Raise error for invalidly-formatted email addresses.

A message's `from_email` and each address in its `to`, `cc`, and `bcc` lists must contain exactly one email address. Previous code would silently ignore additional addresses, leading to unusual behavior. Now, raises new `AnymailInvalidAddress` exception.

Example: `from_email='Widgets, Inc. <widgets@example.com>'` is invalid: it needs double-quotes around the "Widgets, Inc." display-name portion. In earlier versions, this probably would have sent the message from something like "From: Widgets <@localhost>". Now, it will raise an exception.

**Potentially-breaking change:** If your code is using an unquoted display-name containing a comma in an email address, it will now raise an error. In earlier versions, this may have appeared to succeed, but was almost certainly not doing what you intended.

Fixes #44.

anymail/exceptions.py

@@ -101,6 +101,10 @@ class AnymailRecipientsRefused(AnymailError):
         super(AnymailRecipientsRefused, self).__init__(message, *args, **kwargs)
 
 
+class AnymailInvalidAddress(AnymailError, ValueError):
+    """Exception when using an invalidly-formatted email address"""
+
+
 class AnymailUnsupportedFeature(AnymailError, ValueError):
     """Exception for Anymail features that the ESP doesn't support.
 

anymail/utils.py

@@ -2,15 +2,16 @@ import mimetypes
 from base64 import b64encode
 from datetime import datetime
 from email.mime.base import MIMEBase
-from email.utils import formatdate, parseaddr, unquote
+from email.utils import formatdate, getaddresses, unquote
 from time import mktime
 
 import six
 from django.conf import settings
 from django.core.mail.message import sanitize_address, DEFAULT_ATTACHMENT_MIME_TYPE
+from django.utils.encoding import force_text
 from django.utils.timezone import utc
 
-from .exceptions import AnymailConfigurationError
+from .exceptions import AnymailConfigurationError, AnymailInvalidAddress
 
 UNSET = object()  # Used as non-None default value
 
@@ -93,31 +94,39 @@ def getfirst(dct, keys, default=UNSET):
         return default
 
 
+def parse_one_addr(address):
+    # This is email.utils.parseaddr, but without silently returning
+    # partial content if there are commas or parens in the string:
+    addresses = getaddresses([address])
+    if len(addresses) > 1:
+        raise ValueError("Multiple email addresses (parses as %r)" % addresses)
+    elif len(addresses) == 0:
+        return ('', '')
+    return addresses[0]
+
+
 class ParsedEmail(object):
-    """A sanitized, full email address with separate name and email properties"""
+    """A sanitized, full email address with separate name and email properties."""
 
     def __init__(self, address, encoding):
-        self.address = sanitize_address(address, encoding)
-        self._name = None
-        self._email = None
-
-    def _parse(self):
-        if self._email is None:
-            self._name, self._email = parseaddr(self.address)
+        if address is None:
+            self.name = self.email = self.address = None
+            return
+        try:
+            self.name, self.email = parse_one_addr(force_text(address))
+            if self.email == '':
+                # normalize sanitize_address py2/3 behavior:
+                raise ValueError('No email found')
+            # Django's sanitize_address is like email.utils.formataddr, but also
+            # escapes as needed for use in email message headers:
+            self.address = sanitize_address((self.name, self.email), encoding)
+        except (IndexError, TypeError, ValueError) as err:
+            raise AnymailInvalidAddress("Invalid email address format %r: %s"
+                                        % (address, str(err)))
 
     def __str__(self):
         return self.address
 
-    @property
-    def name(self):
-        self._parse()
-        return self._name
-
-    @property
-    def email(self):
-        self._parse()
-        return self._email
-
 
 class Attachment(object):
     """A normalized EmailMessage.attachments item with additional functionality

docs/sending/exceptions.rst

@@ -36,6 +36,26 @@ Exceptions
     your ESP's dashboard. See :ref:`troubleshooting`.)
 
