Object Oriented Programming

Introduction

Object Oriented Programming is a programming paradigm (i.e. a way of organizing code) that is supported in many different programming languages, such as Python.

Understanding OOP is not just about learning the syntax (although that is a big part of it). Having an intuition about design philosphy is also essential for OOP.

If you have programmed in another object-oriented programming language, see the section about comparisons between Java and Python.

Terminology

In this section, we will use the following Python class as an example:

class Computer(object):
    population = 0

    def __init__(self, owner, operating_system):
        self.owner = owner
        self.operating_system = operating_system
        Computer.population += 1

    def boot(self):
        print('Starting up', self.operating_system)
        time = 10
        while time > 0:
            print(time)
            time -= 1

Here's a list of terminology we will be covering:

Basics

A class is like a blueprint that is used to create an object. In this case, we have a class that is used to create a Computer.

An object is an instance of a class. For example, the code above tells Python how to build a Computer, but it won't actually create a new Computer. To create objects, you use the following syntax:

my_computer = Computer('Albert', 'Linux')

The left side is just a variable assignment — nothing new. On the right side, it looks like we are calling Computer as if it was a function. This is known as the constructor of an object. In general, the syntax for creating objects is

ClassName(args)

Methods

One question you might have is, where is the constructor defined? Take a look at __init__ definition in the Computer class:

def __init__(self, owner, operating_system):
    ...

It looks like we're defining a function. When we use a def statement inside of a class, we create a method. A method is very much like a function — we can call it with arguments, and it performs some computations. The only difference is that a method is tied to the class. To see what this means, consider the following code:

my_computer = Computer('Albert', 'Linux')
my_computer.boot()

The second line is how we call the boot method. This will print out "Starting up Linux", followed by a countdown. Notice the syntax we use: we put the instance (my_computer), followed by a dot (.), then the method name (boot). Just like with functions, we call methods with parentheses.

You might be wondering why we passed in zero arguments to boot, when it looks like it should take in one:

def boot(self):
    ...

The self argument refers to the object that is calling boot — in this case, self is my_computer. The dot notation implicitly passes in my_computer as self, so we don't have to put it as an argument when we call boot. We'll talk more about this later.

In Python, the __init__ method is a special method that is used as the constructor for an object. For example, when we create a Computer object like this

my_computer = Computer('Albert', 'Linux')

we are really calling the __init__ method. 'Albert' gets bound to the owner parameter, and 'Linux' gets bound to the operating_system parameter.

Variables

There are three types of variables:

  • Local variables
  • Instance variables
  • Class variables

Which type a variable belongs to depends on the syntax with which it is created.

Local Variables

A local variable exists only in the local frame of a method. For example, the variable time in the boot method is a local variable. Notice that it looks just like a normal variable. When boot is called, the local variable time is initialized to 10, and decremented until it reaches 0. Once we finish calling boot, time gets discarded along with the local frame for boot.

Here's a list of instance variables in the Computer class:

  • Inside __init__:

    • self
    • owner (not to be confused with self.owner)
    • operating_system (not to be confused with self.operating_system)

  • Inside boot:

    • self
    • time

Instance Variables

An instance variable has a couple of properties:

  • persists across method calls: unlike a local variable, an instance variable retains its most recent value even after a method call finishes.
  • each instance has its own instance variables: an instance variable for one instance might have a different value for another instance. Each instance keeps track of its own instance variables.

Notice in the __init__ method that we create two instance variables: self.owner and self.operating_system. Instance variables are always preceded by self.. This means the variable self.owner is a different variable than owner — they just happen to have similar names.

The first property (persistence) is demonstrated with the self.operating_system instance variable. When we call __init__, self.operating_system is assigned to 'Linux' (in the case of my_computer). When we call boot, the value of self.operating_system is still 'Linux'.

The second property is demonstrated with the following:

computer1 = Computer('Albert', 'Linux')
computer2 = Computer('Robert', 'Mac')

For computer1, the self.owner instance variable is assigned to 'Albert'. The self.owner variable for computer2 is different, as it is assigned to 'Robert'.

