Notebook download

Defining your own classes¶

User Defined Types¶

A class is a user-programmed Python type (since Python 2.2!)

It can be defined like:

In [1]:

class Room(object):
    pass

Or:

In [2]:

class Room():
    pass

Or:

In [3]:

class Room:
    pass

What's the difference? Before Python 2.2 a class was distinct from all other Python types, which caused some odd behaviour. To fix this, classes were redefined as user programmed types by extending object, e.g., class room(object).

So most Python 2 code will use this syntax as very few people want to use old style python classes. Python 3 has formalised this by removing old-style classes, so they can be defined without extending object, or indeed without braces.

Just as with other python types, you use the name of the type as a function to make a variable of that type:

In [4]:

zero = int()
type(zero)

Out[4]:

int

In [5]:

myroom = Room()
type(myroom)

Out[5]:

__main__.Room

In the jargon, we say that an object is an instance of a particular class.

__main__ is the name of the scope in which top-level code executes, where we've defined the class Room.

Once we have an object with a type of our own devising, we can add properties at will:

In [6]:

myroom.name = "Living"

In [7]:

myroom.name

Out[7]:

'Living'

The most common use of a class is to allow us to group data into an object in a way that is easier to read and understand than organising data into lists and dictionaries.

In [8]:

myroom.capacity = 3
myroom.occupants = ["Graham", "Eric"]

Methods¶

So far, our class doesn't do much!

We define functions inside the definition of a class, in order to give them capabilities, just like the methods on built-in types.

In [9]:

class Room:
    def overfull(self):
        return len(self.occupants) > self.capacity

In [10]:

myroom = Room()
myroom.capacity = 3
myroom.occupants = ["Graham", "Eric"]

In [11]:

myroom.overfull()

Out[11]:

False

In [12]:

myroom.occupants.append(['TerryG'])

In [13]:

myroom.occupants.append(['John'])

In [14]:

myroom.overfull()

Out[14]:

True

When we write methods, we always write the first function argument as self, to refer to the object instance itself, the argument that goes "before the dot".

This is just a convention for this variable name, not a keyword. You could call it something else if you wanted.

Constructors¶

Normally, though, we don't want to add data to the class attributes on the fly like that. Instead, we define a constructor that converts input data into an object.

In [15]:

class Room:
    def __init__(self, name, exits, capacity, occupants=[]):
        self.name = name
        self.occupants = occupants  # Note the default argument, occupants start empty
        self.exits = exits
        self.capacity = capacity

    def overfull(self):
        return len(self.occupants) > self.capacity

In [16]:

living = Room("Living Room", {'north': 'garden'}, 3)

In [17]:

living.capacity

Out[17]:

Methods which begin and end with two underscores in their names fulfil special capabilities in Python, such as constructors.

Object-oriented design¶

In building a computer system to model a problem, therefore, we often want to make:

classes for each kind of thing in our system
methods for each capability of that kind
properties (defined in a constructor) for each piece of information describing that kind

For example, the below program might describe our "Maze of Rooms" system:

We define a "Maze" class which can hold rooms:

In [18]:

class Maze:
    def __init__(self, name):
        self.name = name
        self.rooms = {}

    def add_room(self, room):
        room.maze = self  # The Room needs to know which Maze it is a part of
        self.rooms[room.name] = room

    def occupants(self):
        return [occupant for room in self.rooms.values()
                for occupant in room.occupants.values()]

    def wander(self):
        """Move all the people in a random direction"""
        for occupant in self.occupants():
            occupant.wander()

    def describe(self):
        for room in self.rooms.values():
            room.describe()

    def step(self):
        self.describe()
        print("")
        self.wander()
        print("")

    def simulate(self, steps):
        for _ in range(steps):
            self.step()

And a "Room" class with exits, and people:

In [19]:

class Room:
    def __init__(self, name, exits, capacity, maze=None):
        self.maze = maze
        self.name = name
        self.occupants = {}  # Note the default argument, occupants start empty
        self.exits = exits  # Should be a dictionary from directions to room names
        self.capacity = capacity

    def has_space(self):
        return len(self.occupants) < self.capacity

    def available_exits(self):
        return [exit for exit, target in self.exits.items()
                if self.maze.rooms[target].has_space()]

    def random_valid_exit(self):
        import random
        if not self.available_exits():
            return None
        return random.choice(self.available_exits())

    def destination(self, exit):
        return self.maze.rooms[self.exits[exit]]

    def add_occupant(self, occupant):
        occupant.room = self  # The person needs to know which room it is in
        self.occupants[occupant.name] = occupant

    def delete_occupant(self, occupant):
        del self.occupants[occupant.name]

    def describe(self):
        if self.occupants:
            print(f"{self.name}: " + " ".join(self.occupants.keys()))

We define a "Person" class for room occupants:

In [20]:

class Person:
    def __init__(self, name, room=None):
        self.name = name

    def use(self, exit):
        self.room.delete_occupant(self)
        destination = self.room.destination(exit)
        destination.add_occupant(self)
        print("{some} goes {action} to the {where}".format(some=self.name,
                                                           action=exit,
                                                           where=destination.name))

    def wander(self):
        exit = self.room.random_valid_exit()
        if exit:
            self.use(exit)

And we use these classes to define our people, rooms, and their relationships:

