Algorithms and OOD (CSC 207 2013F) : Readings

Separating What from How: Interfaces in Java


Summary: Good programming style suggests that you decide what methods a class provides before you decide how the class will provide those methods. In this reading, we consider Java's interfaces, which encourage such separation.

Prerequisites: Class basics.

Introduction: Separating the What from the How

As you should have discovered by now, a pair of key activities in object-oriented design are describing classes of objects and using objects from those classes. I have also recommended a particular strategy for designing classes: Start by deciding what your class should do and then figure out how to make the class do that.

Why do I (and most object-oriented designers) make the what/how distinction? First, because there's good evidence that programmers cannot think carefully about a wide variety of issues simultaneously. That is, if programmers have to think from the start about how they are going to implement a class (and its methods), they may not think as carefully about what methods it should provide. Second, because there's good evidence that programmers often decide upon the methods they will provide based on their choice of how they will implement the class. Third, because it's difficult to decide when rewriting a class what methods are necessary and what methods are tied to the implementation. Finally, because the what/how interface relates closely to the key object-oriented concept of encapsulation. That is, because the clients of a class should only know what the class can do, and not how the class does it, class designers should also carefully distinguish the two.

Interfaces in Java

Many programming languages accept this important need to distinguish between what a piece of code does and how it accomplishes that task, and incorporate in the language a mechanism for distinguishing these two key issues. In Java, the primary mechanism for distinguising what from how is something called an interface. Interfaces contain only method headers (a method declaration without the body) and, optionally, static fields.

Interface declarations look much like class declarations, except: (1) interfaces use the keyword interface, rather than the keyword class; (2) interfaces contain no constructors, object fields, or static methods; (3) method declarations in interfaces lack bodies.

Hence, an interface declaration looks much like the following:

public interface InterfaceName {
    // +-----------+---------------------------------------------------------
    // | Constants |
    // +-----------+

    // ...
  
    // +---------+-----------------------------------------------------------
    // | Methods |
    // +---------+

    // ...

} // interface InterfaceName

For example, here is a portion of an interface declaration for a point in two-space.

/**
 *  Vectors in two-space.
 *
 *  @author Samuel A. Rebelsky
 *  @version 1.1 of September 2013
 */
public interface Point2D {
    // +---------+-----------------------------------------------------------
    // | Methods |
    // +---------+

    /**
     * Get the X coordinate of the point.
     */
    public double getX();
  
    /**
     * Get the Y coordinate of the point.
     */
    public double getY();

    /**
     * Get the distance of this point from the origin.
     */
    public double getDistanceFromOrigin();

    /**
     * Translate this point by deltaX in the x direction and 
     * deltaY in the y direction.  Note that this method
     * mutates the underlying point.
     */
    public void translate(double deltaX, double deltaY);

    /**
     * Make a copy of this point.
     */
    public Point2D clone();
} // interface Point2D

Interfaces and Classes

What happens after you've defined an interface? It's time to build a class that corresponds to the interface. When we build that class, we tell Java that the class corresponds to the interface with the implements keyword. That is, we write

public class ClassName implements InterfaceName {
  // ...
} // class ClassName

For example, if we chose to implement Point2D as a class with a pair of fields, the x coordinate and the y coordinate, we might write

/**
 * An implementation of points in two space using an X coordinate
 * and a Y coordinate.
 *
 * @author Samuel A. Rebelsky
 *  @version 1.1 of September 2013
 */
public class Point2DPair implements Point2D {
    // +--------+------------------------------------------------------------
    // | Fields |
    // +--------+

    /**
     * The x coordinate of the point
     */
    double x;

    /**
     * The y coordinate of the point.
     */
    double y;

    // +--------------+------------------------------------------------------
    // | Constructors |
    // +--------------+

    /**
     * Create the point (x,y).
     */
    public Point2DPair(double x, double y) {
        this.x = x;
        this.y = y;
    } // Point2DPair(double,double)

    // +---------+-----------------------------------------------------------
    // | Methods |
    // +---------+

    /**
     * Get the X coordinate of the point.
     */
    public double getX() {
        return this.x;
    } // getX() 
  
    /**
     * Get the Y coordinate of the point.
     */
    public double getY() {
        return this.y;
    } // getY()

