Maybe a comment as to why the whole try...except block exists wouldn't go amiss. (And then the meaningless comment "Python 3" should be removed)

Actually, I think just having the "# Python 3" comment is sufficient since it indicates that that is how one should import this module in py3k. I am going to go ahead and merge this. Thanks for the patch!

Actually, I think just having the "# Python 3" comment is sufficient since it indicates that that is how one should import this module in py3k.

I disagree. Stylistically it is unelegant (IMHO) to put a no-op (i.e. see if a builtin is available) inside a try statement. I would have preferred (untested) something like:

try:
   from imp import reload
except ImportError:
   pass # with python < 2.7 reload is a builtin 

I'm not going to ask that this gets resolved (because it makes no functional difference and its only 2 lines), but I would like to make my reasoning for raising a review action clear.

@bfroehle: Thanks for contributing, it is really valuable to us that as many people as possible can get stuck into the matplotlib code and make improvements.

@pelson: Thanks for the review. I actually looked through the matplotlib source to see if there was a preferred way to accomplish this. The proposed patch here was quite common, but also common was a construction which you might have found to be preferable:

if sys.version_info[0] >= 3:
    from imp import reload

The actual patch itself here was styled after some other snippets from the source:

try:
    raw_input
except NameError:
    # Python 3
    raw_input = input

try:
    unichr
except NameError:
    # Python 3
    unichr = chr

And there are plenty other "no-op" try statements scattered throughout the source, including:

try:
    set
except NameError:
    from sets import Set as set

try:
    np.copyto
except AttributeError:
    _putmask = np.putmask
else:
    def _putmask(a, mask, values):
        return np.copyto(a, values, where=mask)

@bfroehle: Thanks for looking through the codebase to identify implemented patterns, I found that combined with your possible alternatives really interesting an illuminating. As you guessed, my preference is the explicit version check for is readability, but you are right to point out pre-existing code as a precedent. Thanks again for this PR.

This didn't get moved across with commit 865f1c0. I'll see if I can do a quick PR.