Fix for Python 4

[hugovk]

We don't yet know if 3.10 or 4.0 will follow Python 3.9, but it will probably happen in 2020 when Python 3.9 reaches beta and as six is vendored and used in many codebases, it would be good for six to be ready for both.

There's several places in the codebase which essentially do:

if sys.version_info[0] == 3:
   pass  # Python 3+ stuff
else:
   pass  # Python 2 stuff

When run on Python 4, this will run the Python 2 code!

This PR flips it around:

if sys.version_info[0] == 2:
   pass  # Python 2 stuff
else:
   pass  # Python 3+ stuff

Thanks to @asottile for flake8-2020.

---

As an aside, there's a slight inconsistency here:

# Useful for very coarse version differentiation.
PY2 = sys.version_info[0] == 2
PY3 = sys.version_info[0] == 3
PY34 = sys.version_info[0:2] >= (3, 4)

PY34 is true on Python 4, but PY3 is not. Would it be too big a break to change it?

-PY3 = sys.version_info[0] == 3
+PY3 = sys.version_info[0] >= 3

[benjaminp]

I don't think this is worth worrying about unless Python 4 is actually concretely on the horizon, and people are still using six.

https://bitbucket.org/gutworth/six/issues/22/variable-sixpy3-is-time-bomb

[asottile]

fwiw, there's already indication that python 4 will happen:

>>> from __future__ import annotations
>>> annotations
_Feature((3, 7, 0, 'beta', 1), (4, 0, 0, 'alpha', 0), 1048576)

that is, __future__ annotations will become the default in python 4

why wait to fix this?

[benjaminp]

I was just discussing with GvR this week that the annotations will probably have to move to 3.10 or 3.11.

[asottile]

separately: why wait to fix this?

[benjaminp]

I don't think there's anything to fix.

[asottile]

legacy code lives for a long time, I wouldn't be surprised to see six for a long long time (especially now with the addition of the new ensure_* helpers which have no direct replacement)

[benjaminp]

Yes, but again, for this to be more than a theoretical problem, Python 4 would have to be more than theoretical.

[asottile]

we know it will eventually come, why wait until it's a problem when it's easy to fix now?

~/workspace/cpython $ git grep 'deprecated-removed::.*4\.0' | wc -l
23

[benjaminp]

We don't know it will eventually come. That PEP 563 and deprecation notices mention it just goes to show people consider it a safely vague indefinite future event. It's always "a while off".

[asottile]

that's fair -- my concern is that since six is broken (and therefore a huge percentage of the ecosystem) it will force the issue to make it never happen 🤷‍♂

[benjaminp]

that's fair -- my concern is that since six is broken (and therefore a huge percentage of the ecosystem) it will force the issue to make it never happen man_shrugging

I'm not exactly sure what the antecedents of the "it"s in this sentence are.

[asottile]

that's fair -- my concern is that since six is broken (and therefore a huge percentage of the ecosystem) [six] will force the issue to make [python 4] never happen man_shrugging

I'm not exactly sure what the antecedents of the "it"s in this sentence are.

[asottile]

yeah sorry that sentence was a mess :)