Use `__dict__` or `vars()`?

Builtin function vars() looks more Pythonic to me, but I see __dict__ used more frequently.

The Python documentation indicates that they are equivalent.

One blogger claims that __dict__ is faster than vars().

Which shall I use?

Solution 1:

Generally, you should consider dunder/magic methods to be the implementation and invoking functions/methods as the API, so it would be preferable to use vars() over __dict__, in the same way that you would do len(a_list) and not a_list.__len__(), or a_dict["key"] rather than a_dict.__getitem__('key')

Solution 2:

I'd use vars().

From: https://wiki.python.org/moin/DubiousPython#Premature_Optimization

While a correctly applied optimization can indeed speed up code, optimizing code that is only seldom use [..] can make code harder to read. [..] Write correct code first, then make it fast (if necessary).

From: The Zen of Python

Readability counts.

Solution 3:

I agree vars should be preferred. My rationale is that, as python evolves, vars might be extended to do more than __dict__ does (for example, work for objects with slots, possibly in 3.7).