What is the definition of "interface" in object oriented programming

A friend of mine goes back and forth on what "interface" means in programming.

What is the best description of an "interface"?

To me, an interface is a blueprint of a class. Is this the best 开发者_开发知识库definition?

An interface is one of the more overloaded and confusing terms in development.

It is actually a concept of abstraction and encapsulation. For a given "box", it declares the "inputs" and "outputs" of that box. In the world of software, that usually means the operations that can be invoked on the box (along with arguments) and in some cases the return types of these operations.

What it does not do is define what the semantics of these operations are, although it is commonplace (and very good practice) to document them in proximity to the declaration (e.g., via comments), or to pick good naming conventions. Nevertheless, there are no guarantees that these intentions would be followed.

Here is an analogy: Take a look at your television when it is off. Its interface are the buttons it has, the various plugs, and the screen. Its semantics and behavior are that it takes inputs (e.g., cable programming) and has outputs (display on the screen, sound, etc.). However, when you look at a TV that is not plugged in, you are projecting your expected semantics into an interface. For all you know, the TV could just explode when you plug it in. However, based on its "interface" you can assume that it won't make any coffee since it doesn't have a water intake.

In object oriented programming, an interface generally defines the set of methods (or messages) that an instance of a class that has that interface could respond to.

What adds to the confusion is that in some languages, like Java, there is an actual interface with its language specific semantics. In Java, for example, it is a set of method declarations, with no implementation, but an interface also corresponds to a type and obeys various typing rules.

In other languages, like C++, you do not have interfaces. A class itself defines methods, but you could think of the interface of the class as the declarations of the non-private methods. Because of how C++ compiles, you get header files where you could have the "interface" of the class without actual implementation. You could also mimic Java interfaces with abstract classes with pure virtual functions, etc.

An interface is most certainly not a blueprint for a class. A blueprint, by one definition is a "detailed plan of action". An interface promises nothing about an action! The source of the confusion is that in most languages, if you have an interface type that defines a set of methods, the class that implements it "repeats" the same methods (but provides definition), so the interface looks like a skeleton or an outline of the class.

Consider the following situation:

You are in the middle of a large, empty room, when a zombie suddenly attacks you.

You have no weapon.

Luckily, a fellow living human is standing in the doorway of the room.

"Quick!" you shout at him. "Throw me something I can hit the zombie with!"

Now consider:
You didn't specify (nor do you care) exactly what your friend will choose to toss;
...But it doesn't matter, as long as:

• It's something that can be tossed (He can't toss you the sofa)

• It's something that you can grab hold of (Let's hope he didn't toss a shuriken)

• It's something you can use to bash the zombie's brains out (That rules out pillows and such)

It doesn't matter whether you get a baseball bat or a hammer -
as long as it implements your three conditions, you're good.

To sum it up:

When you write an interface, you're basically saying: "I need something that..."

Interface is a contract you should comply to or given to, depending if you are implementer or a user.

I don't think "blueprint" is a good word to use. A blueprint tells you how to build something. An interface specifically avoids telling you how to build something.

An interface defines how you can interact with a class, i.e. what methods it supports.

In Programming, an interface defines what the behavior a an object will have, but it will not actually specify the behavior. It is a contract, that will guarantee, that a certain class can do something.

Consider this piece of C# code here:

using System;

public interface IGenerate
{
    int Generate();
}

// Dependencies
public class KnownNumber : IGenerate
{
    public int Generate() 
    {
        return 5;
    }   
}

public class SecretNumber : IGenerate
{
    public int Generate()
    {
        return new Random().Next(0, 10);
    }
}

// What you care about
class Game
{
    public Game(IGenerate generator) 
    {
        Console.WriteLine(generator.Generate())
    }
}

new Game(new SecretNumber());
new Game(new KnownNumber());

The Game class requires a secret number. For the sake of testing it, you would like to inject what will be used as a secret number (this principle is called Inversion of Control).

The game class wants to be "open minded" about what will actually create the random number, therefore it will ask in its constructor for "anything, that has a Generate method".

First, the interface specifies, what operations an object will provide. It just contains what it looks like, but no actual implementation is given. This is just the signature of the method. Conventionally, in C# interfaces are prefixed with an I. The classes now implement the IGenerate Interface. This means that the compiler will make sure, that they both have a method, that returns an int and is called Generate. The game now is being called two different object, each of which implementant the correct interface. Other classes would produce an error upon building the code.

Here I noticed the blueprint analogy you used:

A class is commonly seen as a blueprint for an object. An Interface specifies something that a class will need to do, so one could argue that it indeed is just a blueprint for a class, but since a class does not necessarily need an interface, I would argue that this metaphor is breaking. Think of an interface as a contract. The class that "signs it" will be legally required (enforced by the compiler police), to comply to the terms and conditions in the contract. This means that it will have to do, what is specified in the interface.

