Skip to content

Instantly share code, notes, and snippets.

@dmmeteo
Last active October 28, 2024 15:00
Show Gist options
  • Save dmmeteo/f630fa04c7a79d3c132b9e9e5d037bfd to your computer and use it in GitHub Desktop.
Save dmmeteo/f630fa04c7a79d3c132b9e9e5d037bfd to your computer and use it in GitHub Desktop.
SOLID Principles explained in Python with examples.
"""
Single Responsibility Principle
“…You had one job” — Loki to Skurge in Thor: Ragnarok
A class should have only one job.
If a class has more than one responsibility, it becomes coupled.
A change to one responsibility results to modification of the other responsibility.
"""
class Animal:
def __init__(self, name: str):
self.name = name
def get_name(self) -> str:
pass
def save(self, animal: Animal):
pass
"""
The Animal class violates the SRP.
How does it violate SRP?
SRP states that classes should have one responsibility, here, we can draw out two responsibilities: animal database management and animal properties management.
The constructor and get_name manage the Animal properties while the save manages the Animal storage on a database.
How will this design cause issues in the future?
If the application changes in a way that it affects database management functions. The classes that make use of Animal properties will have to be touched and recompiled to compensate for the new changes.
You see this system smells of rigidity, it’s like a domino effect, touch one card it affects all other cards in line.
To make this conform to SRP, we create another class that will handle the sole responsibility of storing an animal to a database:
"""
class Animal:
def __init__(self, name: str):
self.name = name
def get_name(self):
pass
class AnimalDB:
def get_animal(self) -> Animal:
pass
def save(self, animal: Animal):
pass
"""
When designing our classes, we should aim to put related features together,
so whenever they tend to change they change for the same reason.
And we should try to separate features if they will change for different reasons. - Steve Fenton
"""
Open-Closed Principle
Software entities(Classes, modules, functions) should be open for extension, not modification.
"""
class Animal:
def __init__(self, name: str):
self.name = name
def get_name(self) -> str:
pass
animals = [
Animal('lion'),
Animal('mouse')
]
def animal_sound(animals: list):
for animal in animals:
if animal.name == 'lion':
print('roar')
elif animal.name == 'mouse':
print('squeak')
animal_sound(animals)
"""
The function animal_sound does not conform to the open-closed principle because it cannot be closed against new kinds of animals.
If we add a new animal, Snake, We have to modify the animal_sound function.
You see, for every new animal, a new logic is added to the animal_sound function.
This is quite a simple example. When your application grows and becomes complex,
you will see that the if statement would be repeated over and over again
in the animal_sound function each time a new animal is added, all over the application.
"""
animals = [
Animal('lion'),
Animal('mouse'),
Animal('snake')
]
def animal_sound(animals: list):
for animal in animals:
if animal.name == 'lion':
print('roar')
elif animal.name == 'mouse':
print('squeak')
elif animal.name == 'snake':
print('hiss')
animal_sound(animals)
"""
How do we make it (the animal_sound) conform to OCP?
"""
class Animal:
def __init__(self, name: str):
self.name = name
def get_name(self) -> str:
pass
def make_sound(self):
pass
class Lion(Animal):
def make_sound(self):
return 'roar'
class Mouse(Animal):
def make_sound(self):
return 'squeak'
class Snake(Animal):
def make_sound(self):
return 'hiss'
def animal_sound(animals: list):
for animal in animals:
print(animal.make_sound())
animal_sound(animals)
"""
Animal now has a virtual method make_sound. We have each animal extend the Animal class and implement the virtual make_sound method.
Every animal adds its own implementation on how it makes a sound in the make_sound.
The animal_sound iterates through the array of animal and just calls its make_sound method.
Now, if we add a new animal, animal_sound doesn’t need to change.
All we need to do is add the new animal to the animal array.
animal_sound now conforms to the OCP principle.
"""
"""
Another example:
Let’s imagine you have a store, and you give a discount of 20% to your favorite customers using this class:
When you decide to offer double the 20% discount to VIP customers. You may modify the class like this:
"""
class Discount:
def __init__(self, customer, price):
self.customer = customer
self.price = price
def give_discount(self):
if self.customer == 'fav':
return self.price * 0.2
if self.customer == 'vip':
return self.price * 0.4
"""
No, this fails the OCP principle. OCP forbids it. If we want to give a new percent discount maybe, to a diff.
type of customers, you will see that a new logic will be added.
