1) Introduction

• Suppose you want to simulate a small ecosystem, such as a tidal pool, that contains many different kinds of things
  • Plants (don't move)
  • Fish (swim in three dimensions)
  • Crawly things (cling to the surface most of the time)
• The procedural way to do it uses a type-switch

for time in simulation_period:
    for thing in world:
        if type(thing) is plant:
            update_plant(thing, time)
        elif type(thing) is fish:
            update_fish(thing, time)
        elif type(thing) is creepy_crawly:
            update_creepy_crawly(thing, time)

• But:
  • Every time you add a new type of thing, you have to find and update all the type-switches
  • It's very easy to make a cut-and-paste mistake

2) Objects to the Rescue

• Object-oriented programming (OOP) solves both problems
  • Each object knows how to update itself

for time in simulation_period:
    for thing in world:
        thing.update(time)

• Don't have to change old code when adding new types of things
• No chance of calling the wrong function
• Seems like a small change, but it allows programmers to think and design at a higher level
  • Also allows them to make more powerful mistakes...
• Take two lectures to introduce OOP
  • Ideas apply to all modern languages
  • But there's more variation in form than there is with loops, conditionals, and functions

3) You Can Skip This Lecture If...

• You know what a class is
• You know what methods and member variables are
• You know what encapsulation, inheritance, and polymorphism are

4) Abstract Data Types

• Modern languages encourage programmers to define abstract data types (ADTs)
  • "Abstract" because they hide the details of their implementation
• Programmers interact with them through a limited set of operations, rather than by manipulating data directly
  • Fewer things can go wrong
  • Easier to read resulting code
  • Makes code easier to maintain, since internals can be changed without changing calling code

5) Classes and Instances

• An ADT is usually created by defining a class that specifies:
  • How the ADT stores state (its member variables)
  • What it can do (its methods)
  • And yes, classes are also objects, just like functions
• Programmers then create objects that are instances of that class
  • Each object of a particular ADT shares the class's methods, but has its own members
  • So changes to one object do not affect the state of others


Figure 13.1: Memory Model for Classes and Objects

6) Defining a Class

• Define a class in Python using the class keyword
  • Name of the new class is usually followed by object in parentheses
  • We'll see other options later
• Then ":" and an indented block containing the class's contents

class Empty(object):
    pass

• pass means "do nothing", i.e., create an empty class
  • Not particularly useful, but you have to start somewhere

7) Creating an Instance

• Create a new instance of the class by calling the class name as if it were a function

if __name__ == '__main__':
    first = Empty()
    second = Empty()
    print 'first has id', id(first)
    print 'second has id', id(second)

first has id 5086860
second has id 5086892

• id returns the object's hash code
  • Doesn't mean anything: just distinguishes objects
• Note how main body of program is put in a block under if __name__ == '__main__':
  • Otherwise, it will be executed when other programs import the class

8) Methods

• Give the class methods by defining functions inside it
• The object itself is always passed to the method as its first argument
  • Universally called self
  • Unlike this in C++ and Java, the name is just a convention
  • But everyone uses it, and you should too
• object.method(argument) is equivalent to:
  • Find the class C that object is an instance of
  • Call C.method(object, argument)

class Greeting(object):
    def say(self, name):
        print 'Hello, %s!' % name

if __name__ == '__main__':
    greet = Greeting()
    greet.say('object')

Hello, object!

9) Creating Members

• Every object is a new scope for variable names
  • Just like a module, or a function call
• The values in an object's scope are its members
  • Create members use dotted notation: self.x = 3
  • Gives the current object a new member x with the value 3
  • Or overwrites the existing member x with the value 3

class Point(object):

    def set_values(self, x, y):
        self.x = x
        self.y = y

    def get_values(self):
        return (self.x, self.y)

    def norm(self):
        return math.sqrt(self.x ** 2 + self.y ** 2)

if __name__ == '__main__':
    p = Point()
    p.set_values(1.2, 3.5)
    print 'p is', p.get_values()
    print 'norm is', p.norm()

p is (1.2, 3.5)
norm is 3.7

Figure 13.2: Creating a Simple Point

10) Encapsulation

• Encapsulation is one of the three defining principles of OOP
  • Programs are much easier to write, read, and maintain if object members are only ever accessed by methods
• But unlike C++, Java, and C#, Python doesn't allow programmers to hide methods or data members

p = Point()
p.x = 3.5
p.y = 4.25
print 'point is', p.get_values()

point is (3.5, 4.25)

