Syntax for making objects callable in python-CodePudding

I understand that in python user-defined objects can be made callable by defining a __call__() method in the class definition. For example,

class MyClass:
  def __init__(self):
    pass

  def __call__(self, input1):
    self.my_function(input1)

  def my_function(self, input1):
    print(f"MyClass - print {input1}")

my_obj = MyClass()
# same as calling my_obj.my_function("haha")
my_obj("haha") # prints "MyClass - print haha"

I was looking at how pytorch makes the forward() method of a nn.Module object be called implicitly when the object is called and saw some syntax I didn't understand.

In the line that supposedly defines the __call__ method the syntax used is,

__call__ : Callable[..., Any] = _call_impl

This seemed like a combination of an annotation (keyword Callable[ following : ignored by python) and a value of _call_impl which we want to be called when __call__ is invoked, and my guess is that this is a shorthand for,

def __call__(self, *args, **kwargs):
    return self._call_impl(*args, **kwargs)

but wanted to understand clearly how this method of defining functions worked.

My question is: When would we want to use such a definition of callable attributes of a class instead of the usual def myfunc(self, *args, **kwargs)

CodePudding user response：

Functions are normal first-class objects in python. The name to with which you define a function object, e.g. with a def statement, is not set in stone, any more than it would be for an int or list. Just as you can do

a = [1, 2, 3]
b = a

to access the elements of a through the name b, you can do the same with functions. In your first example, you could replace

def __call__(self, input1):
    self.my_function(input1)

with the much simpler

__call__ = my_function

You would need to put this line after the definition of my_function.

The key differences between the two implementations is that def __call__(... creates a new function. __call__ = ... simply binds the name __call__ to the same object as my_function. The noticeable difference is that if you do __call__.__name__, the first version will show __call__, while the second will show my_function, since that's what gets assigned by a def statement.