In [21]:

graham = Person('Graham')
eric = Person('Eric')
terryg = Person('TerryG')
john = Person('John')

In [22]:

living = Room('livingroom', {'outside': 'garden',
                             'upstairs': 'bedroom', 'north': 'kitchen'}, 2)
kitchen = Room('kitchen', {'south': 'livingroom'}, 1)
garden = Room('garden', {'inside': 'livingroom'}, 3)
bedroom = Room('bedroom', {'jump': 'garden', 'downstairs': 'livingroom'}, 1)

In [23]:

house = Maze('My House')

In [24]:

for room in [living, kitchen, garden, bedroom]:
    house.add_room(room)

In [25]:

living.add_occupant(graham)

In [26]:

garden.add_occupant(eric)
garden.add_occupant(terryg)

In [27]:

bedroom.add_occupant(john)

And we can run a "simulation" of our model:

In [28]:

house.simulate(3)

livingroom: Graham
garden: Eric TerryG
bedroom: John

Graham goes north to the kitchen
Eric goes inside to the livingroom
TerryG goes inside to the livingroom
John goes jump to the garden

livingroom: Eric TerryG
kitchen: Graham
garden: John

Eric goes upstairs to the bedroom
TerryG goes outside to the garden
Graham goes south to the livingroom
John goes inside to the livingroom

livingroom: Graham John
garden: TerryG
bedroom: Eric

Graham goes outside to the garden
John goes outside to the garden
TerryG goes inside to the livingroom
Eric goes downstairs to the livingroom

Object oriented design¶

There are many choices for how to design programs to do this. Another choice would be to separately define exits as a different class from rooms. This way, we can use arrays instead of dictionaries, but we have to first define all our rooms, then define all our exits.

In [29]:

class Maze:
    def __init__(self, name):
        self.name = name
        self.rooms = []
        self.occupants = []

    def add_room(self, name, capacity):
        result = Room(name, capacity)
        self.rooms.append(result)
        return result

    def add_exit(self, name, source, target, reverse=None):
        source.add_exit(name, target)
        if reverse:
            target.add_exit(reverse, source)

    def add_occupant(self, name, room):
        self.occupants.append(Person(name, room))
        room.occupancy += 1

    def wander(self):
        "Move all the people in a random direction"
        for occupant in self.occupants:
            occupant.wander()

    def describe(self):
        for occupant in self.occupants:
            occupant.describe()

    def step(self):
        house.describe()
        print("")
        house.wander()
        print("")

    def simulate(self, steps):
        for _ in range(steps):
            self.step()

In [30]:

class Room:
    def __init__(self, name, capacity):
        self.name = name
        self.capacity = capacity
        self.occupancy = 0
        self.exits = []

    def has_space(self):
        return self.occupancy < self.capacity

    def available_exits(self):
        return [exit for exit in self.exits if exit.valid()]

    def random_valid_exit(self):
        import random
        if not self.available_exits():
            return None
        return random.choice(self.available_exits())

    def add_exit(self, name, target):
        self.exits.append(Exit(name, target))

In [31]:

class Person:
    def __init__(self, name, room=None):
        self.name = name
        self.room = room

    def use(self, exit):
        self.room.occupancy -= 1
        destination = exit.target
        destination.occupancy += 1
        self.room = destination
        print("{some} goes {action} to the {where}".format(some=self.name,
                                                           action=exit.name,
                                                           where=destination.name))

    def wander(self):
        exit = self.room.random_valid_exit()
        if exit:
            self.use(exit)

    def describe(self):
        print("{who} is in the {where}".format(who=self.name,
                                               where=self.room.name))

In [32]:

class Exit:
    def __init__(self, name, target):
        self.name = name
        self.target = target

    def valid(self):
        return self.target.has_space()

In [33]:

house = Maze('My New House')

In [34]:

living = house.add_room('livingroom', 2)
bed = house.add_room('bedroom', 1)
garden = house.add_room('garden', 3)
kitchen = house.add_room('kitchen', 1)

In [35]:

house.add_exit('north', living, kitchen, 'south')

In [36]:

house.add_exit('upstairs', living, bed, 'downstairs')

In [37]:

house.add_exit('outside', living, garden, 'inside')

In [38]:

house.add_exit('jump', bed, garden)

In [39]:

house.add_occupant('Graham', living)
house.add_occupant('Eric', garden)
house.add_occupant('TerryJ', bed)
house.add_occupant('John', garden)

In [40]:

house.simulate(3)

Graham is in the livingroom
Eric is in the garden
TerryJ is in the bedroom
John is in the garden

Graham goes outside to the garden
Eric goes inside to the livingroom
TerryJ goes downstairs to the livingroom

Graham is in the garden
Eric is in the livingroom
TerryJ is in the livingroom
John is in the garden

Eric goes north to the kitchen
TerryJ goes outside to the garden
John goes inside to the livingroom

Graham is in the garden
Eric is in the kitchen
TerryJ is in the garden
John is in the livingroom

Graham goes inside to the livingroom
John goes outside to the garden

This is a huge topic, about which many books have been written. The differences between these two designs are important, and will have long-term consequences for the project. That is the how we start to think about software engineering, as opposed to learning to program, and is an important part of this course.

Exercise: Your own solution¶

Compare the two solutions above. Discuss with a partner which you like better, and why. Then, starting from scratch, design your own. What choices did you make that are different from mine?