Improve performance of sub-authority splitting in URL

Author: sethmlarson

src/urllib3/util/url.py

@@ -63,12 +63,12 @@
 BRACELESS_IPV6_ADDRZ_RE = re.compile("^" + IPV6_ADDRZ_PAT[2:-2] + "$")
 ZONE_ID_RE = re.compile("(" + ZONE_ID_PAT + r")\]$")
 
-SUBAUTHORITY_PAT = (u"^(?:(.*)@)?(%s|%s|%s)(?::([0-9]{0,5}))?$") % (
+_HOST_PORT_PAT = ("^(%s|%s|%s)(?::([0-9]{0,5}))?$") % (
     REG_NAME_PAT,
     IPV4_PAT,
     IPV6_ADDRZ_PAT,
 )
-SUBAUTHORITY_RE = re.compile(SUBAUTHORITY_PAT, re.UNICODE | re.DOTALL)
+_HOST_PORT_RE = re.compile(_HOST_PORT_PAT, re.UNICODE | re.DOTALL)
 
 UNRESERVED_CHARS = set(
     "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789._-~"
@@ -365,7 +365,9 @@ def parse_url(url):
             scheme = scheme.lower()
 
         if authority:
-            auth, host, port = SUBAUTHORITY_RE.match(authority).groups()
+            auth, _, host_port = authority.rpartition("@")
+            auth = auth or None
+            host, port = _HOST_PORT_RE.match(host_port).groups()
             if auth and normalize_uri:
                 auth = _encode_invalid_chars(auth, USERINFO_CHARS)
             if port == "":

test/test_util.py

@@ -438,6 +438,16 @@ def test_netloc(self, url, expected_netloc):
                 fragment="hash",
             ),
         ),
+        # Tons of '@' causing backtracking
+        ("https://" + ("@" * 10000) + "[", False),
+        (
+            "https://user:" + ("@" * 10000) + "example.com",
+            Url(
+                scheme="https",
+                auth="user:" + ("%40" * 9999),
+                host="example.com",
+            ),
+        ),
     ]
 
     @pytest.mark.parametrize("url, expected_url", url_vulnerabilities)