URLField serializer is not playing nice with "built in" test server address http://testserver

[yarikoptic]

Initially encountered in our

• Local runtime schema serialization endpoints dandi/dandi-archive#2386
  where our response model contains schema_url which is a URLField and when we plugged in a "proper" URL to local test instance http://testserver/api/schema/latest/dandiset/, then response we were getting from server became with code 400 and .content saying

(Pdb) p resp.content
b'{"schema_url":["Enter a valid URL."]}'

with such minimal reproducer depicting the issue:

from rest_framework import serializers

class MySerializer(serializers.Serializer):
    url = serializers.URLField()

def test_valid_url():
    serializer = MySerializer(data={'url': 'http://testserver/path'})
    serializer.is_valid()

running pytest on which would give you an ugly excepted demanding configuration since i18n would not yet be configured to give a proper localized error...

but replacing that simple hostname testserver with a domain name with the TLD (e.g. testserver.com), which is AFAIK is not required for a URL to be valid, pytest would pass just fine.

❯ sed -i -E 's,(testserver),\1.com,g' /tmp/tttt.py
❯ python -m pytest -s -v /tmp/tttt.py
====================================================================================================== test session starts =======================================================================================================
platform linux -- Python 3.11.12, pytest-8.3.5, pluggy-1.6.0 -- /home/yoh/proj/dandi/dandi-archive/.venv/bin/python
cachedir: .pytest_cache
django: version: 4.2.21
rootdir: /tmp
plugins: django-s3-file-field-1.0.1, Faker-37.3.0, mock-3.14.0, memray-1.7.0, factoryboy-2.7.0, django-4.11.1, cov-6.1.1
collected 1 item                                                                                                                                                                                                                 

../../../../../tmp/tttt.py::test_valid_url PASSED

So

• any way to workaround?
• if I got it right such validation stems from DJANGO itself, so should it be checked/fixed there?

[auvipy]

you should have open a discussion first