Switch to using idna package

hyperlink/_url.py

@@ -20,14 +20,20 @@
 import string
 import socket
 from unicodedata import normalize
-
+try:
+    from socket import inet_pton
+except ImportError:
+    inet_pton = None  # defined below
 try:
     from collections.abc import Mapping
 except ImportError:  # Python 2
     from collections import Mapping
-try:
-    from socket import inet_pton
-except ImportError:
+
+# Note: IDNAError is a subclass of UnicodeError
+from idna import encode as idna_encode, decode as idna_decode, IDNAError
+
+
+if inet_pton is None:
     # based on https://gist.github.com/nnemkin/4966028
     # this code only applies on Windows Python 2.7
     import ctypes
@@ -573,13 +579,59 @@ def _percent_decode(text, normalize_case=False, subencoding='utf-8',
 
 
 def _decode_host(host):
+    """Decode a host from ASCII-encodable text to IDNA-decoded text. If
+    the host text is not ASCII, it is returned unchanged, as it is
+    presumed that it is already IDNA-decoded.
+
+    Some technical details: _decode_host is built on top of the "idna"
+    package, which has some quirks:
+
+    Capital letters are not valid IDNA2008. The idna package will
+    raise an exception like this on capital letters:
+
+    > idna.core.InvalidCodepoint: Codepoint U+004B at position 1 ... not allowed
+
+    However, if a segment of a host (i.e., something in
+    url.host.split('.')) is already ASCII, idna doesn't perform its
+    usual checks. In fact, for capital letters it automatically
+    lowercases them.
+
+    This check and some other functionality can be bypassed by passing
+    uts46=True to idna.encode/decode. This allows a more permissive and
+    convenient interface. So far it seems like the balanced approach.
+
+    Example output (from idna==2.6):
+
+    >> idna.encode(u'mahmöud.io')
+    'xn--mahmud-zxa.io'
+    >> idna.encode(u'Mahmöud.io')
+    Traceback (most recent call last):
+      File "<stdin>", line 1, in <module>
+      File "/home/mahmoud/virtualenvs/hyperlink/local/lib/python2.7/site-packages/idna/core.py", line 355, in encode
+        result.append(alabel(label))
+      File "/home/mahmoud/virtualenvs/hyperlink/local/lib/python2.7/site-packages/idna/core.py", line 276, in alabel
+        check_label(label)
+      File "/home/mahmoud/virtualenvs/hyperlink/local/lib/python2.7/site-packages/idna/core.py", line 253, in check_label
+        raise InvalidCodepoint('Codepoint {0} at position {1} of {2} not allowed'.format(_unot(cp_value), pos+1, repr(label)))
+    idna.core.InvalidCodepoint: Codepoint U+004D at position 1 of u'Mahm\xf6ud' not allowed
+    >> idna.encode(u'Mahmoud.io')
+    'Mahmoud.io'
+
+    # Similar behavior for decodes below
+    >> idna.decode(u'Mahmoud.io')
+    u'mahmoud.io
+    >> idna.decode(u'Méhmoud.io', uts46=True)
+    u'm\xe9hmoud.io'
+    """
+    if not host:
+        return u''
     try:
         host_bytes = host.encode("ascii")
     except UnicodeEncodeError:
         host_text = host
     else:
         try:
-            host_text = host_bytes.decode("idna")
+            host_text = idna_decode(host_bytes, uts46=True)
         except ValueError:
             # only reached on "narrow" (UCS-2) Python builds <3.4, see #7
             # NOTE: not going to raise here, because there's no
@@ -1255,8 +1307,8 @@ def to_uri(self):
 
         For example::
 
-            >>> URL.from_text(u'https://→example.com/foo⇧bar/').to_uri()
-            URL.from_text(u'https://xn--example-dk9c.com/foo%E2%87%A7bar/')
+            >>> URL.from_text(u'https://ايران.com/foo⇧bar/').to_uri()
+            URL.from_text(u'https://xn--mgba3a4fra.com/foo%E2%87%A7bar/')
 
