Understand Liskov Substitution Principle (LSP)

Introduction

Here, I am going to discuss the Liskov’s Substitution Principle of S<code>OLID. Now what does SOLID mean? SOLID is an object oriented design principle, where each letter has it own meaning:

• S-> Single responsibility
• O-> Open Closed
• L-> Liskov substitution
• I-> Interface segregation
• D-> Dependency inversion

According to Wikipedia, the definition of SOLID is:

"SOLID are guidelines that can be applied while working on software to remove code smells by causing the programmer to refactor the software's source code until it is both legible and extensible."

Background

If you read my previous article, it will be very helpful for you to understand SOLID.

1. Understand Open Closed Principle and Dependecy Inversion .
2. Understand Single Responsibility and Interface Segregation

Using the Code

Before starting with the technical discussion, I want to answer the below questions:

What is Liskov substitution principle?

Answer: "objects in a program should be replaceable with instances of their subtypes without altering the correctness of that program"

Is my answer tough to understand? Ok, I will make it more easy...

"It means that we must make sure that new derived classes are extending the base classes without changing their behavior."

Let's consider an example to make it better understood. If you read articles related to the Liskov principle, one example is very popular. Yes that is Rectangular and Square. Here, I will show you the same but will try to explain each and every aspect clearly. Before starting with the technical discussion, look at the below picture and answer the questions.

           

Question: Is Square a Rectangle?
Answer: Yes

If you remember Inheritance, then the word "Is a" relationship familiar to you. So if I convert this to code, it looks like the following:

public class Rectangle
{
    protected int _width;
    protected int _height;
    public int Width
    {
        get { return _width; }
    }
    public int Height
    {
        get { return _height; }
    }

    public virtual void SetWidth(int width)
    {
        _width = width;
    }    
    public virtual void SetHeight(int height)
    {
        _height = height;
    }
    public int getArea()
    {
        return _width * _height;
    }
}  

public class Square : Rectangle  // In an "is a" relationship, the derived class is clearly a
                                  //kind of the base class
{
    public override void SetWidth(int width)
    {
        _width = width;
        _height = width;
    }

    public override void SetHeight(int height)
    {
        _height = height;
        _width = height;
    }
}

So everything is ok. Now we will calculate the area of Figure 1: Rectangle and Square.

public void AreaOfRectangle()
    {
        Rectangle r = RectangleFactory(); // Returns the rectangle type object
        r.SetWidth(7);
        r.SetHeight(3);
        r.getArea();
    }  

So can you tell me the output of the r.getArea(); method. Yes, very simple, the expected output is 7 * 3=21. Now run the program with the below RectangleFactory() method and see whether we get our expected result or not?

public  Rectangle RectangleFactory()
    {
        return new Square();
    }

One thing I want to mention about the RectangleFactory() is that, this method is now exposed to you. But think as if you are getting the Rectangle object just by using a factory DLL or from any service where you have no idea what type of rectangle object will be returned.

Did you see the result? Yes.

This is the output: 9.

So what's wrong there ? Remember as I said earlier:

"we must make sure that new derived classes are extending the base classes without changing their behavior"

If we say specifically:

"we must make sure that Square classes are extending the Rectangle without changing their behavior"

Is the above sentence correct according to our program?

No, the problem is:

without changing their behavior

But we change the behavior of the base Rectangle class. Did you notice how? No, OK, let's see:

public override void SetWidth(int width)
    {
        _width = width; 
        _height = width; //Change the behavior here by assigning the width to Rectangle _height
                         
    }

    public override void SetHeight(int height)
    {
        _height = height;
        _width = height;//Change the behavior here by assigning the height to Rectangle _width 
                        
    } 

Width=Height is must be rule for Square <code>not for Rectangle. So when Square object returns from Rectanglefactory() before calling getArea(), user assigns r.SetHeight(3) and r.SetWidth(7) and waits for a result 21 but gets 9. It changes the result. But the end user was expecting the result of Rectangle area.

So violation of Liskov substitution principle occurs.

Now the solution is to manage the class inheritance hierarchies correctly. Let's introduce another class.

public abstract class Quadrilaterals
{
    abstract public int GetArea();
}

Now change the inheritance hierarchies by making it base class.

public class Rectangle :Quadrilaterals
{    
    public int Width
    {
        get ; set ;
    }
    public int Height
    {
        get ; 
        set ;
    }
    public override int GetArea()
    {
        return Height * Width;
    }    
} 

public class Square : Quadrilaterals  // In an "is a" relationship, the derived class is clearly a
                                      // kind of the base class
{
    public  int Size
    {
        get ; 
        set ;
    }
    public override int GetArea()
    {
        return Size* Size;
    }
}  

Now our factory method also changes.

 public Quadrilaterals QuadrilateralsFactory()
    {
        return new Square();
    }
    public void AreaOfQuadrilateral()
    {
        Quadrilaterals r = QuadrilateralsFactory(); // Returns the Quadrilaterals type object
        r.Height=7;
        r.Width=3;
        r.getArea();
    } 

Points of Interest

So from users point of view, now s/he is expecting the area of <code>Quadrilateral not the Rectangle. By using this input, when output comes 21 is a Rectangle area and when 9 it's a Square .

 Quadrilaterals r = new Rectangle();
        r.Height=7;
        r.Width=3;
        r.getArea();

        Quadrilaterals r = new Square();
        r.Height = 7;
        r.Width = 3;
        r.getArea(); 

Now user gets the expected behavior from the program. So it satisfies the Liskov Substitute Principle (LSP).