capfirst() corrupts text for languages without capital letters

[irakli khitarishvili]

When using Django admin for ka-ge language first characters gets corrupted, because language dose not have capital letters.
For example "username" in Georgian is "მომხმარებლის სახელი" and when first letter is capitalized it's displayed as "Სომხმარებლის სახელი".

You can test this case when "log in" page is displayed for admin.

In django.contrib.admin.forms.AuthenticationForm

    def __init__(self, request=None, *args, **kwargs):
        self.request = request
        self.user_cache = None
        super().__init__(*args, **kwargs)
        self.username_field = UserModel._meta.get_field(UserModel.USERNAME_FIELD)
        self.fields['username'].max_length = self.username_field.max_length or 254
        if self.fields['username'].label is None:
            # This is the line when character gets corrupted
            self.fields['username'].label = capfirst(self.username_field.verbose_name)

[irakli khitarishvili]

screen of the page

[Tim Graham]

Any idea how django.utils.text.capfirst() could be adapted to fix the issue? I was thinking maybe we could check ord() of the first character's upper() but I'm not sure if that's reliable.or not.

[Claude Paroz]

I cannot reproduce this issue on a Linux/Python3.5 system.

>>> ord('მ')
4315
>>> ord('მ'.upper())
4315

What system are you using?

[irakli khitarishvili]

Python : 3.7.0
System : Windows 7, Linux (Ubuntu 17.10)

[Claude Paroz]

That's weird. However, I don't think that's something that Django can change. It must be related to the underlying Unicode library mappings.
Maybe other people can test to see if the issue you are seeing is the exception or the rule!

[Carlton Gibson]

Repeating Claude's test on macOS with Python 3.6 and 3.7 and on Windows 10 (with cmd + Python 3.6 and WSL bash + Python 3.5) I get the same result:

>>> ord('მ')
4315
>>> ord('მ'.upper())
4315

[Tim Graham]

Looks like Python 3.7 is required to reproduce. Bisected to ​https://github.com/python/cpython/commit/4705ea38c900f068fd262aca02943896d1123544.

On Ubuntu 18.04 with a manually installed Python 3.7:

>>> ord('მ'.upper())
7323

[Claude Paroz]

Maybe a Unicode bug, then? Looks like Მ / \u1C9B / ​https://unicode-table.com/en/1C9B/ has no information in Unicode. (not that Unicode does have capitals for Georgian, see e.g. ​https://unicode-table.com/en/10AB/).

Anyway, I fear the resolution will not take place in Django, but in Python or in Unicode data. I'd suggest trying to report the issue in Python.