Here's a list of instance variables in the Computer class:

  • self.owner
  • self.operating_system

Class Variables

The final type of variable is called a class variable, also known as a static variable. A class variable has the following properties:

  • persists across method calls: just like an instance variable
  • same for all instances: every instance shares the same class variable.

Like instance variables, class variables have persistent state. The difference between the two is that class variables are shared by all instances of the class. In the Compter class, the population variable is a class variable, and has an initial value of 0. To illustrate this, consider the following:

computer1 = Computer('Albert', 'Linux')   # population is now 1
computer2 = Computer('Robert', 'Mac')     # population is now 2

Notice that, every time we create a new Computer, the population class variable is incremented by 1.

Syntax

Defining classes

Now that we know the terminology associated with OOP, we have to know how to use it.

To define a class, we do the following:

class ClassName(object):
    ...

By convention, class names are written in CamelCase (e.g. ClassName as opposed to class_name). Don't worry about the object part yet — we'll explain that when we talk about inheritance.

To create methods, use a def statement inside the class:

class ClassName(object):

    def method_name(self, args):
        ...

The first argument of every method must be a reference to the instance (we don't talk about static methods in this class). This parameter is, by convention, called self. You can technically call it anything you want, but you shouldn't. Other than that, writing a method is just like writing a normal function.

The __init__ method, as mentioned above, is used to construct objects of that class:

class ClassName(object):

    def __init__(self, args):
        ...

You must put two underscores (_) before and after init, otherwise Python will not be able to find __init__.

To call a method inside of another method, use dot notation:

class ClassName(object):

    def method1(self, arg):
        self.method2(arg, 10)

    def method2(self, arg1, arg2):
        return arg1 + arg2

We call method2 from within method1. Notice the self. notation.

To create a class variable, we usually put the variable declaration at the top of the class:

class ClassName(object):
    class_variable = 9001

    def __init__(self, args):
        ClassName.class_variable -= 10

Notice that, when declaring the class_variable outside of a method, we just put the variable name without dot notation (i.e. class_variable as opposed to ClassName.class_variable). If we access the class variable from inside a method, we need to use dot notation (i.e. ClassName.class_variable, as opposed to class_variable). Notice we put the name of the class before the dot.

Instance variables must be accessed inside of methods. This is because we need a self variable, and self is a local variable defined inside of methods. Here is some sample syntax using instance variables:

class ClassName(object):

    def method(self, arg):
        self.instance_variable = arg
        local_variable = self.instance_variable + 4
        return self.instance_variable / local_variable

Using a class

To create an instance of a class, we use this syntax:

ClassName(args)

Keep in mind that the constructor calls the __init__ method. The object in

class ClassName(object):
    ...

has nothing to do with the arguments passed into the constructor.

To call an instance's method, use dot notation:

instance = ClassName(arg1, arg2)
instance.method1(foo)

You can also access instance variables with similar syntax:

x = instance.instance_variable
instance.instance_variable = 5

The first line just gets the value associated with instance_variable. The second line reassigns the instance_variable to 5 (remember that every isntance keeps its own copy of instance variables).

To access class variables, use the following syntax:

ClassName.class_variable = 4
x = ClassName.class_variable

Inheritance

A powerful feature of OOP is inheritance. The idea is that a class can "inherit" methods and variables from another class. We can use this to create a hierarchy of classes. For example:

  • Animal

    • Mammal

      • Dog
      • Cat

    • Fish

      • Shark

Notice that each sublist is a more specific version of its predecessor (e.g. Dogs are a more specific type of Mammal).

We can implement this inheritance in code like this:

class Animal(object):
    ...

class Mammal(Animal):
    ...

class Dog(Mammal):
    ...

class Cat(Mammal):
    ...

Notice that we put class names inside the parentheses. For example, class Mammal(Animal) reflects the fact that Mammals are a more specific type of Animal. We call Mammal the subclass (the inheriting class) and Animal the superclass (the inherited class).

When a subclass inherits from a superclass, the subclass has access to all the methods of the superclass. Consider the following example:

class A(object):
    def method1(self, arg):
        print('hi')

class B(A):
    ...