 
+.. exception:: AnymailInvalidAddress
+
+    .. versionadded:: 0.7
+
+    The send call will raise a :exc:`!AnymailInvalidAddress` error if you
+    attempt to send a message with invalidly-formatted email addresses in
+    the :attr:`from_email` or recipient lists.
+
+    One source of this error can be using a display-name ("real name") containing
+    commas or parentheses. Per :rfc:`5322`, you should use double quotes around
+    the display-name portion of an email address:
+
+    .. code-block:: python
+
+        # won't work:
+        send_mail(from_email='Widgets, Inc. <widgets@example.com>', ...)
+        # must use double quotes around display-name containing comma:
+        send_mail(from_email='"Widgets, Inc." <widgets@example.com>', ...)
+
+
 .. exception:: AnymailSerializationError
 
     The send call will raise a :exc:`!AnymailSerializationError`

tests/test_utils.py

@@ -0,0 +1,63 @@
+# Tests for the anymail/utils.py module
+# (not to be confused with utilities for testing found in in tests/utils.py)
+
+from django.test import SimpleTestCase
+
+from anymail.exceptions import AnymailInvalidAddress
+from anymail.utils import ParsedEmail
+
+
+class ParsedEmailTests(SimpleTestCase):
+    """Test utils.ParsedEmail"""
+
+    # Anymail (and Djrill) have always used EmailMessage.encoding, which defaults to None.
+    # (Django substitutes settings.DEFAULT_ENCODING='utf-8' when converting to a mime message,
+    # but Anymail has never used that code.)
+    ADDRESS_ENCODING = None
+
+    def test_simple_email(self):
+        parsed = ParsedEmail("test@example.com", self.ADDRESS_ENCODING)
+        self.assertEqual(parsed.email, "test@example.com")
+        self.assertEqual(parsed.name, "")
+        self.assertEqual(parsed.address, "test@example.com")
+
+    def test_display_name(self):
+        parsed = ParsedEmail('"Display Name, Inc." <test@example.com>', self.ADDRESS_ENCODING)
+        self.assertEqual(parsed.email, "test@example.com")
+        self.assertEqual(parsed.name, "Display Name, Inc.")
+        self.assertEqual(parsed.address, '"Display Name, Inc." <test@example.com>')
+
+    def test_obsolete_display_name(self):
+        # you can get away without the quotes if there are no commas or parens
+        # (but it's not recommended)
+        parsed = ParsedEmail('Display Name <test@example.com>', self.ADDRESS_ENCODING)
+        self.assertEqual(parsed.email, "test@example.com")
+        self.assertEqual(parsed.name, "Display Name")
+        self.assertEqual(parsed.address, 'Display Name <test@example.com>')
+
+    def test_unicode_display_name(self):
+        parsed = ParsedEmail(u'"Unicode \N{HEAVY BLACK HEART}" <test@example.com>', self.ADDRESS_ENCODING)
+        self.assertEqual(parsed.email, "test@example.com")
+        self.assertEqual(parsed.name, u"Unicode \N{HEAVY BLACK HEART}")
+        # display-name automatically shifts to quoted-printable/base64 for non-ascii chars:
+        self.assertEqual(parsed.address, '=?utf-8?b?VW5pY29kZSDinaQ=?= <test@example.com>')
+
+    def test_invalid_display_name(self):
+        with self.assertRaises(AnymailInvalidAddress):
+            # this parses as multiple email addresses, because of the comma:
+            ParsedEmail('Display Name, Inc. <test@example.com>', self.ADDRESS_ENCODING)
+
+    def test_none_address(self):
+        # used for, e.g., telling Mandrill to use template default from_email
+        parsed = ParsedEmail(None, self.ADDRESS_ENCODING)
+        self.assertEqual(parsed.email, None)
+        self.assertEqual(parsed.name, None)
+        self.assertEqual(parsed.address, None)
+
+    def test_empty_address(self):
+        with self.assertRaises(AnymailInvalidAddress):
+            ParsedEmail('', self.ADDRESS_ENCODING)
+
+    def test_whitespace_only_address(self):
+        with self.assertRaises(AnymailInvalidAddress):
+            ParsedEmail(' ', self.ADDRESS_ENCODING)