[python] Why do we use __init__ in Python classes?

I am having trouble understanding the Initialization of classes.

What's the point of them and how do we know what to include in them? Does writing in classes require a different type of thinking versus creating functions (I figured I could just create functions and then just wrap them in a class so I can re-use them. Will that work?)

Here's an example:

class crawler:
  # Initialize the crawler with the name of database
  def __init__(self,dbname):
    self.con=sqlite.connect(dbname)

  def __del__(self):
    self.con.close()

  def dbcommit(self):
    self.con.commit()

Or another code sample:

class bicluster:
  def __init__(self,vec,left=None,right=None,distance=0.0,id=None):
    self.left=left
    self.right=right
    self.vec=vec
    self.id=id
    self.distance=distance

There are so many classes with __init__ I come across when trying to read other people's code, but I don't understand the logic in creating them.

The __init__ function is setting up all the member variables in the class. So once your bicluster is created you can access the member and get a value back:

mycluster = bicluster(...actual values go here...)
mycluster.left # returns the value passed in as 'left'

Check out the Python Docs for some info. You'll want to pick up an book on OO concepts to continue learning.

To contribute my 5 cents to the thorough explanation from Amadan.

Where classes are a description "of a type" in an abstract way. Objects are their realizations: the living breathing thing. In the object-orientated world there are principal ideas you can almost call the essence of everything. They are:

1. encapsulation (won't elaborate on this)
2. inheritance
3. polymorphism

Objects have one, or more characteristics (= Attributes) and behaviors (= Methods). The behavior mostly depends on the characteristics. Classes define what the behavior should accomplish in a general way, but as long as the class is not realized (instantiated) as an object it remains an abstract concept of a possibility. Let me illustrate with the help of "inheritance" and "polymorphism".

    class Human:
        gender
        nationality
        favorite_drink
        core_characteristic
        favorite_beverage
        name
        age

        def love    
        def drink
        def laugh
        def do_your_special_thing                

    class Americans(Humans)
        def drink(beverage):
            if beverage != favorite_drink: print "You call that a drink?"
            else: print "Great!" 

    class French(Humans)
        def drink(beverage, cheese):
            if beverage == favourite_drink and cheese == None: print "No cheese?" 
            elif beverage != favourite_drink and cheese == None: print "Révolution!"

    class Brazilian(Humans)
        def do_your_special_thing
            win_every_football_world_cup()

    class Germans(Humans)
        def drink(beverage):
            if favorite_drink != beverage: print "I need more beer"
            else: print "Lecker!" 

    class HighSchoolStudent(Americans):
        def __init__(self, name, age):
             self.name = name
             self.age = age

jeff = HighSchoolStudent(name, age):
hans = Germans()
ronaldo = Brazilian()
amelie = French()

for friends in [jeff, hans, ronaldo]:
    friends.laugh()
    friends.drink("cola")
    friends.do_your_special_thing()

print amelie.love(jeff)
>>> True
print ronaldo.love(hans)
>>> False

Some characteristics define human beings. But every nationality differs somewhat. So "national-types" are kinda Humans with extras. "Americans" are a type of "Humans " and inherit some abstract characteristics and behavior from the human type (base-class) : that's inheritance. So all Humans can laugh and drink, therefore all child-classes can also! Inheritance (2).

But because they are all of the same kind (Type/base-class : Humans) you can exchange them sometimes: see the for-loop at the end. But they will expose an individual characteristic, and thats Polymorphism (3).

So each human has a favorite_drink, but every nationality tend towards a special kind of drink. If you subclass a nationality from the type of Humans you can overwrite the inherited behavior as I have demonstrated above with the drink() Method. But that's still at the class-level and because of this it's still a generalization.

hans = German(favorite_drink = "Cola")

instantiates the class German and I "changed" a default characteristic at the beginning. (But if you call hans.drink('Milk') he would still print "I need more beer" - an obvious bug ... or maybe that's what i would call a feature if i would be a Employee of a bigger Company. ;-)! )

The characteristic of a type e.g. Germans (hans) are usually defined through the constructor (in python : __init__) at the moment of the instantiation. This is the point where you define a class to become an object. You could say breath life into an abstract concept (class) by filling it with individual characteristics and becoming an object.

