The library doesn't comply with RFC 7643 attribute naming definitions

[NaqGuug]

RFC 7643 §2.1 states the following ABNF rules for attribute names:

ATTRNAME = ALPHA *(nameChar)
nameChar = "$" / "-" / "_" / DIGIT / ALPHA

However, currently it is impossible to have attributes with special characters "-", "_" and "$". The reason is too aggressive normalisation removing all special characters from attributes.

scim2-models/scim2_models/utils.py

Lines 66 to 75 in 279d405

	def _normalize_attribute_name(attribute_name: str) -> str:
	"""Remove all non-alphabetical characters and lowerise a string.
	This method is used for attribute name validation.
	"""
	is_extension_attribute = ":" in attribute_name
	if not is_extension_attribute:
	attribute_name = _NON_WORD_UNDERSCORE.sub("", attribute_name)
	return attribute_name.lower()

For example in the following situation the attributes gets merged together while by RFC 7643 the attributes would be considered different.

from scim2_models import Resource, AnyExtension, URN
from pydantic import Field


class CustomModel(Resource[AnyExtension]):
    __schema__ = URN("urn:ietf:param:scim:something")

    # userName
    user_name: str | None = None

    # For users this field will show as "user_name"
    # which is fine by RFC 7643 rules
    string_field: str | None = Field(None, serialization_alias="user_name")


model = CustomModel.model_validate({"userName": "string1", "user_name": "string2"})
print(model.model_dump())
# "userName" got the value from "user_name", even if they are two separate fields
# {'schemas': ['urn:ietf:param:scim:something'], 'userName': 'string2'}

There is a even simpler example where using alias breaks things. User clearly wanted to set value for "string_field" which has been defined, but the library doesn't allow that. Yes, in this situation serialization_alias works, but as above example shows, it can get merged with another field like "stringfield", "string-field" or even "s$t$r$i$n$g$f$i$e$l$d$"

class CustomModel(Resource[AnyExtension]):
    __schema__ = URN("urn:ietf:param:scim:something")

    # It's impossible to access this value because
    # during validation the incoming field "string_field"
    # gets converted to "stringfield" and raises an error.
    string_field: str | None = Field(None, alias="string_field")


model = CustomModel.model_validate({"string_field": "string"})

# Validation error
# pydantic_core._pydantic_core.ValidationError: 1 validation error for CustomModel
# stringfield
#   Extra inputs are not permitted [type=extra_forbidden, input_value='string', input_type=str]
#     For further information visit https://errors.pydantic.dev/2.13/v/extra_forbidden

[azmeuk]

Thank you for your bug report, you pointed out a big subject :)
This should be fixed with #167