I have seen in many libraries like Spring
which use a lot of interfaces with single methods in them like BeanNameAware
, etc.
You create interfaces not according to the number of methods in it but to define behaviour expected by components of your systems to deliver a single responsibility to their neighbors. If you follow this simple principle/rule, you might or might not end up with single method interfaces, depending on the responsibility you are defining. I like to keep tests stupid simple and the application very flexible so I usually have many of those
With Java 8, keeping the single method interface is quite useful, since single method interfaces will allow the usage of closures and "function pointers". So, whenever your code is written against a single method interface, the client code may hand in a closure or a method (which must have a compatible signature to the method declared in the single method interface) instead of having to create an anonymous class. In contrast, if you make one interface with more than one method, the client code will not have that possibility. It must always use a class that implements all methods of the interface.
So as a common guideline, one can say: If a class that only exposes a single method to the client code might be useful to some client, then using a single method interface for that method is a good idea. A counter example to this is the Iterator
interface: Here, it would not be useful having only a next()
method without a hasNext()
method. Since having a class that only implements one of these methods is no use, splitting this interface is not a good idea here.
Example:
interface SingleMethod{ //The single method interface
void foo(int i);
}
class X implements SingleMethod { //A class implementing it (and probably other ones)
void foo(int i){...}
}
class Y { //An unrelated class that has methods with matching signature
void bar(int i){...}
static void bar2(int i){...}
}
class Framework{ // A framework that uses the interface
//Takes a single method object and does something with it
//(probably invoking the method)
void consume(SingleMethod m){...}
}
class Client{ //Client code that uses the framework
Framework f = ...;
X x = new X();
Y y = new Y();
f.consume(x); //Fine, also in Java 7
//Java 8
//ALL these calls are only possible since SingleMethod has only ONE method!
f.consume(y::bar); //Simply hand in a method. Object y is bound implicitly
f.consume(Y::bar2); //Static methods are fine, too
f.consume(i -> { System.out.println(i); }) //lambda expression. Super concise.
//This is the only way if the interface has MORE THAN ONE method:
//Calling Y.bar2 Without that closure stuff (super verbose)
f.consume(new SingleMethod(){
@Override void foo(int i){ Y.bar2(i); }
});
}
Interfaces with only one (or few) methods is the key to the highly useful Strategy pattern, which is "some design standard which advocates the use of these type of interfaces".
Another common scenario is when you want a callback. Foo calls Bar as an asynchronous task, and when Bar is finished with something, the result is sent back to Foo using a callback -- which can be an interface containing only one method. (An example of this is the many listeners in Android, Event Listeners in Swing...)
Also, if you have two classes that are tightly coupled with one another (let's call them Foo and Bar). Foo uses nearly all of Bar's methods, but Bar only needs some a few of those from Foo. Foo can implement FooInterface
which is then sent to Bar. Now the coupling is looser, because Bar only knows about the FooInterface, but does not care about the other methods the implementing class contains.
Favor Composition over Inheritance
tutorial of Head First Design Pattern
book recommends this approach to add functionality dynamically to a class. Let's take below case:
public interface Quackable {
public void quack();
}
public class Quacks implements Quackable {
public void quack(){
//quack behavior
}
}
public class DontQuack implements Quackable {
public void quack(){
//dont quack
}
}
public class QuackableDuck{
Quackable quack; //add behavior dynamicall
}
So QuackableDuck class can add feature dynamically.
quack = new Quacks();
//or
quack = new DontQuack();
So similarly you can add multiple behavior to the class dynamically.
In what scenarios does it make sense to keep single method interfaces?
In such a scenarios when you need an interface with only one method.
Interfaces are used to encapsulate a common behavior of several classes. So if you have several places in your code where you need to call only limited set of class methods, it's time to introduce an interface. The number of methods depends on what exactly do you need to call. Sometimes you need one method, sometimes two or more, sometimes you don't need methods at all. What matters is that you can separate behavior from implementation.