Difference in "dir" between Python 2 and 3

Question

The following code behaves differently in Python 2 and Python 3, and I'm not sure why.

class Dataset(object):
    def __getattr__(self, item):
        if not item in dir(self):
            print(item)

a = Dataset()
a.Hello

The result in Python 3:

> Hello

The result in Python 2:

__members__
__members__
__methods__
...

ad infinitum until a recursion ceiling is reached. What is the difference in the behavior of "dir"?

Edit: And is there a workaround? self.dict is the obvious choice but it doesn't include functions which turns out to be a problem in my code.

Answer 1

The documentation for dir in Python 2.7 and 3.5 seems identical - there are no implementation details. But clearly, dir() in Python 2 invokes __getattr__ causing the infinite recursion.

However, both documentation sets do say that

Because dir() is supplied primarily as a convenience for use at an interactive prompt, it tries to supply an interesting set of names more than it tries to supply a rigorously or consistently defined set of names, and its detailed behavior may change across releases. For example, metaclass attributes are not in the result list when the argument is a class.

That note about it being a convenience is significant.

If you modify your __getattr__ to look at self.__dict__ instead of using dir(),the problem goes away.

In [5]: class Dataset(object):
          def __getattr__(self, item):
            if not item in self.__dict__:
              print(item)
   ...:             

In [6]: a = Dataset()

In [7]: a.Hello
Hello

Answer 2

Without examining the source code, I can't say why this happens (although I have some hypotheses), but here is a pretty simple workaround:

class Dataset(object):
    def __getattr__(self, item):
        try:
            super(Dataset, self).__getattr__(item)
        except AttributeError:
            print(item)