Introduction to Python Metaclasses

Creating Custom Metaclasses

In Python, we can customize the class creation process by passing the metaclass keyword in the class definition. This can also be done by inheriting a class that has already passed in this keyword.

class MyMeta(type):
    pass

class MyClass(metaclass=MyMeta):
    pass

class MySubclass(MyClass):
    pass

We can see below that the type of MyMeta class is type and that the type of MyClass and MySubClass is MyMeta.

print(type(MyMeta))
print(type(MyClass))
print(type(MySubclass))

<class 'type'>
<class '__main__.MyMeta'>
<class '__main__.MyMeta'>

When defining a class and no metaclass is defined the default type metaclass will be used. If a metaclass is given and it is not an instance of type(), then it is used directly as the metaclass.

__new__ and __init__

Metaclasses can also be defined in one of the two ways shown below. We'll explain the difference between them below.

class MetaOne(type):
    def __new__(cls, name, bases, dict):
        pass

class MetaTwo(type):
    def __init__(self, name, bases, dict):
        pass

__new__ is used when one wants to define dict or bases tuples before the class is created. The return value of __new__is usually an instance of cls. __new__ allows subclasses of immutable types to customize instance creation. It can be overridden in custom metaclasses to customize class creation. __init__ is usually called after the object has been created so as to initialize it.

Metaclass __call__ method

According to the official docs, we can also override other class methods by defining a custom __call__() method in the metaclass that allows custom behavior when the class is called.

Metaclass __prepare__ method

According to Python's data model docs

Once the appropriate metaclass has been identified, then the class namespace is prepared. If the metaclass has a __prepare__ attribute, it is called as namespace = metaclass.__prepare__(name, bases, **kwds) (where the additional keyword arguments, if any, come from the class definition). If the metaclass has no __prepare__attribute, then the class namespace is initialized as an empty ordered mapping. - docs.python.org

Singleton Design using a Metaclass

This is a design pattern that restricts the instantiation of a class to only one object. This could prove useful for example when designing a class to connect to the database. One might want to have just one instance of the connection class.

class SingletonMeta(type):
    _instances = {}
    def __call__(cls, *args, **kwargs):
        if cls not in cls._instances:
            cls._instances[cls] = super(SingletonMeta,cls).__call__(*args, **kwargs)
        return cls._instances[cls]

class SingletonClass(metaclass=SingletonMeta):
    pass

Conclusion

In this article, we have learned about what metaclasses are and how we can implement them in our Python programming. Metaclasses can be applied in logging, registration of classes at creation time and profiling among others. They seem to be quite abstract concepts, and you might be wondering if you need to use them at all. Long-term Pythonista, Tim Peters answers that question best.

"Metaclasses are deeper magic than 99% of users should ever worry about. If you wonder whether you need them, you don’t (the people who actually need them know with certainty that they need them, and don’t need an explanation about why).” - realpython

If you would like to learn more about Python, start DataCamp's Python Programming skill track and check out our Python Classes Tutorial.

References

Link one

Link two