Classes¶
Almost everything in Python is an object. All objects (e.g., variables, functions) have a type. An object's type is also known as its class. A class can also be thought of as a template for creating an object of that type.
Objects can combine holding data with functionality. Anything that you can access in an object is known as an attribute. Attributes can either be data attributes (sometime also known as class properties, class members or class variables) or function attributes (also known as class methods). In Python all of a class' attributes are public, i.e., you can access them from an instance of a class.
To add to this, data attributes can be of two types:
- class attributes - these belong to the class itself and are the same value for every instance of the class (and can even be used without creating a class instance),
- instance attributes - these belong to a specific class instance and may be different (based on user input) between instances.
A new class can be defined using the
class keyword.
Note
You can write perfectly good codes in Python without ever having to define your own object's classes. However, using classes is part of the Object Oriented paradigm and in many cases it makes intuitive and practical sense to bundle certain bits of data and certain functions together, i.e., define a class.
A simple class¶
A simple class containing just class attributes can be created using:
class Electron:
# indents are required to define thing within a class
name = "electron"
charge = -1.6e-19 # electric charge in coulombs
mass = 9.1e-31 # mass in kg
Variables can then be created from this class (also known as class instances) with:
e = Electron() # note the brackets are required
The class attributes of the object can be accessed using a . followed by their name, e.g.,:
print(e.mass)
9.1e-31
Note
Even though class attributes are public (i.e., viewable and changeable by the user), it is not good practice to change them in any class instance. If you want to set data attributes that can be altered you should use instance attributes that are set when initialising the class, or through a class method, or using properties (setters are not covered in this course).
Class attributes can also be accessed directly from the class rather than an instance, e.g.,:
print(Electron.mass)
9.1e-31
A class with initialisation¶
The above class has been created with a fixed set of values using class attributes. Every time an instance of the class is created the attributes will be the same.
It is often useful to be able to create an instance of a class with user defined values rather than
fixed values. Python classes can have a special method called __init__ that defines how the class
is initialised and can take in user supplied arguments. In OOP languages like C++ the __init__
method is equivalent to what is called a
constructor.
class Particle:
def __init__(self, name, charge, mass):
# the initialisation function
self.name = name
self.charge = charge
self.mass = mass
Now, Particle class instances can be created for different particles by supplying their values,
e.g.,:
electron = Particle("electron", -1.6e-19, 9.1e-31)
proton = Particle("proton", 1.6e-19, 1.7e-27)
for p in [electron, proton]:
print(f"{p.name} mass = {p.mass}")
electron mass = 9.1e-31
proton mass = 1.7e-27
The __init__ method of a class can have positional arguments and/or keyword arguments,
just like any other function. Using keyword arguments allows the setting of default initialisation
values if no user supplied values are given, e.g.,
class Particle:
def __init__(self, name="electron", charge=-1.6e-19, mass=9.1e-31):
# if no values are supplied the default values will be used
self.name = name
self.charge = charge
self.mass = mass
myparticle = Particle()
print(myparticle.name)
electron
self¶
The __init__ method, and any other regular methods defined in a class, takes self as its first
argument. But when creating the new object above nothing was passed for self, i.e., the brackets
were empty! When defining a class method the method has to have the class instance explicitly
passed to it. This allows that method to access all the attributes of the current class instance via
self. But, when using a method self is passed implicitly, i.e., the user does not have to supply
it as it is supplied automatically.
Note
The usage of the word "self" is just convention. In reality any word can be used in place
of self as long as it is consistently used throughout the class, e.g.,
class whatsit:
def __init__(cheesy, name="Blah"): # using cheesy instead of self!
cheesy.name = name
def show_name(cheesy):
print(f"My name is {cheesy.name}")
It is recommended to stick to using self!
Adding methods¶
Methods are defined just like standard functions, but within the class definition. The first
argument that they take must be self, so that all other class attributes are available within it.
class Particle:
def __init__(self, name, charge, mass, spin=None):
# the initialisation function
self.name = name
self.charge = charge
self.mass = mass
self.spin = spin
def fermion_or_boson(self):
"""
Determine if the particle is a fermion or a boson.
"""
# note: the method is able to use the instance's "spin" attribute via self
if self.spin is not None:
if self.spin % 0.5 == 0.0:
return "fermion"
elif self.spin % 1.0 == 0.0:
return "boson"
else:
print("Particle is neither a fermion or a boson")
return None
else:
return None
Like the data attributes, this method can then be accessed using a . followed by the method name,
e.g.,:
electron = Particle("electron", -1.6e-19, 9.1e-31, spin=1/2)
print(electron.fermion_or_boson()) # note the brackets when accessing the method
fermion
The fermion_or_boson method does not take in any arguments (other than the implicit self). A
method that could be added to the particle is one that calculates the Lorentz
force on the particle in an electric and magnetic
field:
class Particle:
def __init__(self, name, charge, mass, spin=None):
# the initialisation function
self.name = name
self.charge = charge
self.mass = mass
self.spin = spin
def fermion_or_boson(self):
"""
Determine if the particle is a fermion or a boson.
