Normalize RFC 4007 IPv6 Zone IDs

Author: sethmlarson

src/rfc3986/abnf_regexp.py

@@ -115,14 +115,14 @@
     UNRESERVED_RE + SUB_DELIMITERS_RE + ':'
 )
 
-
 # RFC 6874 Zone ID ABNF
 ZONE_ID = '(?:[' + UNRESERVED_RE + ']|' + PCT_ENCODED + ')+'
-IPv6_ADDRZ_RE = IPv6_RE + '%25' + ZONE_ID
 
-IP_LITERAL_RE = r'\[({0}|(?:{1})|{2})\]'.format(
-    IPv6_RE,
-    IPv6_ADDRZ_RE,
+IPv6_ADDRZ_RFC4007_RE = IPv6_RE + '(?:(?:%25|%)' + ZONE_ID + ')?'
+IPv6_ADDRZ_RE = IPv6_RE + '(?:%25' + ZONE_ID + ')?'
+
+IP_LITERAL_RE = r'\[({0}|{1})\]'.format(
+    IPv6_ADDRZ_RFC4007_RE,
     IPv_FUTURE_RE,
 )
 

src/rfc3986/misc.py

@@ -59,6 +59,12 @@
 
 
 IPv4_MATCHER = re.compile('^' + abnf_regexp.IPv4_RE + '$')
+IPv6_MATCHER = re.compile(r'^\[' + abnf_regexp.IPv6_ADDRZ_RFC4007_RE + r'\]$')
+
+# Used by host validator
+IPv6_NO_RFC4007_MATCHER = re.compile(r'^\[%s\]$' % (
+    abnf_regexp.IPv6_ADDRZ_RE
+))
 
 # Matcher used to validate path components
 PATH_MATCHER = re.compile(abnf_regexp.PATH_RE)

src/rfc3986/normalizers.py

@@ -49,6 +49,21 @@ def normalize_password(password):
 
 def normalize_host(host):
     """Normalize a host string."""
+    if misc.IPv6_MATCHER.match(host):
+        percent = host.find('%')
+        if percent != -1:
+            percent_25 = host.find('%25')
+
+            # Replace RFC 4007 IPv6 Zone ID delimiter '%' with '%25'
+            # from RFC 6874. If the host is '[<IPv6 addr>%25]' then we
+            # assume RFC 4007 and normalize to '[<IPV6 addr>%2525]'
+            if percent_25 == -1 or percent < percent_25 or \
+                    (percent == percent_25 and percent_25 == len(host) - 4):
+                host = host.replace('%', '%25', 1)
+
+            # Don't normalize the casing of the Zone ID
+            return host[:percent].lower() + host[percent:]
+
     return host.lower()
 
 

src/rfc3986/uri.py

@@ -171,6 +171,12 @@ def authority_info(self):
             # valid bytes, it is an InvalidAuthority.
             raise exc.InvalidAuthority(self.authority.encode(self.encoding))
 
+        if (host and misc.IPv6_MATCHER.match(host) and not
+                misc.IPv6_NO_RFC4007_MATCHER.match(host)):
+            # If it's an IPv6 address that has RFC 4007 IPv6
+            # Zone IDs then it's invalid.
+            raise exc.InvalidAuthority(self.authority.encode(self.encoding))
+
         return matches
 
     @property

src/rfc3986/validators.py

@@ -306,6 +306,8 @@ def authority_is_valid(authority, host=None, require=False):
     validated = is_valid(authority, misc.SUBAUTHORITY_MATCHER, require)
     if validated and host is not None and misc.IPv4_MATCHER.match(host):
         return valid_ipv4_host_address(host)
+    elif validated and host is not None and misc.IPv6_MATCHER.match(host):
+        return misc.IPv6_NO_RFC4007_MATCHER.match(host) is not None
     return validated
 
 

tests/conftest.py

@@ -12,6 +12,8 @@
     '[::1%25lo]',  # With ZoneID
     '[FF02:0:0:0:0:0:0:2%25en01]',  # With ZoneID
     '[FF02:30:0:0:0:0:0:5%25en1]',  # With ZoneID
+    '[FF02:30:0:0:0:0:0:5%25%26]',  # With ZoneID
+    '[FF02:30:0:0:0:0:0:5%2525]',  # With ZoneID
     '[21DA:D3:0:2F3B:2AA:FF:FE28:9C5A]',
     '[FE80::2AA:FF:FE9A:4CA2]',
     '[FF02::2]',
@@ -30,9 +32,9 @@
     '[FF02::3::5]',  # IPv6 can only have one ::
     '[FADF:01]',  # Not properly compacted (missing a :)
     '[FADF:01%en0]',  # Not properly compacted (missing a :), Invalid ZoneID
-    '[FADF::01%en0]',  # Invalid ZoneID separator
+    '[FADF::01%en0]',  # ZoneID is per RFC 4007
     '[FADF::01%]',  # Invalid ZoneID separator and no ZoneID
-    '[FADF::01%25]',  # Missing ZoneID
+    '[FADF::01%25]',  # Missing ZoneID in RFC 6974, is 25 in RFC 4007
     'localhost:80:80:80',  # Too many ports
     '256.256.256.256',  # Invalid IPv4 Address
     SNOWMAN.decode('utf-8')

tests/test_normalizers.py

@@ -4,7 +4,7 @@
 from rfc3986.uri import URIReference
 from rfc3986.normalizers import (
     normalize_scheme, normalize_percent_characters,
-    remove_dot_segments, encode_component,
+    remove_dot_segments, encode_component, normalize_host
     )
 
 
@@ -93,3 +93,20 @@ def test_fragment_normalization():
 )
 def test_detect_percent_encoded_component(component, encoded_component):
     assert encode_component(component, 'utf-8') == encoded_component
+
+
+@pytest.mark.parametrize(
+    ["host", "normalized_host"],
+    [
+    ('LOCALHOST', 'localhost'),
+    ('[::1%eth0]', '[::1%25eth0]'),
+    ('[::1%25]', '[::1%2525]'),
+    ('[::1%%25]', '[::1%25%25]'),
+    ('[::1%25%25]', '[::1%25%25]'),
+    ('[::Af%Ff]', '[::af%25Ff]'),
+    ('[::Af%%Ff]', '[::af%25%Ff]'),
+    ('[::Af%25Ff]', '[::af%25Ff]'),
+    ]
+)
+def test_normalize_host(host, normalized_host):
+    assert normalize_host(host) == normalized_host