I like where this is going, but I think we're actually losing some flexibility by enforcing isalpha() and isdigit(), etc. Right now, you can subclass _timelex() and override the relevant character sets, but in your version, that's no longer possible.

That said, I think that this is the right default approach. Today I'll make a PR against your branch with a suggested improvement.

@jarondl This seems like a major change - any thoughts on the matter?

See also my PR at ms-boom/dateutil#1

I'm going to go ahead and merge this. We can always undo it later if it causes some significant problems.

Hi @pganssle, and @ms-boom,
a post-factum +1 from me. It is much better than what we had.