How are Python metaclasses different from regular class inheritance? [duplicate]

Question

This question already has answers here:

What are metaclasses in Python? (24 answers)

Closed 8 years ago.

This might be too much of an open ended question, but I'm just now learning about metaclasses in Python and I don't understand how a metaclass is any different than just having a child class inherit from the parent class, like

class child(parent):

Wouldn't this serve the same purpose of a metaclass? I guess maybe I don't understand the purpose of a metaclass.

Answer 1

class AccessorType(type):
    def __init__(self, name, bases, d):
        type.__init__(self, name, bases, d)
        accessors = {}
        prefixs = ["get_", "set_", "del_"]

        for k in d.keys():
            v = getattr(self, k)
            for i in range(3):
                if k.startswith(prefixs[i]):
                    accessors.setdefault(k[4:], [None, None, None])[i] = v

        for name, (getter, setter, deler) in accessors.items():
            # create default behaviours for the property - if we leave
            # the getter as None we won't be able to getattr, etc..
            # [...] some code that implements the above comment
            setattr(self, name, property(getter, setter, deler, ""))

Answer 2

The difference is that inheriting from a class does not affect how the class is created, it only affects how instances of the class are created. If you do:

class A(object):
    # stuff

class B(A):
    # stuff

then A does not have any opportunity to "hook in" when B is created. Methods of A may be called when an instance of B is created, but not when the class B itself is created.

Metaclasses allow you to define custom behavior for when a class is created. Refer to the question I marked as duplicate for examples of how metaclasses work, and convince yourself that there are effects in those examples that you can't achieve with normal inheritance.