Defining objects: classes

Defining a class in Python involves giving a class name and a base class. By default, we’ll use object as our base class—it’s the default choice.

Here’s a completely empty class. The pass keyword does nothing; we need it because you can’t have an empty declaration.

The format for the class header is: class NAME(BASE), where NAME is the name of the class we’re defining and BASE is the name of the base class. For now BASE should always be object, and NAME should always be written in CamelCase, i.e., each word’s first letter is capitalized, but the rest is lowercase.

The body of a class is always indented a level in—notice that pass is indented.

By default, Empty will have a constructor that does nothing. We can create new instances of Empty, and we can observe that the constructor returns a new object each time. But… that’s about it. To make things more interesting, we should define a constructor. To do so, we define a method called __init__.

There’s quite a bit to unpack here. To define the __init__ method, we indented a level and used def, just like if we were defining a function. But __init__ also takes an argument, self. Which is pretty weird, considering that the call to the constructor doesn’t take any arguments!

The key to object-oriented programming is understanding self-reference, here using the self keyword. (Other languages use this, among other things.) Every method in a class takes an extra argument, self. Since every method can see self, it means that the code and data are grouped together.

Here we don’t have any data stored in self yet, but we can observe that the self inside the constructor is the same thing that the constructor returns. Here’s a run (you may get different id’s):

I'm being constructed! My id is 140598014959136.
140598014959136
I'm being constructed! My id is 140598014957360.
140598014957360

Notice that the id printed out inside of the method is the same as the one immediately returned.

__init__ is a weird name

In Python, lots of special things have the form __NAME__. We’ll learn a few more of these special names today, and we’ll meet many more in the last week of the course.

In the name of honesty, I should say that __init__ isn’t exactly like a constructor in, say, Java. If we’re going to be precise, __init__ is an initializer. But most developers go ahead and call __init__ a constructor. Python’s object construction works quite differently from in Java, and while it has something that you’d call a “constructor” like Java’s, it’s not usually necessary to think about it. None of this is very important, but it’s good to have in the back of your mind that the waters here are a little deeper than we’re making it seem.