But because every object is an instance of a class they share all some basic characteristic-types and some behavior. This is a major advantage of the object-orientated concept.

To protect the characteristics of each object you encapsulate them - means you try to couple behavior and characteristic and make it hard to manipulate it from outside the object. That's Encapsulation (1)

It seems like you need to use __init__ in Python if you want to correctly initialize mutable attributes of your instances.

See the following example:

>>> class EvilTest(object):
...     attr = []
... 
>>> evil_test1 = EvilTest()
>>> evil_test2 = EvilTest()
>>> evil_test1.attr.append('strange')
>>> 
>>> print "This is evil:", evil_test1.attr, evil_test2.attr
This is evil: ['strange'] ['strange']
>>> 
>>> 
>>> class GoodTest(object):
...     def __init__(self):
...         self.attr = []
... 
>>> good_test1 = GoodTest()
>>> good_test2 = GoodTest()
>>> good_test1.attr.append('strange')
>>> 
>>> print "This is good:", good_test1.attr, good_test2.attr
This is good: ['strange'] []

This is quite different in Java where each attribute is automatically initialized with a new value:

import java.util.ArrayList;
import java.lang.String;

class SimpleTest
{
    public ArrayList<String> attr = new ArrayList<String>();
}

class Main
{
    public static void main(String [] args)
    {
        SimpleTest t1 = new SimpleTest();
        SimpleTest t2 = new SimpleTest();

        t1.attr.add("strange");

        System.out.println(t1.attr + " " + t2.attr);
    }
}

produces an output we intuitively expect:

[strange] []

But if you declare attr as static, it will act like Python:

[strange] [strange]

class Dog(object):

    # Class Object Attribute
    species = 'mammal'

    def __init__(self,breed,name):
        self.breed = breed
        self.name = name

In above example we use species as a global since it will be always same(Kind of constant you can say). when you call __init__ method then all the variable inside __init__ will be initiated(eg:breed,name).

class Dog(object):
    a = '12'

    def __init__(self,breed,name,a):
        self.breed = breed
        self.name = name
        self.a= a

if you print the above example by calling below like this

Dog.a
12

Dog('Lab','Sam','10')
Dog.a
10

That means it will be only initialized during object creation. so anything which you want to declare as constant make it as global and anything which changes use __init__

Classes are objects with attributes (state, characteristic) and methods (functions, capacities) that are specific for that object (like the white color and fly powers, respectively, for a duck).

When you create an instance of a class, you can give it some initial personality (state or character like the name and the color of her dress for a newborn). You do this with __init__.

Basically __init__ sets the instance characteristics automatically when you call instance = MyClass(some_individual_traits).

Following with your car example: when you get a car, you just don't get a random car, I mean, you choose the color, the brand, number of seats, etc. And some stuff is also "initialize" without you choosing for it, like number of wheels or registration number.

class Car:
    def __init__(self, color, brand, number_of_seats):
        self.color = color
        self.brand = brand
        self.number_of_seats = number_of_seats
        self.number_of_wheels = 4
        self.registration_number = GenerateRegistrationNumber()

So, in the __init__ method you defining the attributes of the instance you're creating. So, if we want a blue Renault car, for 2 people, we would initialize or instance of Car like:

my_car = Car('blue', 'Renault', 2)

This way, we are creating an instance of the Car class. The __init__ is the one that is handling our specific attributes (like color or brand) and its generating the other attributes, like registration_number.

• More about classes in Python
• More about the __init__ method

It is just to initialize the instance's variables.

E.g. create a crawler instance with a specific database name (from your example above).