Difference with re, when repl returns None

[brondsem]

It's not correct for a repl function to return None, but I encountered some code doing that and the stdlib re silently allows it and regex raises an error. Should it be made consistent? Thanks

$ python
Python 3.7.10 (default, Feb 24 2021, 20:00:24)
[GCC 4.8.5 20150623 (Red Hat 4.8.5-36)] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import regex, re
>>> def repl(m):
...     return None
...
>>> re.sub(r'a', repl, 'abc')
'bc'
>>> regex.sub(r'a', repl, 'abc')
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/var/local/env-py37/lib/python3.7/site-packages/regex/regex.py", line 278, in sub
    return pat.sub(repl, string, count, pos, endpos, concurrent, timeout)
TypeError: expected str instance, NoneType found

[mrabarnett]

Well, .findall will never return None for empty groups:

>>> re.search(r'(.)|(.)', 'a').groups()
('a', None)
>>> re.findall(r'(.)|(.)', 'a')
[('a', '')]

For compatibility with re, perhaps regex should do the same, even though it could hide a bug.

I'm a little unsure at this point.

Has re's behaviour ever been reported on the Python bug tracker? If re isn't going to change, if it's intentional behaviour, then I might as well copy it.

[brondsem]

I couldn't find an Python bug report for it.

Either way is ok with me, not a huge deal in my situation. Maybe a middle-ground option would be to copy the behavior but issue a Warning?