When we create a B object, we will be able to do the following:

b = B()
b.method1(4)

The second line will print out hi. Notice that, even though we don't define method1 explicitly in B, we can still call it. When looking for b.method1, Python will first look in the B class for a method call method1. If it can't find one (as in this case), it looks in the superclass (A) for method1.

This is useful for writing concise code — we don't need to repeat something we've already written.

Overriding methods

If we want a subclass to have a method that shares the same name as a superclass method, we can override that method:

class A(object):
    def method1(self, arg):
        print(arg)

class B(A):
    def method1(self):
        print('hello')

Both A and B have a method called method1. Consider the following example:

b = B()
b.method1()

When Python looks for b.method1, it looks in the current class first (i.e. B). Because it's able to find method1 in B, it immediately uses that version of method1. That's why the second line takes 0 arguments and always prints hello.

You can also call a superclass method, even if you override it:

class A(object):
    def method1(self, arg):
        print(arg)

class B(A):
    def method1(self):
        A.method1(self, 'hello')

The last line introduces syntax for calling a superclass method. We use dot notation, where the superclass name (A) precedes the dot. Also notice that, when we call A.method1, we have to explicitly pass in self as the first argument.

Note 1: this implementation of B will also print out hello, just like in the previous implementation. The only difference is that this implementation calls the superclass method.

Note 2: we don't cover super() in this class. If you're interested, see this.

object

You'll undoubtedly have noticed the object in most of our class definitions:

class ClassName(object):
    ...

What is object? Notice the syntax is the same as if we were using inheritance. In fact, that's exactly what is happening: when we put object there, we inherit from the object class.

object is the superclass of all classes in Python — all classes inherit from object. The object class has some general methods defined, such as __init__.

Design Principles

In addition to understanding the terminology and syntax of OOP, we also need to understand some design principles for OOP. The power of OOP comes from several qualities:

  • data abstraction: classes can be used to represent literally any object the programmer wants. In addition, selectors and constructors are logically grouped inside of classes
  • inheritance: class hierarchies and method overriding can make code more concise, and in some cases make class behavior more intuitive
  • Polymorphism and interfaces: method overriding allows methods of the same name to do different things. This means other programmers need only use one method to achieve different behavior for different types of objects (it's easier just to remember one method name and makes code cleaner)

With that being said, here are some fundamental design principles. These principles apply to any object-oriented language, not just Python.

Use inheritance whenever possible

If you find that two or more classes share a lot of the same code, you can make one class the superclass. This way, the other classes can just inherit methods, making your code concise.

In addition to aesthetiscm, there is a practical use to inheritance. If you need to update a method (to make it more efficient, do something different, etc.) that is shared by many classes, inheritance allows you to change it in just one class. If you don't use inheritance, you'll have to go through each class that shares that method — this makes it easier to make mistakes! For this reason, inheritance makes code maintenance a lot easier.

Consider the following example:

class Animal(object):
    def __init__(self, name, age):
        self.name = name
        self.age = age

class Mammal(Animal):
    def __init__(self, name, age):
        self.name = name
        self.age = age

Notice that the __init__ method of both classes takes exactly the same arguments, and also creates the exact same instance variables. Using inheritance, we could omit the __init__ definition in the Mammal class completely!

Call superclass methods when possible

This is similar to the first principle. Instead of duplicating code inside of a method, just call the superclass method instead:

class Animal(object):
    def __init__(self, name, age):
        self.name = name
        self.age = age

class Pet(Animal):
    def __init__(self, name, age, owner):
        self.name = name
        self.age = age
        self.owner = owner

In this example, defining an __init__ in Pet is necessary, since it takes in one more argument than Animal, and also creates another instance variable. However, we can call the superclass method to make this more concise:

class Animal(object):
    def __init__(self, name, age):
        self.name = name
        self.age = age

class Pet(Animal):
    def __init__(self, name, age, owner):
        Animal.__init__(self, name, age)
        self.owner = owner

Don't forget to pass in self when calling a superclass method.

Instance vs. Class variables

Deciding whether a variable should be an instance variable or a class variable is an important consideration. Part of it depends on the programmer and the application, but there are some general guidelines:

  • If a variable is different for every instance of a class, it should be an instance variable
  • If a variable is always the same for every instance, it should be a class variable

Methods vs. Variables

In general, methods can be thought of as "actions", while variables can be thought of as "data." For example, consider the following class:

class Dog(object):
    def __init__(self, age):
        self.age = age

    def bark(self):
        return 'woof!'

