Java and Object-Oriented Programming

Many seasoned Java developers will scoff at the fact that this section even exists in this book. It is here for two very important reasons. The first is that I continually run across Java applications built with a procedural mind-set. The fact that you know Java doesn't mean that you have the ability to transform that knowledge into well-designed object-oriented systems. As both an instructor and consultant, I see many data-processing shops send COBOL and/or Visual Basic developers to a three-day class on UML and a five-day class on Java and expect miracles. Case in point: I was recently asked to review a Java application to assess its design architecture and found that it had only two classes—SystemController and ScreenController—which contained over 70,000 lines of Java code.

The second reason for the emphasis on how the language maps to object-oriented principles is that people like language comparisons and how they stack up to their counterparts. To appease those that live and die by language comparisons, let's put Java under the scrutiny of what constitutes an object-oriented language.

No definitive definition of what makes a language object-oriented is globally accepted. However, a common set of criteria I personally find useful is that the language must support the following:

  • Classes

  • Complex types (Java reference types)

  • Message passing

  • Encapsulation

  • Inheritance

  • Polymorphism

These are discussed in the next subsections.

Java and Classes

Java allows classes to be defined. There are no stray functions floating around in Java. A class is a static template that contains the defined structure (attributes) and behavior (operations) of a real-world entity in the application domain. At runtime, the class is instantiated, or brought to life, as an object born in the image of that class. In my seminars, when several folks new to the object world are in attendance, I often use the analogy of a cookie cutter. The cookie cutter is merely the template used to stamp out what will become individually decorated and unique cookies. The cookie cutter is the class; the unique blue, green, and yellow gingerbread man is the object (which I trust supports a bite operation).

Java exposes the class to potential outside users through its public interface. A public interface consists of the signatures of the public operations supported by the class. A signature is the operation name and its input parameter types (the return type, if any, is not part of the operation's signature).

Good programming practice encourages developers to declare all attributes as private and allow access to them only via operations. As with most other languages, however, this is not enforced in Java. Figure 2-1 outlines the concept of a class and its interface.

Figure 2-1. Public interface of a class


The figure uses a common eggshell metaphor to describe the concept of the class's interface, as well as encapsulation. The internal details of the class are hidden from the outside via a well-defined interface. In this case, only four operations are exposed in the classes interface (Operation_A, B, C, and D). The other attributes and operations are protected from the outside world. Actually, to the outside world, it's as if they don't even exist.

Suppose you want to create an Order class in Java that has three attributes—orderNumber, orderDate, and orderTotal—and two operations—calcTotalValue() and getInfo(). The class definition could look like this:

/**
 * Listing 1
 * This is the Order class for the Java/UML book
 */
package com.jacksonreed;
import java.util.*;

public class Order
{
  private Date orderDate;
  private long orderNumber;
  private long orderTotal;

  public Order()
  {
  }

  public boolean getInfo()
  {
    return true;
  }

  public long calcTotalValue()
  {
    return 0;
  }

  public Date getOrderDate()
  {
    return orderDate;
  }

  public void setOrderDate(Date aOrderDate)
  {
    orderDate = aOrderDate;
  }

  public long getOrderNumber()
  {
    return orderNumber;
  }
  public void setOrderNumber(long aOrderNumber)
  {
    orderNumber = aOrderNumber;
  }

  public long getOrderTotal()
  {
    return orderTotal;
  }

  public void setOrderTotal(long aOrderTotal)
  {
    orderTotal = aOrderTotal;
  }

  public static void main(String[] args)
  {
    Order order = new Order();
    System.out.println("instantiated Order");
    System.out.println(order.getClass().getName());
    System.out.println(order.calcTotalValue());

    try {
      Thread.currentThread().sleep(5*1000);
      }  catch(InterruptedException e) { }
  }
}

A few things are notable about the first bit of Java code presented in this book. Notice that each of the three attributes has a get and a set operation to allow for the retrieval and setting of the Order object's properties. Although doing so is not required, it is common practice to provide these accessor-type operations for all attributes defined in a class. In addition, if the Order class ever wanted to be a JavaBean, it would have to have “getters and setters” defined in this way.

Some of the method code in the main() operation does a few things of note. Of interest is that a try block exists at the end of the operation that puts the current thread to sleep for a bit. This is to allow the console display to freeze so that you can see the results.

If you type in this class and then compile it and execute it in your favorite development tool or from the command prompt with

javac order.java //* to compile it
java order //* to run it

you should get results that look like this:

instantiated Order
com.jacksonreed.Order
0

Note

Going forward, I promise you will see no code samples with class, operation, or attribute names of foo, bar, or foobar.


