Mukesh Yadav
  22 Nov 17

Interface Segregation Principle

Previous Principle: Liskov’s Substitution Principle

A client should never be forced to implement an interface that it doesn’t use or clients shouldn’t be forced to depend on methods they do not use.

In OOD we provide abstraction to module using interfaces, so we create an interface for a module and an implementation class. But now if we have to create an another module which only have some of the functionality as of previous module then we are forced to implement the whole interface and add some dummy implementation or throw Exception. This is known as fat interface.

So the Interface Segregation Principle states that in place of using one fat interface we create multiple small interfaces for submodules.

Let’s take an example of a printer which can print, fax and scan. So we have a system like this:

interface ISmartPrinter {
    void print();
    void fax();
    void scan();
}

class SmartPrinter implements ISmartPrinter {

    public void print() {
         // Printing code.
    }
    
    public void fax() {
         // Fax Code.
    }
    
    public void scan() {
         // Scanning code.
    }
}

Now suppose we also have a not so good printer which can only print, so we’re forced to implement the whole interface like:

class EconomicPrinter implements ISmartPrinter {
    public void print() {
        //Yes I can print.
    }
    
    public void fax() {
        throw new NotSupportedException();
    }
    
    public void scan() {
        throw new NotSupportedException();
    }
}

This is not good.

So what we can do is apply ISP and divide the fat interface ISmartPrinter into three smaller interfaces, IPrinter, IFax and IScanner like this:

interface IPrinter {
    void print();
}

interface IFax {
    void fax();
}

interface IScanner {
    void scan();
}


class SmartPrinter implements IPrinter, IFax, IScanner {

    public void print() {
         // Printing code.
    }
    
    public void fax() {
         // Fax Code.
    }
    
    public void scan() {
         // Scanning code.
    }
}


class EconomicPrinter implements IPrinter {
    public void print() {
        //Yes I can print.
    }
}

Keep in mind that using ISP can result in many small interfaces in code, but it will reduce the complexity, increase testability and also improve code quality. Smaller interfaces are easier to implement, improving flexibility and the possibility of reuse. But ISP can also be used based on experience, like identifying the areas which are more likely to have an extension in future.

Next: Dependency Inversion Principle