To make it follow the OCP principle, we will add a new class that will extend the Discount.
In this new class, we would implement its new behavior:
"""
class Discount:
def __init__(self, customer, price):
self.customer = customer
self.price = price
def get_discount(self):
return self.price * 0.2
class VIPDiscount(Discount):
def get_discount(self):
return super().get_discount() * 2
"""
If you decide 80% discount to super VIP customers, it should be like this:
You see, extension without modification.
"""
class SuperVIPDiscount(VIPDiscount):
def get_discount(self):
return super().get_discount() * 2
"""
Liskov Substitution Principle
A sub-class must be substitutable for its super-class.
The aim of this principle is to ascertain that a sub-class can assume the place of its super-class without errors.
If the code finds itself checking the type of class then, it must have violated this principle.
Let’s use our Animal example.
"""
def animal_leg_count(animals: list):
for animal in animals:
if isinstance(animal, Lion):
print(lion_leg_count(animal))
elif isinstance(animal, Mouse):
print(mouse_leg_count(animal))
elif isinstance(animal, Pigeon):
print(pigeon_leg_count(animal))
animal_leg_count(animals)
"""
To make this function follow the LSP principle, we will follow this LSP requirements postulated by Steve Fenton:
If the super-class (Animal) has a method that accepts a super-class type (Animal) parameter.
Its sub-class(Pigeon) should accept as argument a super-class type (Animal type) or sub-class type(Pigeon type).
If the super-class returns a super-class type (Animal).
Its sub-class should return a super-class type (Animal type) or sub-class type(Pigeon).
Now, we can re-implement animal_leg_count function:
"""
def animal_leg_count(animals: list):
for animal in animals:
print(animal.leg_count())
animal_leg_count(animals)
"""
The animal_leg_count function cares less the type of Animal passed, it just calls the leg_count method.
All it knows is that the parameter must be of an Animal type, either the Animal class or its sub-class.
The Animal class now have to implement/define a leg_count method.
And its sub-classes have to implement the leg_count method:
"""
class Animal:
def leg_count(self):
pass
class Lion(Animal):
def leg_count(self):
pass
"""
When it’s passed to the animal_leg_count function, it returns the number of legs a lion has.
You see, the animal_leg_count doesn’t need to know the type of Animal to return its leg count,
it just calls the leg_count method of the Animal type because by contract a sub-class of Animal class must implement the leg_count function.
"""
"""
Interface Segregation Principle
Make fine grained interfaces that are client specific
Clients should not be forced to depend upon interfaces that they do not use.
This principle deals with the disadvantages of implementing big interfaces.
Let’s look at the below IShape interface:
"""
class IShape:
def draw_square(self):
raise NotImplementedError
def draw_rectangle(self):
raise NotImplementedError
def draw_circle(self):
raise NotImplementedError
"""
This interface draws squares, circles, rectangles. class Circle, Square or Rectangle implementing the IShape
interface must define the methods draw_square(), draw_rectangle(), draw_circle().
"""
class Circle(IShape):
def draw_square(self):
pass
def draw_rectangle(self):
pass
def draw_circle(self):
pass
class Square(IShape):
def draw_square(self):
pass
def draw_rectangle(self):
pass
def draw_circle(self):
pass
class Rectangle(IShape):
def draw_square(self):
pass
def draw_rectangle(self):
pass
def draw_circle(self):
pass
"""
It’s quite funny looking at the code above. class Rectangle implements methods (draw_circle and draw_square) it has no use of,
likewise Square implementing draw_circle, and draw_rectangle, and class Circle (draw_square, draw_rectangle).
If we add another method to the IShape interface, like draw_triangle(),
"""
class IShape:
def draw_square(self):
raise NotImplementedError
def draw_rectangle(self):
raise NotImplementedError
def draw_circle(self):
raise NotImplementedError
def draw_triangle(self):
raise NotImplementedError
"""
the classes must implement the new method or error will be thrown.
We see that it is impossible to implement a shape that can draw a circle but not a rectangle or a square or a triangle.
