Why deepcopy doesnt create new references to lambda function?

Question

Found this strange in python:

class SomeClass():
    def __init__(self):
        pass

a = [SomeClass()]
b = copy.deepcopy(a)

Output:

>>> a
[<__main__.Some instance at 0x10051b1b8>]
>>> b
[<__main__.Some instance at 0x10051b092>]

This is just as expected- deepcopy created new SomeClass() object for b.

But if,

f = lambda x:x+1
a = [f]
b = copy.deepcopy(a)

I get:

>>> a
[<function <lambda> at 0x10056e410>]
>>> b
[<function <lambda> at 0x10056e410>]

Why deepcopy doesnt create a new lambda instance in the second case? does that mean lambda functions are atomic?

Answer 1

This applies not just to lambdas, but to functions without state more generally.

>>> def some_function(word): print word
>>> a = [some_function]
>>> a
[<function some_function at 0x1007026e0>]
>>> copy.deepcopy(a)
[<function some_function at 0x1007026e0>]

Because the functions do not store state, deepcopy does not create a new reference for them. An interesting discussion of topics similar to this issue (though not exactly the same issue) is recorded here: http://bugs.python.org/issue1515