• Any function or method can see and modify any object's internals
  • Resist the temptation to program this way!
  • If you manipulate an object's internals directly, you have to change your program when you change the object's implementation

11) Constructors

• If a class has a method called __init__, Python will call it when building new instances
  • Hence the name constructor
  • Simpler than creating a blank object, then initializing its members
  • And there's less chance the programmer will forget to do the initialization

class Point(object):

    def __init__(self, x=0, y=0):
        self.reset(x, y)

    def reset(self, x, y):
        assert (type(x) is int) and (x >= 0), 'x is not non-negative integer'
        assert (type(y) is int) and (y >= 0), 'y is not non-negative integer'
        self.x = x
        self.y = y

    def get(self):
        return (self.x, self.y)

    def norm(self):
        return math.sqrt(self.x ** 2 + self.y ** 2)

if __name__ == '__main__':
    p = Point(1, 1)
    print 'point is initially', p.get()
    p.reset(1, 1)
    print 'p moved to', p.get()

point is initially (1, 1)
p moved to (1, 1)

12) Constructor Style

• A class can only have one constructor
  • Some languages allow classes to have several, distinguished by argument types
  • But since Python doesn't use type declarations, this wouldn't work
• It's good style to create all of the object's members in the constructor
  • So that people only have to look in one place to find what members exist
• Note how the class checks values before changing the object's state
  • Remember: fail early, fail often

13) Special Methods

• __init__ is just one example of a special method
  • All have names beginning and ending with double underscore
  • Give programmers a way to make their data types look like those built into Python
• Most widely used is __str__
  • When Python needs a text representation of an object, it:
  • Calls __str__ if it exists, or
  • Creates a default representation that shows the object's location in memory

class Point(object):

    def __init__(self, x=0, y=0):
        self.x = x
        self.y = y

    def __str__(self):
        return '(%4.2f, %4.2f)' % (self.x, self.y)

if __name__ == '__main__':
    p = Point(3, 4)
    print 'point is', p

point is (3, 4)

14) New Classes from Old

• Suppose we have a class Organism that represents living things
  • Common name, scientific name, ...
• Want to create a class Mammal
  • Body temperature, gestation period, ...
• Wrong: copy Organism's definition and add more members and methods
  • "Anything repeated in two or more places will eventually be wrong in at least one."
• Right: use inheritance
  • The second defining principle of OOP
  • Derive a child class from a parent
  • The child has all the members and methods of its parents, plus whatever else we give it

15) Inheritance Example

class Organism(object):

    def __init__(self, common_name, sci_name):
        self.common_name = common_name
        self.sci_name = sci_name

    def get_common_name(self):
        return self.common_name

    def get_sci_name(self):
        return self.sci_name

    def __str__(self):
        return '%s (%s)' % (self.common_name, self.sci_name)

class Mammal(Organism):

    def __init__(self, common_name, sci_name, body_temp, gest_period):
        Organism.__init__(self, common_name, sci_name)
        self.body_temp = body_temp
        self.gest_period = gest_period

    def get_body_temp(self):
        return self.body_temp

    def get_gest_period(self):
        return self.gest_period

    def __str__(self):
        extra = ' %4.2f degrees / %d days' % (self.body_temp, self.gest_period)
        return Organism.__str__(self) + extra

if __name__ == '__main__':
    creature = Mammal('wolf', 'canis lupus', 38.7, 63)
    print creature

wolf (canis lupus) 38.70 degrees / 63 days

Figure 13.3: Memory Model for Inheritance

16) Overriding Methods

• Mammal's constructor calls Organism's to initialize the organism-ish bits of the object
• And Mammal defines its own __str__ method
  • Overrides the one defined by Organism
  • Mammal.__str__ calls Organism.__str__ for the same reason that Mammal.__init__ calls Organism.__init__
• Python always calls the most specific method
  • Keep the memory model in mind when figuring out what this will be

