Username suggestion for non-English names

[CodeWithEmad]

While a user tries to register, Authentication MFE sends a request to LMS_HOST/api/user/v1/validation/registration to suggest 3 usernames but usernames can only contain letters (A-Z, a-z), but, the suggested name does not follow this rule for non-English names.

The generate_username_suggestions function in edx-platform/openedx/core/djangoapps/user_authn/utils.py is responsible for generating these names.

edx-platform/openedx/core/djangoapps/user_authn/utils.py

Lines 79 to 122 in b07464b

	def generate_username_suggestions(name):
	""" Generate 3 available username suggestions """
	username_suggestions = []
	max_length = USERNAME_MAX_LENGTH
	names = name.split(' ')
	if names:
	first_name = remove_special_characters_from_name(names[0].lower())
	last_name = remove_special_characters_from_name(names[-1].lower())
	if first_name != last_name and first_name:
	# username combination of first and last name
	suggestion = f'{first_name}{last_name}'[:max_length]
	if not username_exists_or_retired(suggestion):
	username_suggestions.append(suggestion)
	# username is combination of first letter of first name and last name
	suggestion = f'{first_name[0]}-{last_name}'[:max_length]
	if not username_exists_or_retired(suggestion):
	username_suggestions.append(suggestion)
	if len(first_name) >= 2:
	short_username = first_name[:max_length - 6] if max_length is not None else first_name
	short_username = short_username.replace('_', '').replace('-', '')
	int_ranges = [
	{'min': 0, 'max': 9},
	{'min': 10, 'max': 99},
	{'min': 100, 'max': 999},
	{'min': 1000, 'max': 9999},
	{'min': 10000, 'max': 99999},
	]
	for int_range in int_ranges:
	for _ in range(10):
	random_int = random.randint(int_range['min'], int_range['max'])
	suggestion = f'{short_username}_{random_int}'
	if not username_exists_or_retired(suggestion):
	username_suggestions.append(suggestion)
	break
	if len(username_suggestions) == 3:
	break
	return username_suggestions

[CodeWithEmad]

IDK what's the best solution here.

• We can use unidecode package

from unidecode import unidecode

name = 'عماد'  # Persian word for "Emad"
transliterated_name = unidecode(name)
print(transliterated_name)  # Output: 'mad

There are other packages like slugify but I guess they're using something similar under the hood.

• We can check on the frontend side, and send the request only if the word consists of Aa-Zz 9-0 - _

Do you have any idea @kdmccormick? Maybe this was fixed somewhere in the platform before, but this piece was left out.

[kdmccormick]

Hmm, interesting, thanks for the report @CodeWithEmad .

As a user of the platform in a non-ascii language, what would be your preferred behavior? If you enter your name in Arabic script, would you like that platform to try to suggest a phonetically similar ascii username, or would you rather it simply suggest no usernames?

[regisb]

My 2 cents: when I enter "Régis Behmo" the suggested username is "régisb" and it would make sense to convert that into "regisb". ("é" is not a supported character)

The year is 2024, and I think that the "right" fix for this would be to support unicode characters in user names.

[CodeWithEmad]

@kdmccormick I usually enter my name in English everywhere so I never faced this issue, but I tested a couple of services like x.com and they're using a similar solution to the unidecode approach I mentioned above.
I've entered "عماد راد" which is "Emad Rad" in Persian, and the username was "mdrd1264007" (not my actual account. I'll delete this later)

@regisb I like the idea of supporting the unicode characters for username, but I'm afraid it introduces some complexity later where one small or big package/service under the hood doesn't support unicode and it breaks something.

Suggesting a phonetically similar ASCII username is a reasonable workaround for this issue, but I am open to any solution.