"""
if self.spin is not None:
if self.spin % 0.5 == 0.0:
return "fermion"
elif self.spin % 1.0 == 0.0:
return "boson"
else:
print("Particle is neither a fermion or a boson")
return None
else:
return None
def lorentz_force(self, E, B=[0.0, 0.0, 0.0], v=[0.0, 0.0, 0.0]):
"""
Calculate the Lorentz force on the particle.
Parameters
----------
E: array
A vector giving the x, y and z components of the electric field
(required).
B: array
A vector giving the x, y and z components of the magnetic field
(defaults to zero).
v: array
A vector giving the x, y and z components of the particle's
velocity (defaults to zero)
Returns
-------
F: list
The 3d Lorentz force vector.
"""
if len(E) != 3:
# check E is the right length
raise ValueError("E is not the right length")
F = 3 * [0.0] # initialise F as list of 3 zeros
# calculate F = q * (E + v x B)
F[0] = self.charge * (E[0] + v[1] * B[2] - v[2] * B[1])
F[1] = self.charge * (E[1] + v[2] * B[0] - v[0] * B[2])
F[2] = self.charge * (E[2] + v[0] * B[1] - v[1] * B[0])
return F
electron = Particle("electron", -1.6e-19, 9.1e-31)
First, we can see what happens if we used the method incorrectly (error checking is discussed in
more detail in the "Error checking and debugging" tutorial).
The lorentz_force method requires one positional argument, but if we do not supply one then:
# calculate the Lorentz force (forgetting the required positional argument!)
F = electron.lorentz_force()
---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
<ipython-input-23-14143ee08c6d> in <module>
----> 1 F = electron.lorentz_force()
TypeError: lorentz_force() missing 1 required positional argument: 'E'
The positional argument must be a list containing three values, but if we try and use two, then:
# try again (with E being the wrong length!)
F = electron.lorentz_force([0.1, 0.2])
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
<ipython-input-24-966d81233b8b> in <module>
----> 1 F = electron.lorentz_force([0.1, 0.2])
<ipython-input-20-98635895214d> in lorentz_force(self, E, B, v)
42 if len(E) != 3:
43 # check E is the right length
---> 44 raise ValueError("E is not the right length")
45
46 F = 3 * [0.0] # initialise F as zeros
ValueError: E is not the right length
Finally, we will use it correctly:
# try again!
F = electron.lorentz_force([0.1, 0.2, 0.3])
print(F)
[-1.6e-20, -3.2e-20, -4.8e-20]
The lorentz_force method above takes one positional argument and two keyword arguments (that give
default values). Any number of positional or keyword arguments could be used.
The lorentz_force method could be simplified using NumPy (see the NumPy
tutorial.
Special methods¶
There are a set of special method names (sometimes
called "dunder" methods as they start and end with a double underscore, but also known as magic
methods) like __init__ that can be defined in a class. These can be used:
- to allow comparisons of objects of specific class;
- define how mathematical operators work on a class (see Operator overloading);
- access attributes within a class;
- provide representations of class.
The full set of special methods can be found here.
One particular special method that we will use in this course is
__str__.
This defines how to display a class instance as a string, for example, if trying to print the
object:
class Particle:
def __init__(self, name="electron", charge=-1.6e-19, mass=9.1e-31):
# if no values are supplied the default values will be used
self.name = name
self.charge = charge
self.mass = mass
def __str__(self):
# a string representing the object
vowel = self.name[0].lower() in ["a", "e", "i", "o", "u"]
firstword = "An" if vowel else "A" # shorthand if else statement!
return f"{firstword} {self.name} with mass of {self.mass} kg and charge of {self.charge} C"
myparticle = Particle(name="positron", charge=1.6e-19)
print(myparticle)
A positron with mass of 9.1e-31 kg and charge of 1.6e-19 C
A similar method that can be used instead is __repr__.
Static methods¶
Static methods are functions within a class that can be used without creating a new instance of that class. As they do not use an instance of the class they do not have access to any of the other class attributes, i.e., they are standalone and must be supplied with all the variables they require.
Unlike normal methods they do not get passed the self argument. To make a method static you use
the @staticmethod
decorator on the line above the method definition, e.g.,:
class Line2D:
"""
A class defining a line in 2d coordinates.
Parameters
----------
point1: tuple
A tuple containing the (x, y) coordinates of one end of the line.
point2: tuple
A tuple containing the (x, y) coordinates of the other end of the line.
