url_regex doesn't correctly match all allowed characters in a URI

[jonathanunderwood]

Bug

from pydantic.networks import url_regex
m=url_regex.match("postgresql+psycopg2://postgres:postgres@localhost:5432/hatch")
parts = m.groupdict()
parts

produces the following incorrect matching:

{'scheme': None,
 'user': 'postgresql+psycopg2',
 'password': '//postgres:postgres',
 'ipv4': None,
 'ipv6': None,
 'domain': 'localhost',
 'port': '5432',
 'path': '/hatch',
 'query': None,
 'fragment': None}

This is with Python 3.7.5 on OSX, pydantic version 1.2

[jonathanunderwood]

Note that according to Appendix A of rfc2396 a URI scheme is defined as:

scheme        = alpha *( alpha | digit | "+" | "-" | "." )

(search the document for "Collected BNF for URI").

Edit: RFC3986 is probably more up to date. Appendix A of that document describes scheme as:

scheme        = ALPHA *( ALPHA / DIGIT / "+" / "-" / "." )

In any case the +, - and . characters are all valid in the scheme part.

[samuelcolvin]

Thanks for reporting, please could you submit a PR for this?

[jonathanunderwood]

I'll look into it. Will need to get work clearance etc tho, so it might take a little while.

[samuelcolvin]

Thank you, let me know if it's too much work to be worth it and/or urgent, if so I can do it.

[jonathanunderwood]

It would require legal review, so if you didn't mind doing it, that would be very much appreciated.

[samuelcolvin]

wow 😴, there are advantages to working for a startup.

I'll do it.

[samuelcolvin]

Fixed via #1148, please review and check that will work for you.