(Easiest Guide) What Are Python Metaclasses?

by

A Python metaclass is a class that creates and controls other classes, similar to how classes create and control objects. It’s essentially a “class of a class”.

Metaclass Metaphor 1: Toy Factory

Imagine you’re running a toy factory. The blueprint for each type of toy (like a car, doll, or train) is like a class in Python. Each individual toy made from the blueprint is like an object or instance of that class.

Now, suppose you need a way to modify the blueprints themselves – maybe you want to ensure every toy you produce, no matter its type, comes with a ‘Made in Toy Factory’ stamp. You could manually modify each blueprint, but that’s tedious and error-prone.

Here’s where metaclasses come in. In this metaphor, a metaclass is like a blueprint-making machine. You can program this machine to automatically add the ‘Made in Toy Factory’ stamp to every blueprint it creates. So instead of altering each blueprint (class) individually, you alter the blueprint-making machine (metaclass) once, and it applies the change to all the blueprints (classes) it creates.

That’s a metaclass in Python – a way to define common behavior across multiple classes at the level of the class definitions themselves.

Metaclass Metaphor 2: Russian Dolls

Imagine a set of Russian dolls, where each doll is nested inside a bigger one. The dolls come in three sizes – small, medium, and large.

  1. Small Doll (Object/Instance): The smallest doll represents an instance or an object of a class. This is your standard Python object, such as a string, list, or an instance of a custom class you’ve created.
  2. Medium Doll (Class): The medium-sized doll represents the class. Just as the medium doll can have multiple small dolls identical to the original (you can imagine duplicates if it helps), a class can create multiple instances. When you define a class in Python, you’re effectively creating a ‘blueprint’ for making instances.
  3. Large Doll (Metaclass): The largest doll represents the metaclass. This doll ‘creates’ the medium doll, just like a metaclass in Python defines the behavior of a class. It sets the rules for how the medium doll (class) is structured – in the same way, a metaclass can define methods, attributes, and other properties of a class.
  4. The Act of Nesting (Instantiation): The act of opening a larger doll and revealing a smaller one inside represents the process of instantiation in Python – creating an instance of a class or a class from a metaclass.

This nested dolls analogy can help illustrate the concept of instances, classes, and metaclasses in Python, showing how each level creates and contains the level below it.

Why Do You Need Them?

You don’t. Most Python programmers don’t need metaclasses.

However, they can be useful for:

  1. Dynamically creating or modifying classes: Metaclasses let you automatically add or change attributes/methods when a class is created.
  2. Implementing Design Patterns: Some advanced design patterns, like Singleton or Abstract Base Classes, can be implemented using metaclasses.
  3. Frameworks & ORMs: Libraries or frameworks might use them for internal workings or to provide user-friendly interfaces. Django, for example, uses metaclasses for creating models from database schema.

But with power comes complexity. Misused metaclasses can lead to difficult-to-debug code. Use them sparingly and only when necessary.

Still here? Let’s have a look at the minimal example:

Minimal Python Example

This Python code defines a metaclass Meta that adds an attribute my_attr with value of 100 for any class it creates. MyClass is created with Meta as its metaclass, so my_attr is automatically added to it. When an instance obj of MyClass is created, you can access my_attr on it.

class Meta(type):
    def __init__(cls, name, bases, attrs):
        attrs['my_attr'] = 100
        super().__init__(name, bases, attrs)


class MyClass(metaclass=Meta):
    pass


obj = MyClass()
print(obj.my_attr)  # Prints: 100

Metaclasses and Inheritance

If a class has a metaclass, its subclasses also use the same metaclass.

class SubClass(MyClass):
    pass

print(isinstance(SubClass, MyMeta))  # True

Metaclasses vs Classes

Metaclasses control classes, offering an extra level of indirection. Classes control objects.

Alternatives

Consider class decorators or monkey-patching for less complex modifications to classes.

💡 Recommended: Introduction to Python Classes