Stop using NonASCIILocalPartDefect in email._header_value_parser

When parsing email messages from Unicode strings (but not bytes),
get_local_part() recorded a NonASCIILocalPartDefect for non-ASCII
characters. RFC 5322 permits such addresses.

This change:
- removes the parse-time detection for a non-ASCII local-part
  (and a related test)
- adds tests for passing a non-ASCII addr_spec to
  email.headerregistry.Address.__init__()
- marks the (undocumented) email.errors.NonASCIILocalPartDefect
  as unused and deprecated

 This affected parsing email messages from Unicode strings
(but not from bytes), and also prevented

Author: medmunds

Lib/email/_header_value_parser.py

@@ -1493,11 +1493,6 @@ def get_local_part(value):
             local_part.defects.append(errors.ObsoleteHeaderDefect(
                 "local-part is not a dot-atom (contains CFWS)"))
         local_part[0] = obs_local_part
-    try:
-        local_part.value.encode('ascii')
-    except UnicodeEncodeError:
-        local_part.defects.append(errors.NonASCIILocalPartDefect(
-                "local-part contains non-ASCII characters)"))
     return local_part, value
 
 def get_obs_local_part(value):

Lib/email/errors.py

@@ -109,9 +109,9 @@ class ObsoleteHeaderDefect(HeaderDefect):
     """Header uses syntax declared obsolete by RFC 5322"""
 
 class NonASCIILocalPartDefect(HeaderDefect):
-    """local_part contains non-ASCII characters"""
-    # This defect only occurs during unicode parsing, not when
-    # parsing messages decoded from binary.
+    """Unused. Note: this error is deprecated and may be removed in the future."""
+    # RFC 6532 permits a non-ASCII local-part. _header_value_parser previously
+    # treated this as a parse-time defect (when parsing Unicode, but not bytes).
 
 class InvalidDateDefect(HeaderDefect):
     """Header has unparsable or invalid date"""

Lib/email/headerregistry.py

@@ -45,12 +45,8 @@ def __init__(self, display_name='', username='', domain='', addr_spec=None):
                 raise ValueError("Invalid addr_spec; only '{}' "
                                  "could be parsed from '{}'".format(
                                     a_s, addr_spec))
-            relevant_defects = [
-                defect for defect in a_s.all_defects
-                if not isinstance(defect, errors.NonASCIILocalPartDefect)
-            ]
-            if relevant_defects:
-                raise relevant_defects[0]
+            if a_s.all_defects:
+                raise a_s.all_defects[0]
             username = a_s.local_part
             domain = a_s.domain
         self._display_name = display_name

Lib/test/test_email/test__header_value_parser.py

@@ -1235,17 +1235,6 @@ def test_get_local_part_valid_and_invalid_qp_in_atom_list(self):
             '@example.com')
         self.assertEqual(local_part.local_part, r'\example\\ example')
 
-    def test_get_local_part_unicode_defect(self):
-        # Currently this only happens when parsing unicode, not when parsing
-        # stuff that was originally binary.
-        local_part = self._test_get_x(parser.get_local_part,
-            'exá[email protected]',
-            'exámple',
-            'exámple',
-            [errors.NonASCIILocalPartDefect],
-            '@example.com')
-        self.assertEqual(local_part.local_part, 'exámple')
-
     # get_dtext
 
     def test_get_dtext_only(self):

Misc/NEWS.d/next/Library/2024-07-30-19-19-33.gh-issue-81074.YAeWNf.rst

@@ -1 +1,8 @@
-Allow non-ASCII addr_spec in email.headerregistry.Address constructor
+The :mod:`email` module no longer treats email addresses with non-ASCII
+characters as defects when parsing a Unicode string or in the ``addr_spec``
+parameter to :class:`email.headerregistry.Address`. :rfc:`5322` permits such
+addresses, and they were already supported when parsing bytes and in the Address
+``username`` parameter.
+
+The (undocumented) :exc:`!email.errors.NonASCIILocalPartDefect` is no longer
+used and should be considered deprecated.