More on Java and Classes

A class can also have what are called class-level operations and attributes. Java supports these with the static keyword. This keyword would go right after the visibility (public, private, protected) component of the operation or attribute. Static operations and attributes are needed to invoke either a service of the class before any real instances of that class are instantiated or a service that doesn't directly apply to any of the instances. The classic example of a static operation is the Java constructor. The constructor is what is called when an object is created with the New keyword. Perhaps a more business-focused example is an operation that retrieves a list of Customer instances based on particular search criteria.

A class-level attribute can be used to store information that all instances of that class may access. This attribute might be, for example, a count of the number of objects currently instantiated or a property about Customer that all instances might need to reference.

Java and Complex Types (Java Reference Types)

A complex type, which in Java is called a reference type, allows variables typed as something other than primitive types (e.g., int and boolean) to be declared. In Java, these are called reference types. In object-oriented systems, variables that are “of” a particular class, such as Order, Customer, or Invoice, must be defined. Taken a step further, Order could consist of other class instances, such as OrderHeader and OrderLine.

In Java, you can define different variables that are references to runtime objects of a particular class type:

Public Order myOrder;
Public Customer myCustomer;
Public Invoice myInvoice;

Such variables can then be used to store actual object instances and subsequently to serve as recipients of messages sent by other objects. In the previous code fragment, the variable myOrder is an instance of Order. After the myOrder object is created, a message can be sent to it and myOrder will respond, provided that the operation is supported by myOrder's interface.

Java and Message Passing

Central to any object-oriented language is the ability to pass messages between objects. In later chapters you will see that work is done in a system only by objects that collaborate (by sending messages) to accomplish a goal (which is specified in a use-case) of the system.

Java doesn't allow stray functions floating around that are not attached to a class. In fact, Java demands this. Unfortunately, as my previous story suggested, just saying that a language requires everything to be packaged in classes doesn't mean that the class design will be robust, let alone correct.

Java supports message passing, which is central to the use of Java's object-oriented features. The format closely resembles the syntax of other languages, such as C++ and Visual Basic. In the following code fragment, assume that a variable called myCustomer, of type Customer, is defined and that an operation called calcTotalValue() is defined for Customer. Then the calcTotalValue() message being sent to the myCustomer object in Java would look like this:

myCustomer.calcTotalValue();

Many developers feel that, in any other structured language, this is just a fancy way of calling a procedure. Calling a procedure and sending a message are similar in that, once invoked, both a procedure and a message implement a set of well-defined steps. However, a message differs in two ways:

  1. There is a designated receiver, the object. Procedures have no designated receiver.

  2. The interpretation of the message—that is, the how-to code (called the method) used to respond to the message—can vary with different receivers. This point will become more important later in the chapter, when polymorphism is reviewed.

The concepts presented in this book rely heavily on classes and the messaging that takes place between their instances, or objects.

Java and Encapsulation

Recall that a class exposes itself to the outside world via its public interface and that this should be done through exposure to operations only, and not attributes. Java supports encapsulation via its ability to declare both attributes and operations as public, private, or protected. In UML this is called visibility.

Using the code from the previous Order example, suppose you want to set the value of the orderDate attribute. In this case, you should do so with an operation. An operation that gets or sets values is usually called a getter or a setter, respectively, and collectively such operations are called accessors. The local copy of the order date, orderDate, is declared private. (Actually, all attributes of a class should be declared private or protected, so that they are accessible only via operations exposed as public to the outside world.)

Encapsulation provides some powerful capabilities. To the outside world, the design can hide how it derives its attribute values. If the orderTotal attribute is stored in the Order object, the corresponding get operation defined previously looks like this:

public long getOrderTotal()
  {
    return orderTotal;
  }

This snippet of code would be invoked if the following code were executed by an interested client:

private long localTotal;
private Order localOrder;
localOrder = New Order();
localTotal = localOrder.getOrderTotal()

However, suppose the attribute orderTotal isn't kept as a local value of the Order class, but rather is derived via another mechanism (perhaps messaging to its OrderLine objects). If Order contains OrderLine objects (declared as a Vector or ArrayList of OrderLine objects called myOrderLines) and OrderLine knows how to obtain its line totals via the message getOrderLineTotal(), then the corresponding get operation for orderTotal within Order will look like this:

public long getOrderTotal()
  {
    long totalAmount=0;

    for (int i=0; i < myOrderLines.length; i++)
    {
      totalAmount = totalAmount +
              myOrderLines[i].getOrderLineTotal();
    }
    return totalAmount;
  }