Notice that bark is a method, even though it always returns the same thing ('woof!'), because barking is an action. Conversely, age is an instance variable because it is piece of information about the Dog.

Other Features

Python includes a variety of features for OOP, some of which are shared by other languages, some of which are not.

Variable lookup

When we ask Python to lookup a variable like this:

my_instance.variable

where my_instance is an instance of a class, Python will first look for an instance variable called variable. If it finds one, it just uses that.

If Python can't find an instance variable by that name, it will look for a class variable called variable. If it can't find one, Python looks in the superclass for a class variable called variable. Python continues this process until it hits object, at which point it raises an AttributeError to signal that it couldn't find that variable.

Property methods

Python supports a feature called property methods. Consider the following code:

class A(object):

    @property
    def age(self):
        ...

The @property decorator labels the age method as a property method. Now when we want to use age, we access it as if it were a variable, not a method:

a = A()
x = a.age
y = a.age()   # Error!

Notice that you cannot call a property method.

Property methods are useful for a number of reasons:

  • Data protection: you cannot reassign a property method. This prevents other users from tampering with your classes (e.g. you can't do a.age = 5 in the code above)
  • Polymorphism: say an external program is interacting with many different classes, some of which have age as an instance variable, some of which have age as a property method. The fact that age is a property method for some classes makes no difference for the external program, since it access age the same way it does an instance variable.

Static methods

We don't talk about static methods in 61A, but they do exist in Python. In Python, there is a distinction between static methods and class methods. Each one is created with a decorator:

class A(object):

    def normal_method(self, arg):
        ...

    @staticmethod
    def static_method(arg):
        ...

    @classmethod
    def class_method(cls, arg):
        ...

Compare each type of method with the normal method. The static method does not take in a self, and thus has no way to reference the instance. You can call static methods by using dot notation, with the class name preceding the dot:

A.static_method(4)
a = A()
a.static_method(4)  # Error!

Class methods also do not take in a self, so they don't have access to the instance. However, they do take in a cls, which is a reference to the class. To call a class method, you can use dot notation with either an instance or the class:

A.class_method(4)
a = A()
a.class_method(4)

Multiple Inheritance

We don't talk about multiple inheritance in 61A, but Python does support it. When defining a class, you can make it inherit from multiple superclasses:

class Sub(Sup1, Sup2, Sup3):
    ...

Python allows an arbitrary number of superclasses.

Multiple inheritance gets complicated when two or more superclasses have methods of the same name. Python (version 3 and beyond) resolves name conflicts by always looking at superclasses from left to right. For example, if Sup1 and Sup3 both have a method called method, Python will use Sup1's version of method.

Java: a comparison

If you have programmed in Java, you will be familiar with many of the OOP concepts present in Python. There are some differences however:

  • There is no this in Python. self is a close equivalent, but self is always required, whereas this in Java is optional.
  • self is always required in non-static methods in Python. Furthermore, self is actually bound to an object.
  • Variables and methods can be dynamically created in Python. If an attribute was not created in the __init__, it can still be created inside of other methods, or even by external programs.
  • There are no protection modifiers in Python. There isn't a formal way to declare a variable private; in other words, all variables are public. There is a partial remedy: if a variable name begins with two underscores and ends with at most one underscore, Python name-mangles the variable so that it has a different name. If an external program can guess the new name (which is easy to do), it will still have access to the variable.
  • When a variable is declared in Python, it must be immediately initialized with a value. This is true of Python variables in general.
  • There are no final variables in Python.
  • There is no method overloading. A name can be bound to only one method at a time. There's a partial remedy to this in the form of default arguments.
  • Methods and variables share the same namespace. That means a method and an instance/class variable cannot share the same name.
  • super behaves differently in Python. There is a super function, but its syntax is different than super in Java.
  • There are no abstract classes in Python.
  • Only class names are CamelCase in Python. Methods and variables, by convention, should use lower_case_and_underscores.
  • Python supports multiple inheritance.