"""
def __init__(self, point1, point2):
self.point1 = point1
self.point2 = point2
# set the gradient and y-intercept of the line using the static method
m, c = Line2D.coefficients(self.point1, self.point2)
self.grad = m
self.yintercept = c
@staticmethod
def coefficients(point1, point2):
"""
Get the coefficients of the linear equation:
y = C_1 x + C_0
defined by the two points, where C_1 defines the gradient and C_0 defines
the y-intercept.
"""
# Note that self is not passed to the method.
# check lengths
if len(point1) != 2 or len(point2) != 2:
raise ValueError("Points on the line must be 2D")
# get differences
dx = point2[0] - point1[0]
dy = point2[1] - point1[1]
C1 = dy / dx
C0 = point2[1] - C1 * point2[0]
return C1, C0
To get the linear equation coefficients we could then do:
point1 = (8, 10)
point2 = (10, 12)
# use staticmethod without creating an instance of Line2D
m, c = Line2D.coefficients(point1, point2)
print(f"Gradient is {m}, y-intercept is {c}")
Gradient is 1.0, y-intercept is 2.0
The static method can also be used by an instance of the class, e.g.,
point1 = (-9, 10)
point2 = (-2, 5)
# create a line
line = Line2D(point1, point2)
print(line.grad, line.yintercept)
-0.7142857142857143 3.571428571428571
# work out gradient and y-intercept of a new line
mnew, cnew = line.coefficients((1, 2), (2, 3))
print(mnew, cnew)
1.0 1.0
# because it's a static method it hasn't altered anything in "line"
print(line.point1, line.point2)
print(line.grad, line.yintercept)
(-9, 10) (-2, 5)
-0.7142857142857143 3.571428571428571
Class inheritance¶
You may want to define a new class that is very similar to an already existing class, but adds new attributes. Rather than redefining all of the aspects of the existing class in the new class you can inherit them from the existing class.
Suppose we have a Galaxy class:
class Galaxy:
"""
A class defining a galaxy.
Parameters
----------
mass: float
The galaxy mass (in solar masses)
distance: float
The distance (in Mpc)
type: str
The type of galaxy, e.g., "spiral"
name: str
The galaxy's name. Defaults to None.
"""
def __init__(self, mass, distance, type, name=None):
self.mass = mass
self.distance = distance
self.type = type
self.name = name
def redshift(self, H0=70.0):
"""
Calculate the redshift using Hubble's law.
Parameters
----------
H0: float
Hubble's constant (defaults to 70 km/s/Mpc)
"""
# recession velocity
v = H0 * self.distance
return v / 3e5 # velocity / speed of light (km/s)
Now, suppose we want a class specifically for a spiral galaxy, but that keeps the attributes of a
Galaxy, i.e., Galaxy is the parent class (or superclass) and SpiralGalaxy will be its
child (or subclass). We can create a new class with:
class SpiralGalaxy(Galaxy): # this is where the Galaxy gets inherited
"""
A class defining a spiral galaxy.
"""
def __init__(self, bulge_mass, disc_mass, halo_mass, distance, name=None, barred=False):
# the special "super" function allows initialisation of the common
# Galaxy attributes
super().__init__(bulge_mass + halo_mass + disc_mass, distance, "spiral", name=name)
# add spiral specific properties
self.bulge_mass = bulge_mass
self.disc_mass = disc_mass
self.halo_mass = halo_mass
self.barred = barred # has it got a bar
def disc_circular_velocity(self, Rd, r):
""""
Calculate the contribution to the circular velocity of the disc (see Eqn. 1
of astro-ph/9909252).
Parameters
----------
Rd: float
The disc scale-length (kpc)
r: array_like
A set of positive radial values at which to calculate the velocity (kpc)
Returns
-------
velocity: array_like
A set of circular velocity values (km/s).
"""
# import modified Bessel functions from scipy
from scipy.special import iv, kn
from math import sqrt
disc_mass_si = self.disc_mass * 1.99e30 # disc mass in kg
velocities = [] # list to hold velocities
for rval in r:
x = rval / Rd
B = iv(0, x / 2) * kn(0, x / 2) - iv(1, x / 2) * kn(1, x / 2)
G = 6.67e-11 # Newton's gravitational constant
v = sqrt(0.5 * G * (disc_mass_si / (Rd * 3.086e19)) * x ** 2 * B)
velocities.append(v / 1e3) # convert to km/s
return velocities
If we create a SpiralGalaxy:
bulge_mass = 3.0e8 # solar masses
disc_mass = 6.0e9
halo_mass = 5.0e10
distance = 0.84 # Mpc
m33 = SpiralGalaxy(bulge_mass, disc_mass, halo_mass, distance, name="M33")
we can then use attributes from the Galaxy class like:
z = m33.redshift()
print(f"{m33.name}'s redshift is {z}")
M33's redshift is 0.000196
or use the new attributes:
rs = list(range(1, 16)) # range of distances
Rd = 1.2 # disk scale in kpc
vs = m33.disc_circular_velocity(Rd, rs)
from matplotlib import pyplot as plt
plt.plot(rs, vs)
plt.xlabel("Distance from Galactic centre (kpc)")
plt.ylabel("Circular velocity (km/s)")
plt.show()
Note
Just for correctness, I should note that M33's actual observed redshift is not 0.000196! That's what its cosmological redshift would be, but it's in our local group of galaxies and so is subject to local gravitational accelerations. Therefore its redshift is dominated by that local motion with respect to the Milky Way.