This code cycles through the myOrderLines collection, which contains all the Orderline objects related to the Order object, sending the getOrderLineTotal() message to each of Order's OrderLine objects. The getOrderTotal() operation will be invoked if the following code is executed by an interested client:

long localTotal;
Order myOrder;
myOrder = new Order();
localTotal = localOrder.getOrderTotal()

Notice that the “client” code didn't change. To the outside world, the class still has an orderTotal attribute. However, you have hidden, or encapsulated, just how the value was obtained. This encapsulation allows the class's interface to remain the same (hey, I have an orderTotal that you can ask me about), while the class retains the flexibility to change its implementation in the future (sorry, how we do business has changed and now we must derive orderTotal like this). This kind of resiliency is one of the compelling business reasons to use an object-oriented programming language in general.

Java and Inheritance

The inclusion of inheritance is often the most cited reason for granting a language object-oriented status. There are two kinds of inheritance: interface and implementation. As we shall see, Java is one of the few languages that makes a clear distinction between the two.

Interface inheritance (Figure 2-2) declares that a class that is inheriting an interface will be responsible for implementing all of the method code of each operation defined in that interface. Only the signatures of the interface are inherited; there is no method or how-to code.

Figure 2-2. Interface inheritance


Implementation inheritance (Figure 2-3) declares that a class that is inheriting an interface may, at its option, use the method code implementation already established for the interface. Alternatively, it may choose to implement its own version of the interface. In addition, the class inheriting the interface may extend that interface by adding its own operations and attributes.

Figure 2-3. Implementation inheritance


Each type of inheritance should be scrutinized and used in the appropriate setting. Interface inheritance is best used under the following conditions:

  • The base class presents a generic facility, such as a table lookup, or a derivation of system-specific information, such as operating-system semantics or unique algorithms.

  • The number of operations is small.

  • The base class has few, if any, attributes.

  • Classes realizing or implementing the interface are diverse, with little or no common code.

Implementation inheritance is best used under the following conditions:

  • The class in question is a domain class that is of primary interest to the application (i.e., not a utility or controller class).

  • The implementation is complex, with a large number of operations.

  • Many attributes and operations are common across specialized implementations of the base class.

Some practitioners contend that implementation inheritance leads to a symptom called the fragile base class problem. Chiefly, this term refers to the fact that over time, what were once common code and attributes in the superclass may not stay common as the business evolves. The result is that many, if not all, of the subclasses, override the behavior of the superclass. Worse yet, the subclasses may find themselves overriding the superclass, doing their own work, and then invoking the same operation again on the superclass. These practitioners espouse the idea of using only interface inheritance. Particularly with the advent of Java and its raising of the interface to a first-class type, the concept and usage of interface-based programming have gained tremendous momentum.

As this book evolves, keeping in mind the pointers mentioned here when deciding between the two types of inheritance will be helpful. Examples of both constructs will be presented in the theme project that extends throughout this book.

Implementation Inheritance

Java supports implementation inheritance with the extends keyword. A class wanting to take advantage of implementation inheritance simply adds an extendsClassName statement to its class definition. To continue the previous example, suppose you have two different types of orders, both warranting their own subclasses: Commercial and Retail. You would still have an Order class (which isn't instantiated directly and which is called abstract). The previous fragment showed the code for the Order class. Following is the code for the Commercial class.

package com.jacksonreed;
public class Commercial extends Order
{
  public Commercial()
  {
  }

  /* Unique Commercial code goes here */
}

Implementation inheritance allows the Commercial class to utilize all attributes and operations defined in Order. This will be done automatically by the Java Virtual Machine (JVM) in conjunction with the language environment. In addition, implementation inheritance has the ability to override and/or extend any of Order's behavior. Commercial may also add completely new behavior if it so chooses.

Interface Inheritance

Java supports interface inheritance with the implements keyword. A class wanting to realize a given interface (actually being responsible for the method code) simply adds an implements InterfaceName statement. However, unlike extension of one class by another class, implementation of an interface by a class requires that the interface be specifically defined as an interface beforehand.

Looking again at the previous example with Order, let's assume that this system will contain many classes—some built in this release, and some built in future releases—that need the ability to price themselves. Remember from earlier in this chapter that one of the indicators of using interface inheritance is the situation in which there is little or no common code but the functional intent of the classes is the same. This pricing functionality includes three services: the abilities to calculate tax, to calculate an extended price, and to calculate a total price. Let's call the operations for these services calcExtendedPrice(), calcTax(), and calcTotalPrice(), respectively, and assign them to a Java interface called IPrice. Sometimes interface names are prefixed with the letter I to distinguish them from other classes:

package com.jacksonreed;

interface IPrice
{
  long calcExtendedPrice();
  long calcTax();
  long calcTotalPrice();
}

Notice that the interface contains only operation signatures; it has no implementation code. It is up to other classes to implement the actual behavior of the operations. For the Order class to implement, or realize, the IPrice interface, it must include the implements keyword followed by the interface name:

public class Order implements IPrice
{
}

If you try to implement an interface without providing implementations for all of its operations, your class will not compile. Even if you don't want to implement any method code for some of the operations, you still must have the operations defined in your class.

