Quiz time! Consider the following piece of code:
class A:
def f(self):
print("A.f called")
super().f()
class B:
def f(self):
print("B.f called")
class C(A, B):
pass
C().f()
Will this code crash when run? Obviously A doesn't inherit from anything, so that super().f() call is suspect. But at runtime a "super" of sorts does exist in C - namely, B. Place your bets now…
The answer, to my surprise, is that this code runs fine. My intuition of super() was a sort of static implementation, where if a class doesn't inherit from anything then any super call will fail (well, get forwaded to object, which obviously doesn't implement f).
But in reality, super() looks at the mro at runtime and chooses the next matching function to call. That mro is dependent on the class the function belongs to (C), not the class the function was originally defined in (A). If we dynamically modify the mro of C to take B out of the picture, the call to f fails:
class A:
def f(self):
print("A.f called")
super().f()
class B:
def f(self):
print("B.f called")
class ChangeMROMeta(type):
def mro(cls):
return (cls, A, object)
class C(A, B, metaclass=ChangeMROMeta):
pass
c = C()
print(C.__mro__)
c.f()
This means you can have a method in a class which always fails except when subclasssed in a very particular way. I can't think of a black magic use case for this, but maybe someone else can!