17) Polymorphism

• Polymorphism means "having more than one form"
  • In object-oriented programming, it means handling specific objects in generic ways
  • The third and final defining principle of OOP
• Derive a new class Bird from Organism
  • As long as it only uses common methods, a single piece of code can work with both mammals and birds

class Bird(Organism):

    def __init__(self, common_name, sci_name, incubate_period):
        Organism.__init__(self, common_name, sci_name)
        self.incubate_period = incubate_period

    def get_incubate_period(self):
        return self.incubate_period

    def __str__(self):
        extra = ' %d days' % self.incubate_period
        return Organism.__str__(self) + extra

if __name__ == '__main__':
    creatures = [
        Bird('loon', 'gavia immer', 27),
        Mammal('grizzly bear', 'ursus arctos horribilis', 38.0, 210)
    ]
    for c in creatures:
        print c

loon (gavia immer) 27 days
grizzly bear (ursus arctos horribilis) 38.00 degrees / 210 days

18) Duck Typing

• Most languages only permit polymorphism via inheritance
  • Lowest common ancestor of two classes defines how interchangeable they are
• In Python, any two classes that define the same set of methods can be used interchangeably
  • Duck typing: "If it walks like a duck, and quacks like a duck, it might as well be a duck."

class Mineral(object):

    def __init__(self, common_name, sci_name, formula):
        self.common_name = common_name
        self.sci_name = sci_name
        self.formula = formula

    def get_common_name(self):
        return self.common_name

    def get_sci_name(self):
        return self.sci_name

    def __str__(self):
        return '%s/%s: %s' % (self.common_name, self.sci_name, self.formula)

if __name__ == '__main__':
    things = [
        Mammal('arctic hare', 'Lepus arcticus', 40.1, 50),
        Mineral("fool's gold", 'iron pyrite', 'FeS2')
    ]
    for t in things:
        print t.get_common_name(), 'is', t.get_sci_name()

arctic hare is Lepus arcticus
fool's gold is iron pyrite

• Allows you to create plug-in replacements for files, strings, and other classes after the fact
• But makes it harder to figure out exactly what can be used in place of what

19) The Liskov Substitution Principle

• The Liskov Substitution Principle states that it must always be possible to use an instance of a child class in place of an instance of its parent
• Means that Child.meth may ignore some of Parent.meth's pre-conditions, but may not impose more
  • Equivalently, Child.meth accepts everything thatParent.meth did, and possibly more
  • So any code that could call Parent.meth correctly is guaranteed to call Child.meth correctly too
• And Child.meth must satisfy all the post-conditions of Parent.meth, and may impose more
  • So Child.meth's possible output is a subset of Parent.meth's
  • And any code that works correctly on the output of Parent.meth will still work if given an instance of Child instead
• The same constraint applies when a class evolves over time

20) Tidal Pools Revisited

• How to represent the creatures in a tidal pool?
  • Each species is a class
  • Use inheritance to separate plants from animals
  • Derive both Plant and Animal from Organism
• How to handle movement?
  • Give Organism two methods: can_move and move
    • Plant.can_move() returns False
    • Plant.move() raises an exception
  • Give Organism one method: move
    • Plant.move() does nothing
• Second simplifies code
  • Uses polymorphism instead of a conditional
• On the other hand, the existence of Plant.move implies that plants can do something they can't
  • Can't really choose between them without knowing what the rest of the code needs
• Usually doesn't make sense to design one class on its own

21) Class, Responsibility, Collaborator

• Many programmers use CRC cards when designing OO systems
  • Stands for "class, responsibility, collaborator"
• Standard 3x5 index cards
  • Top is the class name
  • Left side is point-form description of what the class can do
  • Right side lists other classes that this one interacts with


Figure 13.4: CRC Cards

• Designed so that you won't take them too seriously
  • Lay them out on a table
  • Talk through your program's execution
  • Move cards around, scribble new responsibilities and collaborators on them
  • Create new cards as needed

22) Summary

• Classes and objects are just another way to modularize programs
  • But used well, they can make programs much simpler, and much more adaptable
• Remember: the goal is to simplify, not to dazzle