This is all due to the statically typed nature of some OO languages, as it is the case with Java or C#. In Python on the other hand, another mechanism is used:

import random

# Dependencies
class KnownNumber(object):
    def generate(self):
        return 5

class SecretNumber(object):
    def generate(self):
        return random.randint(0,10)

# What you care about
class SecretGame(object):
    def __init__(self, number_generator):
        number = number_generator.generate()
        print number

Here, none of the classes implement an interface. Python does not care about that, because the SecretGame class will just try to call whatever object is passed in. If the object HAS a generate() method, everything is fine. If it doesn't: KAPUTT! This mistake will not be seen at compile time, but at runtime, so possibly when your program is already deployed and running. C# would notify you way before you came close to that.

The reason this mechanism is used, naively stated, because in OO languages naturally functions aren't first class citizens. As you can see, KnownNumber and SecretNumber contain JUST the functions to generate a number. One does not really need the classes at all. In Python, therefore, one could just throw them away and pick the functions on their own:

# OO Approach
SecretGame(SecretNumber())
SecretGame(KnownNumber())

# Functional Approach

# Dependencies
class SecretGame(object):
    def __init__(self, generate):
        number =  generate()
        print number

SecretGame(lambda: random.randint(0,10))
SecretGame(lambda: 5)

A lambda is just a function, that was declared "in line, as you go". A delegate is just the same in C#:

class Game
{
    public Game(Func<int> generate) 
    {
        Console.WriteLine(generate())
    }
}    

new Game(() => 5);
new Game(() => new Random().Next(0, 10));

Side note: The latter examples were not possible like this up to Java 7. There, Interfaces were your only way of specifying this behavior. However, Java 8 introduced lambda expressions so the C# example can be converted to Java very easily (Func<int> becomes java.util.function.IntSupplier and => becomes ->).

To me an interface is a blueprint of a class, is this the best definition?

No. A blueprint typically includes the internals. But a interface is purely about what is visible on the outside of a class ... or more accurately, a family of classes that implement the interface.

The interface consists of the signatures of methods and values of constants, and also a (typically informal) "behavioral contract" between classes that implement the interface and others that use it.

Technically, I would describe an interface as a set of ways (methods, properties, accessors... the vocabulary depends on the language you are using) to interact with an object. If an object supports/implements an interface, then you can use all of the ways specified in the interface to interact with this object.

Semantically, an interface could also contain conventions about what you may or may not do (e.g., the order in which you may call the methods) and about what, in return, you may assume about the state of the object given how you interacted so far.

Personally I see an interface like a template. If a interface contains the definition for the methods foo() and bar(), then you know every class which uses this interface has the methods foo() and bar().

Let us consider a Man(User or an Object) wants some work to be done. He will contact a middle man(Interface) who will be having a contract with the companies(real world objects created using implemented classes). Few types of works will be defined by him which companies will implement and give him results. Each and every company will implement the work in its own way but the result will be same. Like this User will get its work done using an single interface. I think Interface will act as visible part of the systems with few commands which will be defined internally by the implementing inner sub systems.

An interface separates out operations on a class from the implementation within. Thus, some implementations may provide for many interfaces.

People would usually describe it as a "contract" for what must be available in the methods of the class.

It is absolutely not a blueprint, since that would also determine implementation. A full class definition could be said to be a blueprint.

An interface defines what a class that inherits from it must implement. In this way, multiple classes can inherit from an interface, and because of that inherticance, you can

• be sure that all members of the interface are implemented in the derived class (even if its just to throw an exception)
• Abstract away the class itself from the caller (cast an instance of a class to the interface, and interact with it without needing to know what the actual derived class IS)

for more info, see this http://msdn.microsoft.com/en-us/library/ms173156.aspx

In my opinion, interface has a broader meaning than the one commonly associated with it in Java. I would define "interface" as a set of available operations with some common functionality, that allow controlling/monitoring a module.

In this definition I try to cover both programatic interfaces, where the client is some module, and human interfaces (GUI for example).

As others already said, an interface always has some contract behind it, in terms of inputs and outputs. The interface does not promise anything about the "how" of the operations; it only guarantees some properties of the outcome, given the current state, the selected operation and its parameters.

As above, synonyms of "contract" and "protocol" are appropriate.

The interface comprises the methods and properties you can expect to be exposed by a class.

So if a class Cheetos Bag implements the Chip Bag interface, you should expect a Cheetos Bag to behave exactly like any other Chip Bag. (That is, expose the .attemptToOpenWithoutSpillingEverywhere() method, etc.)

Conventional Definition - An interface is a contract that specifies the methods which needs to be implemented by the class implementing it.

The Definition of Interface has changed over time. Do you think Interface just have method declarations only ? What about static final variables and what about default definitions after Java 5.

Interfaces were introduced to Java because of the Diamond problem with multiple Inheritance and that's what they actually intend to do.

Interfaces are the constructs that were created to get away with the multiple inheritance problem and can have abstract methods , default definitions and static final variables.

http://www.quora.com/Why-does-Java-allow-static-final-variables-in-interfaces-when-they-are-only-intended-to-be-contracts

A boundary across which two systems communicate.

Interfaces are how some OO languages achieve ad hoc polymorphism. Ad hoc polymorphism is simply functions with the same names operating on different types.

In short, The basic problem an interface is trying to solve is to separate how we use something from how it is implemented. But you should consider interface is not a contract. Read more here.