We can just implement the methods to throw an error that shows the operation cannot be performed.
ISP frowns against the design of this IShape interface. clients (here Rectangle, Circle, and Square) should not be forced to depend on methods that they do not need or use.
Also, ISP states that interfaces should perform only one job (just like the SRP principle) any extra grouping of behavior should be abstracted away to another interface.
Here, our IShape interface performs actions that should be handled independently by other interfaces.
To make our IShape interface conform to the ISP principle, we segregate the actions to different interfaces.
Classes (Circle, Rectangle, Square, Triangle, etc) can just inherit from the IShape interface and implement their own draw behavior.
"""
class IShape:
def draw(self):
raise NotImplementedError
class Circle(IShape):
def draw(self):
pass
class Square(IShape):
def draw(self):
pass
class Rectangle(IShape):
def draw(self):
pass
"""
We can then use the I -interfaces to create Shape specifics like Semi Circle, Right-Angled Triangle, Equilateral Triangle, Blunt-Edged Rectangle, etc.
"""
"""
Dependency Inversion Principle
Dependency should be on abstractions not concretions
A. High-level modules should not depend upon low-level modules. Both should depend upon abstractions.
B. Abstractions should not depend on details. Details should depend upon abstractions.
There comes a point in software development where our app will be largely composed of modules.
When this happens, we have to clear things up by using dependency injection.
High-level components depending on low-level components to function.
"""
class XMLHttpService(XMLHttpRequestService):
pass
class Http:
def __init__(self, xml_http_service: XMLHttpService):
self.xml_http_service = xml_http_service
def get(self, url: str, options: dict):
self.xml_http_service.request(url, 'GET')
def post(self, url, options: dict):
self.xml_http_service.request(url, 'POST')
"""
Here, Http is the high-level component whereas HttpService is the low-level component.
This design violates DIP A: High-level modules should not depend on low-level level modules. It should depend upon its abstraction.
This Http class is forced to depend upon the XMLHttpService class.
If we were to change to change the Http connection service, maybe we want to connect to the internet through cURL or even Mock the http service.
We will painstakingly have to move through all the instances of Http to edit the code and this violates the OCP principle.
The Http class should care less the type of Http service you are using. We make a Connection interface:
"""
class Connection:
def request(self, url: str, options: dict):
raise NotImplementedError
"""
The Connection interface has a request method. With this, we pass in an argument of type Connection to our Http class:
"""
class Http:
def __init__(self, http_connection: Connection):
self.http_connection = http_connection
def get(self, url: str, options: dict):
self.http_connection.request(url, 'GET')
def post(self, url, options: dict):
self.http_connection.request(url, 'POST')
"""
So now, no matter the type of Http connection service passed to Http it can easily connect to a network
without bothering to know the type of network connection.
We can now re-implement our XMLHttpService class to implement the Connection interface:
"""
class XMLHttpService(Connection):
xhr = XMLHttpRequest()
def request(self, url: str, options:dict):
self.xhr.open()
self.xhr.send()
"""
We can create many Http Connection types and pass it to our Http class without any fuss about errors.
"""
class NodeHttpService(Connection):
def request(self, url: str, options:dict):
pass
class MockHttpService(Connection):
def request(self, url: str, options:dict):
pass
"""
Now, we can see that both high-level modules and low-level modules depend on abstractions.
Http class(high level module) depends on the Connection interface(abstraction) and
the Http service types(low level modules) in turn, depends on the Connection interface(abstraction).
Also, this DIP will force us not to violate the Liskov Substitution Principle:
The Connection types Node-XML-MockHttpService are substitutable for their parent type Connection.
"""
@amarynets
Copy link

Thank you for this gist.
One comment related to DI: https://gist.github.com/dmmeteo/f630fa04c7a79d3c132b9e9e5d037bfd#file-5-dip-py-L37
We have def request(self, url: str, options: dict): and when we call self.http_connection.request(url, 'POST'). The code is correct from the interpreter, but a bit confusing from type annotation and variable name. It pass 'POST' as a string, but we hope that there will be dict and the last one - the name of the variable in base class defined as options but we pass HTTP method.
Maybe it would be better to change the base signature to
def request(self, url: str, method: str, options: dict):

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment