python.lsh: Avoid capturing underscores as numeric literals

[James (jagourq)]

The definition of numeric literals in python.lsh currently captures underscores, so identifiers like __init__ and ._foo are considered to be numeric literals immediately followed by identifiers and are discarded as invalid.

I've changed the regex for numeric literals by replacing [\d_]+ with \d[\d_]*.

[Leonard Hecker (lhecker)]

Actually I realize now that your explanation doesn't make sense. How would __init__ match this clause? It doesn't even contain any digits!

Can you explain what this PR solves?

[James (jagourq)]

This PR fixes the highlighting of methods beginning with _ and variables consisting solely of _ in Python source files.
The regex for digits is designed to allow digits to contain _, but the pattern used also captures _, __ and so on.

The __ in __main__ is currently captured by [\d_]+ in the regex for digits on line 55, main__ is then captured by \w+ on line 56, and everything is discarded as an "Invalid numeric literal" (see the comment on line 57), preventing __main__ from being parsed as a method afterwards.

Currently, the code

_ = 3.14
def __main__(): pass
def main(): pass
f._foo()
f.foo()

renders as

_ is incorrectly highlighted as a digit, and __main__() and ._foo() are not highlighted as methods.
After these commits, it renders as

with _ highlighted similarly to other variables and __main__() and ._foo() highlighted as methods, fixing the previous issues.

[Leonard Hecker (lhecker)]

Well, time to fix all the others too then!

[Leonard Hecker (lhecker)]

James (@jagourq) We can merge this after you agree to the contributor's agreement above.

[James (jagourq)]

@microsoft-github-policy-service agree

[Leonard Hecker (lhecker)]

Thanks! :)