Saturday, August 19, 2006

Swich Statement code smell and Polymorphism

One of the symptoms of object-oriented programming is the lack of switch or case statements. Imagine that we have some client class that calculates the area and perimeter of particular geometrical shapes.

public class Client {
private double a;
private double b;
private double r;
...
public double calculateArea(int shape) {
double area = 0;
switch(shape) {
case SQUARE:
area = a * a;
break;
case RECTANGLE:
area = a * b;
break;
case CIRCLE:
area = Math.PI * r * r;
break;
}
return area;
}

public double calculatePerimeter(int shape) {
double perimeter = 0;
switch(shape) {
case SQUARE:
perimeter = 4 * a;
break;
case RECTANGLE:
perimeter = 2 * (a + b);
break;
case CIRCLE:
perimeter = 2 * Math.PI * r;
break;
}
return perimeter;
}
...
}

The previous code is clearly a poor design that limits the current client to only work with three types of shapes. The problem with switch statements is the duplication and that is a code smell. There are usually several places in the code where the behaviour slightly deviates and these switch statements are present (e.g. in calculateArea and calculatePerimeter methods). Even worse case is if we work with objects where the switch is replaced by multiple instanceof conditions.


public class Client {
...
public double calculateArea(Object shape) {
double area = 0;
if (shape instanceof Square) {
Square square = (Square) shape;
area = square.getA() * square.getA();
}
else if (shape instanceof Rectangle) {
Rectangle rectangle = (Rectangle) shape;
area = rectangle.getA() * rectangle.getB();
}
else if (shape instanceof Circle) {
Circle circle = (Circle) shape;
area = Math.PI * cirle.getR() * cirle.getR();
}
return area;
}

public double calculatePerimeter(Object shape) {
double perimeter = 0;
if (shape instanceof Square) {
Square square = (Square) shape;
perimeter = 4 * square.getA();
}
else if (shape instanceof Rectangle) {
Rectangle rectangle = (Rectangle) shape;
perimeter = 2 * (rectangle.getA() + rectangle.getB());
}
else if (shape instanceof Circle) {
Circle circle = (Circle) shape;
perimeter = 2 * Math.PI * cirle.getR();
}
return perimeter;
}
}

To improve the desing we make Square, Rectangle and Circle have a commont root. By that I mean that they either extend the same class, such as AbstractShape or that they implement a common interface, such as Shape or ideally both.

Then we can eliminate the switch statement code smell very easily. We replace it with polymorphism.

Shape.java file
public interface Shape {
public double getArea();
public double getPerimeter();
}

Square.java file
public class Square implements Shape {
private double a;
...
public double getArea() {
return a * a;
}
public double getPerimeter() {
return 4 * a;
}
}

Rectangle.java file
public class Rectangle implements Shape {
private double a;
private double b;
...
public double getArea() {
return a * b;
}
public double getPerimeter() {
return 2 * (a + b);
}
}

Circle.java file
public class Circle implements Shape {
private double r;
...
public double getArea() {
return Math.PI * r * r;
}
public double getPerimeter() {
return 2 * Math.PI * r;
}
}

And such refactoring simplifies the Client code. It also allows for easy extensibility by implementations of other shapes without changing a single line of Client code.

public class Client {
private Shape shape;
...
public double calculateArea() {
return shape.getArea();
}
public double calculatePerimeter() {
return shape.getPerimeter();
}
}

Another way of looking at it is from the responsibility point of view. Why should the client be responsible for calculating the area or the perimeter; or have a knowledge about shape's internals (e.g. number of sides, radius, etc.) It is the responsibility of each shape and all that the client needs to know is that a shape has a perimeter and area.

To sum this all up:

6 comments:

Anonymous said...

Obviously a programming genius.
if you implement an interface , each shape has to provide functions of the interface.

so how many sides does a circle have?
and whats the radius of a square?

Unknown said...

Dude, inheritance is good if your objects have something in common (e.g. all shapes have an area).

Obviously, inheritance is not suitable for everything, it isn't a silver bullet to code re-use.

If you are interested in specifics of each implementation or your objects do not have much in common you'll need to use something else.

Rob Meyer said...

I think the point here is that a switch statement (pretty much) always represents some conceptual commonality. For anonymous, that means that you don't put GetNumberOfSides() or GetRadius() into the common interface. You also wouldn't write a method like this:

int GetRadius(shape)
switch(shape.Type)
Type.Circle:
// calculate radius
Type.Square:
// What happens here?

You can implement as many interfaces as you want, so if something represents a first class concept in your conceptual model, no reason not to elevate it to an interface. Dushan is right of course that it's likely unmanageable to create a single interface for every switch statement, but finding the ones that make sense for your application to avoid switch or big if/then blocks is why design isn't easy.

Anonymous said...

Hey,

Thanks for this - Someone told me they were asked this at interview so it's a good one to know!

Unknown said...

So how do you fix this?


if (obj instanceof String) {
// it's a String
} else if (obj instanceof Integer) {
// it's an Integer
}

ADTC said...

Mark,

If you're having to do that, it's BAD design.

Your method should accept a String or an int/Integer, NOT an arbitrary object which can be anything. Create two methods, one for String, one for int/Integer if necessary. Don't create one method for Object - that's horrible design.


Creative Commons License This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 2.5 License.