    /**
     * Get the distance of this point from the origin.
     */
    public double getDistanceFromOrigin() {
        return Math.sqrt(this.x*this.x + this.y*this.y);
    } // getDistanceFromOrigin()

    /**
     * Translate this point by deltaX in the x direction and
     * deltaY in the y direction.  Note that this method
     * mutates the underlying point.
     */
    public void translate(double deltaX, double deltaY) {
        this.x = this.x + deltaX;
        this.y = this.y + deltaY;
    } // translate(double, double)

    /**
     * Make a copy of this point.
     */
    public Point2D clone() {
        return new Point2DPair(this.x, this.y);
    } // clone()
} // class Point2DPair

Alternative Implementations

One particularly important aspect of Java's interface mechanism is that you, as programmer, are not limited to only one implementation per interface. That is, you can implement the same interface in multiple ways and then later choose with implementation is best for a particular task. For example, if we found that we were frequently checking the distance of points from the origin, we might instead choose to implement them as a radius/theta pair.

/**
 *  An implementation of points in two space using a radius/theta pair
 *  which specify the distance of the point from the origin and the
 *  angle between the positive x-axis and a line from the origin to
 *  the point.
 *
 *  @author Samuel A. Rebelsky
 *  @version 1.1 of September 2013
 */
public class Point2DPolar implements Point2D {
    // +--------+------------------------------------------------------------
    // | Fields |
    // +--------+
  
    /**
     * The radius, which represents the distance of the point from 
     * the origin.
     */
    double radius;
  
    /**
     * Theta, which represents the angle between the positive X axis
     * and a line from the origin to the point.
     */
    double theta;
  
    // +--------------+------------------------------------------------------
    // | Constructors |
    // +--------------+
  
    /**
     * Create the point (x,y)
     */
    public Point2DPolar(double x, double y) {
        this.radius = Math.sqrt(x*x + y*y);
        this.theta = Math.atan(y/x);
    } // Point2DPolar(double,double)
  
    /**
     * Create a point radius 0 and angle 0 (with the expectation that
     * we'll change it almost immediately).
     */
    private Point2DPolar() {
        this.radius = 0.0;
        this.theta = 0.0;
    } // Point2DPolar()
  
    // +---------+-----------------------------------------------------------
    // | Methods |
    // +---------+
  
    /**
     * Get the X coordinate of the point.
     */
    public double getX() {
        return this.radius * Math.cos(this.theta);
    } // getX() 
    
    /**
     * Get the Y coordinate of the point.
     */
    public double getY() {
        return this.radius * Math.sin(this.theta);
    } // getY()
  
    /**
     * Get the distance of this point from the origin.
     */
    public double getDistanceFromOrigin() 
    {
      return this.radius;
    } // getDistanceFromOrigin()
  
    /**
     * Translate this point by deltaX in the x direction and
     * deltaY in the y direction.  Note that this method
     * mutates the underlying point.
     */
    public void translate(double deltaX, double deltaY) {
        // Left as an exercise to the reader.
    } // translate(double, double)
  
    /**
     * Make a copy of this point.
     */
    public Point2D clone() {
        Point2DPolar tmp = new Point2DPolar();
        tmp.radius = this.radius;
        tmp.theta = this.theta;
        return tmp;
    } // clone()
  
} // class Point2DPolar

As suggested above, we can use either of these implementations and expect that they will behave the same (except, of course, in terms of efficiency). How do clients write code that accepts either implementation? The they write code that uses the interface, rather than a particular class. For example, we can write a method to compare points by distance from the origin as follows:

/**
 * Compare two points.  Return a negative number if p1 is closer to
 * the origin than p2, 0 if the two are the same distance from the
 * origin, and a positive number if p1 is further from the origin.
 */
public int compare(Point2D p1, Point2D p2) {
    return Double.compare(p1.distanceFromOrigin(), p2.distanceFromOrigin());
} // compare(Point2D, Point2D)

It is now perfectly acceptable to call this compare method with two Point2DPair objects, with two Point2DPolar objects, or even with one object of each class. We return to this idea in the reading on polymorphism.

Copyright (c) 2013 Samuel A. Rebelsky.

Creative Commons License

This work is licensed under a Creative Commons Attribution 3.0 Unported License. To view a copy of this license, visit http://creativecommons.org/licenses/by/3.0/ or send a letter to Creative Commons, 543 Howard Street, 5th Floor, San Francisco, California, 94105, USA.