changing the class of a python object (casting)

Question

On this python doc page it says:

Like its identity, an object’s type is also unchangeable.

And I try this script,

#!python3  

class Foo:
  num = 1
  pass

class Bar:
  num = 2
  pass

f1,f2= Foo(), Foo()

f2.__class__ = Bar
print( type(f1), type(f2), f1.num, f2.num )

The result shows:

<class '__main__.Foo'> <class '__main__.Bar'> 1 2

I think I changed the type of f2.
What's wrong, What am I missing?

Answer 1

The footnotes one that page says:

[1] It is possible in some cases to change an object’s type, under certain controlled conditions. It generally isn’t a good idea though, since it can lead to some very strange behaviour if it is handled incorrectly.

If you try to change the __class__ of f2 to list:

f2.__class__ = list

A TypeError raised:

TypeError: __class__ assignment: only for heap types

Answer 2

When and how to do this

Changing the type ("casting") makes sense if you want to add functionality to an object created by some code you cannot change.

Assume some statement obj = some_call_to_a_library() gives you an object of class A. You want it to have additional functionality, say, mymethod(). Then you could introduce a subclass MyA like this (Python 3 style):

class MyA(A):
    @classmethod
    def cast(cls, some_a: A):
        """Cast an A into a MyA."""
        assert isinstance(some_a, A)
        some_a.__class__ = cls  # now mymethod() is available
        assert isinstance(some_a, MyA)
        return some_a

    def mymethod(self):
        ...

and then write obj = MyA.cast(some_call_to_a_library()). If MyA relies on additional attributes, cast (which is a factory method) should create them.

I just did something like this when I needed a version of requests.Response that could persist and retrieve responses to/from a file.

Answer 3

I was asked this question by a colleague today. He had a parent class that wanted to promote itself automatically to be one of its children based on an input at init time. The following script worked as a proof of concept:

class ClassB(object):

    def __init__(self):
        self.__class__ = ClassA

    def blah2(self,t):
        print('I give you',t)
        return 'You are welcome'

class ClassA(ClassB):

   def blah(self, t):
       print('you gave me',t)
       return 'Thankyou'



a = ClassB()
print(type(a))
print(a.blah('sausage'))
print(a.blah2('cabbage'))

The result shows:

<class '__main__.ClassA'>
you gave me sausage
Thankyou
I give you cabbage
You are welcome

Which shows that both the parent and child functions are now available to A.

Answer 4

Old question but here's a technique for safely extending the type of a python object.

def extendObject(obj):
    # Our extended type is explicitly derived from our existing type.
    class Extended(obj.__class__):
        # Whatever extensions you want to make.
        def someMethod(self):
            pass
    # Change the type of the object.
    obj.__class__ = Extended
    return obj