Operator overloading¶
Mathematical operators (e.g., +, -, *, /) can be applied to number classes like ints or
floats. But, if it's sensible to do so, you can define how the mathematical operators act on any
class you define.
If you want to be able to define a way to, for example, add two instances of a particular object you
can used the special __add__
method. The are equivalent methods for the other mathematical
operators and logical
expressions.
Suppose you had a class representing a vector in 3D Cartesian coordinates:
class Vector:
"""
A class representing a vector in 3D Cartesian coordinates.
Parameters
----------
x: int, float
The component of the vector in the x-dimension
y: int, float
The component of the vector in the y-dimension
z: int, float
The component of the vector in the z-dimension
"""
def __init__(self, x, y, z):
self.vector = [] # store vector as a list
for v in [x, y, z]:
# test that we've given the class numbers
if not isinstance(v, (int, float)):
raise ValueError("Vector can only contain numbers")
self.vector.append(v)
@property
def x(self):
return self.vector[0]
@property
def y(self):
return self.vector[1]
@property
def z(self):
return self.vector[2]
def norm(self):
"""
Return the length of the vector.
"""
from math import sqrt
return sqrt(sum([v ** 2 for v in self.vector]))
def unit(self):
"""
Return the unit vector.
"""
norm = self.norm()
return [v / norm for v in self.vector]
def __len__(self):
# the length of the vector (__len__ is another magic method for
# returning the "length" of a class, if applicable)
return len(self.vector)
Note
In the above definition it has used the @property function decorator. This is a way to
define methods in a class that can allow aspects of a current data attributes to be accessed
with a different name. Here it is handy to store the vector as a list-type data attribute,
but it is also nice to be able to access the individual components in an intuitively named
manner. Hence defining properties for x, y and z that only return those components.
These attributes can be accessed with, e.g.,:
v1 = Vector(1, 2, 3)
print(v1.x) # despite being defined as a function the x property can be access without using ()
1
It would be useful to be able to add two of these vectors together and return a new vector. But, trying this leads to:
v1 = Vector(1, 2, 3)
v2 = Vector(4, 5, 6)
v3 = v1 + v2 # try adding the vector
---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
<ipython-input-7-13b475227a38> in <module>
----> 1 v3 = v1 + v2
TypeError: unsupported operand type(s) for +: 'Vector' and 'Vector'
However, the __add__ special method can be used to define how to do standard vector addition
(i.e., adding each component separately), e.g.,:
class Vector:
"""
A class representing a vector in 3D Cartesian coordinates.
Parameters
----------
x: int, float
The component of the vector in the x-dimension
y: int, float
The component of the vector in the y-dimension
z: int, float
The component of the vector in the z-dimension
"""
def __init__(self, x, y, z):
self.vector = [] # store vector as a list
for v in [x, y, z]:
# test that we've given the class numbers
if not isinstance(v, (int, float)):
raise ValueError("Vector can only contain numbers")
self.vector.append(v)
@property
def x(self):
return self.vector[0]
@property
def y(self):
return self.vector[1]
@property
def z(self):
return self.vector[2]
def norm(self):
"""
Return the length of the vector.
"""
from math import sqrt
return math.sqrt(sum([v ** 2 for v in self.vector]))
def unit(self):
"""
Return the unit vector.
"""
norm = self.norm()
return [v / norm for v in self.vector]
def __len__(self):
# the length of the vector
return len(self.vector)
def __add__(self, other):
# use the argument "other" to represent the other vector to be added
# check that other is actually also a vector
if not isinstance(other, Vector):
raise TypeError("Can only add two Vectors")
# return a new Vector object
return Vector(self.x + other.x, self.y + other.y, self.z + other.z)
then we could do:
v1 = Vector(1, 2, 3)
v2 = Vector(4, 5, 6)
v3 = v1 + v2
print(v3.vector)
[5, 7, 9]
Note
In reality, for something like a vector, the NumPy library already has useful classes called arrays for which the mathematical operators are all defined.
If you wanted to use the += operator, e.g., to change v1 in-place
v1 += v2
you would also have to define the
__iadd__ function.
Logical operators and comparison operation can also be overloaded.