Classes are templates for making objects. Objects are instances of the 'class' that defines them. There can be many objects that are instances of a class, like there are many individual people that are humans. Classes define the attributes and functionality of objects. In Python, the functions defined in a class are known as 'methods' and the 'type' of an object is the class that defines it.
Python class definitions can be embedded throughout programs, but it is considered good practice to organise classes in 'modules'.
Modules are source code files, the name of the module is given by the filename which should be a short word all in lowercase. Modules are commonly stored in a directory alongside the main script that is run. If they are simply placed in the same directory as the script that is run, then there can be name collisions with modules in the standard library and this is best avoided. It is good to organise code into modules as this makes it more reusable, and having lots of code in one file becomes cumbersome. (Details on modules is provided after the next ABM practical.)
The name of a class should start with a capital letter and be in 'CamelCase' (with no spaces and mostly in lower case except for capitalised first letters for each word).
A class is defined using the keyword 'class' followed by the name of the class, parentheses, and a colon, for example:
class Agent():As mentioned, in Python, functions inside classes are called 'methods'. A class in Python has one and only one constructor method named '__init__'. The constuctor method is used to initialise attributes of an instance when it is constructed. A default parameter, by convention called 'self' is always first parameter. This 'self' parameter effectively refers to a generic instance of the class and is used within the class to define and refer to attributes and methods of the class. For example, instances of the following 'Agent' class are instantiated with 'x' and 'y' attribute variables that are initially set to be zero:
class Agent():
def __init__(self):
self.x = 0
self.y = 0
Suppose this class was in a file named 'agentframework.py' - and so in a module named 'agentframework', it could be instantiated from a script in the same directory as 'agentframework.py' as follows:
import agentframework
a = agentframework.Agent()
And the 'x' and 'y' aatribute variables of 'a' - the instance of the class could then be accessed using the dot operator '.' as follows:
print(a.x) # <-- Prints the value of a.x
print(a.y) # <-- Prints the value of a.y
a.x = 3 # <-- Sets the values of a.x to be 3
a.y = a.y + 1 # <-- Increases the value of a.y by 1
A class constructor method may accept multiple arguments. For example, the initial values for 'x' and 'y' could be passed in as follows:
class Agent():
def __init__(self, x, y):
self.x = x
self.y = y
When instantiating an instance of this Agent class, it is necessary to pass in two argument as the constructor method does not set defaults of these values. The following code will instantiate an Agent object using the latest class definition above:
import agents
a = agents.Agent(0, 0)
The parameters 'x' and 'y' do not have to have names corresponding to the attributes, and they can be given default values making it optional as to whether these are supplied, by changing the constructor method as follows:
class Agent():
def __init__(self, x = 0, y = 0):
self.x = x
self.y = y
In Python, all classes inherit (attributes and methods) from the 'type' metaclass. A 'metaclass' can be thought of as a factory for producing classes.
The classes from which a class inherits are known as 'super classes', and the classes that inherit from another class are known as 'subclasses'.
Inheritance is heirarchical and allows for things in common to multiple classes to be stored in a super class avoiding having to define attributes and methods repeatedly in subclasses. Those attributes and methods in a superclass are accessible directly from subclasses.
Methods can be 'overridden' by defining a method with the same name as the superclass method. The function super() can be used, to access an overriden method. In this way an overridden method can access the superclass method that it is overriding as shown in the example below.
Python supports 'multiple inheritance' where a subclass may inherit from multiple superclasses. To do this, the classes being inherited from are specified in the class definition, each separated by a comma.
The following defines 'Goat' and 'Wolf' as classes that inherit from the 'Agent' class:
class Agent():
def __init__(self, x, y):
self.x = x
self.y = y
class Goat(Agent):
def __init__(self, x, y):
super().__init__(x, y)
self.hungry = True
class Wolf(Agent):
def __init__(self, x, y, pack):
super().__init__(x, y)
self.pack = pack
The constructor methods of the Goat and Wolf classes call the superclass constructor method. Each instance of Goat would be instantiated with the attribute 'self.hungry' equal to 'True'. Each instance of Wolf is given a parameter called 'pack' when constructed. This is a variable which is assigned to the attribute with the same name. In this example, 'pack' could be a identifier or a list or something else.
Inheritance allows for a class (call it 'A') to be extended to create different subclasses, and a further class to be defined as the subclass of multiple of the subclasses of A. In such a case, there is an order in the way inheritence works and how methods get overridden. For more details of this see: The Python Tutorial on Classes Section on Multiple Inheritance.
Python does not have an direct way to declare the access level or control access to variables or functions/methods as is common in other languages.
Instead there is a naming convention which is that variables that have a name starting with an underscore and functions starting with a double underscore are not to be accessed directly from outside the class or module.
The function property() can be used with a specifically named get method to provide indirect access to a object attribute. This can optionally allow for setting the attribute value and deleting the attribute too. In the following example the function property() is set up to pass, modify and delete the attribute _x:
class Agent():
def __init__(self, ax, ay):
self._x = ax
self._y = ay
def getx(self):
return self._x
def setx(self, value):
self._x = value
def delx(self):
del self._x
x = property(getx, setx, delx, "I'm the 'x' property.")
This could be used from another module as follows:
import agents:
a = agent.Agent(0, 0):
a.x = 3
print(a.x) # <-- Prints 3
So, rather than access the attribute _x directly, this is done via a function which is aliased as x.
Decorators can alternatively be used to achieve the same thing as follows:
class Agent():
def __init__(self, ax, ay):
self._x = ax
self._y = ay
@property
def x(self):
"""I'm the 'x' property."""
return self._x
@x.setter
def x(self, value):
self._x = value
@x.deleter
def x(self):
del self._x
By default, when printing an object using the print function, the memory address of the object is printed, but rarely is this useful. The '__str__' method can be overriden to provide a more useful string representation of an object, for example:
class Agent():
def __init__(self, x, y):
self.x = x
self.y = y
def __str__(self):
return "x=" + self.str(x) + ", y=" + self.str(y)
class Goat(Agent):
def __init__(self, x, y):
super().__init__(x, y)
self.hungry = True
def __str__(self):
return super().__str__() + ", hungry=" + self.str(hungry)
More details on customisation can be found in The Python 3 Language Reference Data Model Chapter Section on Basic Customization.
For a more in depth consideration of classes, see: The Python Tutorial on Classes.