One very powerful aspect of interface inheritance is that a class can implement many interfaces at the same time. For example, Order could implement the IPrice interface and perhaps a search interface called ISearch. However, a Java class may extend from only one other class.

Java and Polymorphism

Polymorphism is one of those $50 words that dazzles the uninformed and sounds really impressive. In fact, polymorphism is one of the most powerful features of any object-oriented language.

Roget's II: The New Thesaurus cross-references the term polymorphism to the main entry of variety. That will do for starters. Variety is the key to polymorphism. The Latin root for polymorphism means simply “many forms.” Polymorphism applies to operations in the object-oriented context. So by combining these two thoughts, you could say that operations are polymorphic if they are identical (not just in name but also in signatures) but offer variety in their implementations.

Polymorphism is the ability of two different classes each to have an operation that has the same signature, while having two very different forms of method code for the operation. Note that to take advantage of polymorphism, either an interface inheritance or an implementation inheritance relationship must be involved.

In languages such as COBOL and FORTRAN, defining a routine to have the same name as another routine will cause a compile error. In object-oriented languages such as Java and C++, several classes might have an operation with the same signature. Such duplication is in fact encouraged because of the power and flexibility it brings to the design.

As mentioned previously, the implements and extends keywords let the application take advantage of polymorphism. As we shall see, the sample project presented later in this book is an order system for a company called Remulak Productions. Remulak sells musical equipment, as well as other types of products. There will be a Product class, as well as Guitar, SheetMusic, and Supplies classes.

Suppose, then, that differences exist in the fundamental algorithms used to determine the best time to reorder each type of product (called the economic order quantity, or EOQ). I don't want to let too much out of the bag at this point, but there will be an implementation inheritance relationship created with Product as the ancestor class (or superclass) and the other three classes as its descendants (or subclasses). The scenario that follows uses implementation inheritance with a polymorphic example. Note that interface inheritance would yield the same benefits and be implemented in the same fashion.

To facilitate extensibility and be able to add new products in the future in a sort of plug-and-play fashion, we can make calcEOQ() polymorphic. To do this in Java, Product would define calcEOQ() as abstract, thereby informing any inheriting subclass that it must provide the implementation. A key concept behind polymorphism is this: A class implementing an interface or inheriting from an ancestor class can be treated as an instance of that ancestor class. In the case of a Java interface, the interface itself is a valid type.

For example, assume that a collection of Product objects is defined as a property of the Inventory class. Inventory will support an operation, getAverageEOQ(), that needs to calculate the average economic order quantity for all products the company sells. To do this requires that we iterate over the collection of Product objects called myProducts to get each object's unique economic order quantity individually, with the goal of getting an average:

public long getAverageEOQ()
  {
    long totalAmount=0;

    for (int i=0; i < myProducts.length; i++)
    {
      totalAmount = totalAmount + myProducts[i].calcEOQ();
    }
    return totalAmount / myProducts.length;
  }

But wait! First of all, how can Inventory have a collection of Product objects when the Product class is abstract (no instances were ever created on their own)? Remember the maxim from earlier: Any class implementing an interface or extending from an ancestor class can be treated as an instance of that interface or extended class. A Guitar “is a” Product, SheetMusic “is a” Product, and Supplies “is a” Product. So anywhere you reference Guitar, SheetMusic, or Supplies, you can substitute Product.

Resident in the array myProducts within the Inventory class are individual concrete Guitar, SheetMusic, and Supplies objects. Java figures out dynamically which object should get its own unique calcEOQ() message. The beauty of this construct is that later, if you add a new type of Product—say, Organ—it will be totally transparent to the Inventory class. That class will still have a collection of Product types, but it will have four different ones instead of three, each of which will have its own unique implementation of the calcEOQ() operation.

This is polymorphism at its best. At runtime, the class related to the object in question will be identified and the correct “variety” of the operation will be invoked. Polymorphism provides powerful extensibility features to the application by letting future unknown classes implement a predictable and well-conceived interface without affecting how other classes deal with that interface.

..................Content has been hidden....................

You can't read the all page of ebook, please click here login for view all page.
Reset
3.21.46.78