Frazer McLean noticed that even though the version parsing regex in
PEP 440 Appendix B was explicitly restricted to ASCII digits, the
`is_canonical` regex was more permissive and nominally allowed
arbitrary Unicode [Nd] characters.
The restrictive interpretation was the correct one (and reflects what
tools actually support), so I've merged Frazer's change at
to fix the omission and update the `is_canonical` regex to be
consistent with the version parsing one.
Nick Coghlan | ncoghlan(a)gmail.com | Brisbane, Australia