class changes the serialize process would be aborted, the following code shows how to add the ID number:

  public class Tot_price implements Serializable {
     ....
   static final long serialVersionUID = 200120012001L;
     ....
  }

Every time Java serializes an object, the ID number is computed through reflection, this causes to slow down the Serialization-De-serialization process, for this reason it is recommended to define in the code the ID number.

How is the Main Issue Solved?

Since the Proxy Pattern is a private class then its inheritance is also private, if there is an immutable class, then its immutable implementation is also private and its details remains hidden after de-serialization the proxy pattern.

In this way the immutable properties are kept even in the other end side application and the variables cannot be modified.

We recommend to read the following paper work Serializing Immutables and Singletons with a Serialization Proxy written by LingPipe which covers also Singletons, it is very well explained and with java examples to implement Serialization Proxy in Immutable classes.

References

http://stackoverflow.com/questions/702357/what-is-the-serialization-proxy-pattern

http://java.sun.com/javase/6/docs/technotes/guides/serialization/

http://my.safaribooksonline.com/9781598635655/ch10lev1sec4

http://my.safaribooksonline.com/0596006209/jenut3-CHP-10

http://my.safaribooksonline.com/0130844667/ch01lev1sec4

http://en.wikipedia.org/wiki/Object_serialization

http://en.wikipedia.org/wiki/Immutable_class

New

http://java.sun.com/docs/books/tutorial/essential/concurrency/immutable.html

http://java.sun.com/developer/technicalArticles/Programming/serialization/

http://en.wikipedia.org/wiki/Serialization

http://en.wikipedia.org/wiki/Proxy_pattern