Any language that supports object orientation will let you create objects, either by instantiating them from classes or by building them from prototypes. Class-based object orientation becomes significantly more powerful when you are able to use inheritance and polymorphism. It is also a great convenience if you can use multiple constructors. Python permits all these.

In the real world, we often create hierarchies of things. For example, a vehicle is “a machine that is used to carry people or goods from one place to another”.¹⁾ Based on this general concept, we may define a category of two-wheel vehicles that includes bicycles and motorcycles, a category called car that includes sedans, coupes, and convertibles, a truck category that includes buses, vans, tractors, and so on.

In these kinds of hierarchies, we typically start with a general class of things at the root of the tree, and all the other classes of things are more specialized versions of the general class of things.

This is the essence of inheritance in object oriented programming. For example, to solve a particular programming problem, we might define a Person class, and based on that definition we might then define an Employee class, and then based on that Employee class we might define Manager, Staff, and Hourly employee classes.

Or, we might define a general Shape class. Then based on the Shape class we might define Rectangle and Ellipse classes. Then based on the Rectangle class we might define a Square class (a square is rectangle with equal height and width), and based on the Ellipse class we might define a Circle (a circle is an ellipse with zero eccentricity).

In both of the above cases, the classes closest to the root of the tree are more general than the classes toward the bottom. A Square is a special kind of Rectangle and a Rectangle is a special kind of Shape. An Hourly employee is a special kind of Employee, and an Employee is a special kind or Person. For this reason, we often say that inheritance defines “is a” relationships.

We call trees of classes like the above class heirarchies. The class at the root of the tree is a base class, and classes that inherit from a base classes are derived classes.²⁾

One of the advantages of implementing classes with inheritance is code reuse. Almost all class-based object oriented languages let you define derived classes without having to re-write the base class code. The idea is that you write the base class code once. Then when you write the derived classes, you add only whatever new code is required for the derived class or new definitions for old code that must be overridden.

Since you do not need to rewrite the code that is common to both base and derived classes, you end up writing less code. But even more important, when you fix a bug in the base class, it automatically propagates to the derived classes.

Another advantage of using inheritance is that with statically typed languages (such as C++ and Java), it facilitates polymorphism (discussed below). Dynamically typed languages (like Python) behave polymorphically almost by definition.

To demonstrate the use of inheritance in Python, we are going to create a specialized version of ClickerCounter called ClickUpDown.

In addtion to the two buttons found on a ClickerCounter (for incrementing and resetting the count), a ClickUpDown object will add another that resets the count. In terms of a Python model, a ClickUpDown is identical to a ClickerCounter except that it has an additional method: click_down.

Here is the base class:

# base class definition
class ClickerCounter():
    def __init__(self):
        self.count = 0
 
    # accessor for count
    def get_count(self):
        return self.count
 
    # click the counter
    def click(self):
        self.count = self.count + 1
 
    # reset the count
    def reset(self):
        self.count = 0

And here is the derived class:

# derived class definition
class ClickUpDown(ClickerCounter):
    # click down the counter
    def clickdown(self):
        self.count = self.count - 1

That's it!

Using the new class:

b = ClickUpDown()       # instantiate a ClickUpDown object
b.click()               # 1
b.click()               # 2
b.click()               # 3
b.clickdown()           # should be 2
print b.get_count()
 
a = ClickerCounter()    # instantiate a ClickerCounter object
a.click()               # 1
a.click()               # 2
a.click()               # 3
a.clickdown()           # can't do that!
print a.get_count()

In statically typed programming languages, the behavior of an object bound to a variable might be:
a) the behavior associated with the class of the variable used to reference the object, or
b) the behavior associated with the class of the object to which the variable is bound.

We call b) above polymorphism. In other words, when a language behaves polymorphically, the behavior of objects is based on the class to which the object belongs—not to the class of the variable by which you are accessing the object.

Because Python is dynamically typed, Python is polymorphic without any extra effort on the part of the programmer. When variables are bound to objects, the type of the variable changes—therefore the behavior is always determined by the object's class.

a = ClickUpDown()       # a now references a ClickUpDown object
a = ClickerCounter()    # a now references a ClickerCounter object

In fact, it's hard to make Python behave non-polymorphically.

Let's now add a click_limit feature to our ClickerCounter. When the user clicks past the click_limit, the count will automatically reset to zero. By default, the click_limit will be 100,000. Let us also let the user set the click_limit when she or he instantiates a ClickerCounter.

To do this, we will need to add an instance variable to store the click_limit, add some logic to the click method, and add a parameterized constructor:

class ClickerCounter():        
    # parameterized constructor
    def __init__(self, click_limit = 100000):
        self.count = 0
        self.click_limit = click_limit
 
    # accessor for count
    def get_count(self):
        return self.count
 
    # click the counter
    def click(self):
        if self.count < self.click_limit:
            self.count = self.count + 1
        else:
            self.reset()
 
    # reset the count
    def reset(self):
        self.count = 0

To use it:

a = ClickerCounter(3)       # make a clicker-counter that counts up to 3
a.click()
a.click()
a.click()
print a.get_count()         # should be 3
a.click()                   # should go from 3 to 0
print a.get_count()         # should be 0

Copyright © 2011-2012 Mithat Konar. All rights reserved.