         Returns:
             URL: A new instance with its path segments, query parameters, and
@@ -1267,9 +1319,10 @@ def to_uri(self):
                                   self.userinfo.split(':', 1)])
         new_path = _encode_path_parts(self.path, has_scheme=bool(self.scheme),
                                       rooted=False, joined=False, maximal=True)
+        new_host = self.host if not self.host else idna_encode(self.host, uts46=True).decode("ascii")
         return self.replace(
             userinfo=new_userinfo,
-            host=self.host.encode("idna").decode("ascii"),
+            host=new_host,
             path=new_path,
             query=tuple([tuple(_encode_query_part(x, maximal=True)
                                if x is not None else None
@@ -1285,9 +1338,9 @@ def to_iri(self):
         Percent-encoded Unicode and IDNA-encoded hostnames are
         decoded, like so::
 
-            >>> url = URL.from_text(u'https://xn--example-dk9c.com/foo%E2%87%A7bar/')
+            >>> url = URL.from_text(u'https://xn--mgba3a4fra.example.com/foo%E2%87%A7bar/')
             >>> print(url.to_iri().to_text())
-            https://→example.com/foo⇧bar/
+            https://ايران.example.com/foo⇧bar/
 
         .. note::
 

hyperlink/test/test_url.py

@@ -1167,3 +1167,17 @@ def test_str(self):
         else:
             assert isinstance(str(url), unicode)
             assert isinstance(bytes(url), bytes)
+
+    def test_idna_corners(self):
+        text = u'http://abé.com/'
+        url = URL.from_text(text)
+        assert url.to_iri().host == u'abé.com'
+        assert url.to_uri().host == u'xn--ab-cja.com'
+
+        url = URL.from_text("http://ドメイン.テスト.co.jp#test")
+        assert url.to_iri().host == u'ドメイン.テスト.co.jp'
+        assert url.to_uri().host == u'xn--eckwd4c7c.xn--zckzah.co.jp'
+
+        assert url.to_uri().get_decoded_url().host == u'ドメイン.テスト.co.jp'
+
+        assert URL.from_text('http://Example.com').to_uri().get_decoded_url().host == 'example.com'

requirements-test.txt

@@ -1,3 +1,5 @@
+coverage==4.4.1
+idna==2.5
 pytest==2.9.2
 pytest-cov==2.3.0
 tox==2.6.0

setup.py

@@ -11,15 +11,15 @@
 
 
 __author__ = 'Mahmoud Hashemi and Glyph Lefkowitz'
-__version__ = '17.3.2dev'
+__version__ = '18.0.0dev'
 __contact__ = '[email protected]'
 __url__ = 'https://github.com/python-hyper/hyperlink'
 __license__ = 'MIT'
 
 
 setup(name='hyperlink',
       version=__version__,
-      description="A featureful, correct URL for Python.",
+      description="A featureful, immutable, and correct URL for Python.",
       long_description=__doc__,
       author=__author__,
       author_email=__contact__,
@@ -29,6 +29,7 @@
       zip_safe=False,
       license=__license__,
       platforms='any',
+      install_requires=['idna>=2.5'],
       classifiers=[
           'Topic :: Utilities',
           'Intended Audience :: Developers',

tox.ini

@@ -4,7 +4,7 @@ envlist = py26,py27,py34,py35,py36,pypy,coverage-report,packaging
 [testenv]
 changedir = .tox
 deps = -rrequirements-test.txt
-commands = coverage run --parallel --rcfile {toxinidir}/.tox-coveragerc -m pytest --doctest-modules {envsitepackagesdir}/hyperlink {posargs}
+commands = coverage run --parallel --omit 'flycheck__*' --rcfile {toxinidir}/.tox-coveragerc -m pytest --doctest-modules {envsitepackagesdir}/hyperlink {posargs}
 
 # Uses default basepython otherwise reporting doesn't work on Travis where
 # Python 3.6 is only available in 3.6 jobs.