Invalid regexp in URLValidator can't handle file:// schemes

[Marcin Nowak]

Regexp does not allow to use file scheme in URLValidator.

Steps to reproduce

from django.core.validators import URLValidator
URLValidator(schemes=['file'])('file:///tmp/somefile')

Expected result

No exception should be raised.

Current result

ValidationError: [u'Enter a valid URL.']

[Tim Graham]

Can this be fixed without special casing the file scheme in the validation?

[Claude Paroz]

It's URLValidator not URIValidator, I think you might need your own custom validator for this. The current regex is already bloated, I'm not sure we should continue to complexify it...

[Tim Graham]

I had a similar skepticism. If we don't make a change, then let's clarify the documentation to prevent further tickets about this.

[Johannes Maron]

I thinks this is both a documentation and code problem. The URLValidator should raise a value error, if a unsupported schema is set. In the documentation it should be explained with RFC 1738 schemas are supported. Currently one gets the impression that all are.

[Johannes Maron]

Interestingly enough, if you include the optional host parameter in RFC1738 it works: file://localhost/tmp/file/name

[Adam Zapletal]

I opened a PR adding a warning about this to the documentation. I'm happy to handle it in a different way if that'd be better.

For what it's worth, if someone out there wants to validate local file URIs in addition to URLs, here is an idea to start with:

@deconstructible
class CustomURLValidator(URLValidator):
    def __init__(self, **kwargs):
        super().__init__(**kwargs)

        self.schemes.append('file')

    def __call__(self, value):
        if value.startswith('file:///'):
            value = value.replace('file:///', 'file://localhost/', 1)

        return super().__call__(value)

[Mariusz Felisiak <felisiak.mariusz@…>]

In 7326513a:

Fixed #25595 -- Doc'd that URLValidator rejects ​file:// URIs without a host.

[Mariusz Felisiak <felisiak.mariusz@…>]

In 710ca576:

[5.0.x] Fixed #25595 -- Doc'd that URLValidator rejects ​file:// URIs without a host.

Backport of 7326513a8f5d4d4e0aeec28540f9451b939b1dda from main