From a284d69de1d1a42714576d4a9562145a94e62127 Mon Sep 17 00:00:00 2001
From: Ben Kallus <benjamin.p.kallus.gr@dartmouth.edu>
Date: Sat, 12 Nov 2022 15:43:33 -0500
Subject: [PATCH 1/2] gh-99418: Prevent urllib.parse.urlparse from accepting
 schemes that don't begin with an alphabetical ASCII character.

---
 Lib/test/test_urlparse.py                                              |   18 ++++++++++
 Lib/urllib/parse.py                                                    |    7 +++
 Misc/NEWS.d/next/Library/2022-11-12-15-45-51.gh-issue-99418.FxfAXS.rst |    2 +
 3 files changed, 26 insertions(+), 1 deletion(-)

--- a/Lib/test/test_urlparse.py
+++ b/Lib/test/test_urlparse.py
@@ -676,6 +676,24 @@ class UrlParseTestCase(unittest.TestCase
                         with self.assertRaises(ValueError):
                             p.port
 
+    def test_attributes_bad_scheme(self):
+        """Check handling of invalid schemes."""
+        for bytes in (False, True):
+            for parse in (urllib.parse.urlsplit, urllib.parse.urlparse):
+                for scheme in (".", "+", "-", "0", "http&", "६http"):
+                    with self.subTest(bytes=bytes, parse=parse, scheme=scheme):
+                        url = scheme + "://www.example.net"
+                        if bytes:
+                            if urllib.parse.isascii(url):
+                                url = url.encode("ascii")
+                            else:
+                                continue
+                        p = parse(url)
+                        if bytes:
+                            self.assertEqual(p.scheme, b"")
+                        else:
+                            self.assertEqual(p.scheme, "")
+
     def test_attributes_without_netloc(self):
         # This example is straight from RFC 3261.  It looks like it
         # should allow the username, hostname, and port to be filled
--- a/Lib/urllib/parse.py
+++ b/Lib/urllib/parse.py
@@ -35,6 +35,7 @@ __all__ = ["urlparse", "urlunparse", "ur
            "urlsplit", "urlunsplit", "urlencode", "parse_qs",
            "parse_qsl", "quote", "quote_plus", "quote_from_bytes",
            "unquote", "unquote_plus", "unquote_to_bytes",
+           "isascii",
            "DefragResult", "ParseResult", "SplitResult",
            "DefragResultBytes", "ParseResultBytes", "SplitResultBytes"]
 
@@ -79,6 +80,10 @@ scheme_chars = ('abcdefghijklmnopqrstuvw
 # Unsafe bytes to be removed per WHATWG spec
 _UNSAFE_URL_BYTES_TO_REMOVE = ['\t', '\r', '\n']
 
+# Python >= 3.7 shim
+def isascii(word):
+    return all([ord(c) < 128 for c in word])
+
 # XXX: Consider replacing with functools.lru_cache
 MAX_CACHE_SIZE = 20
 _parse_cache = {}
@@ -435,7 +440,7 @@ def urlsplit(url, scheme='', allow_fragm
         clear_cache()
     netloc = query = fragment = ''
     i = url.find(':')
-    if i > 0:
+    if i > 0 and isascii(url[0]) and url[0].isalpha():
         if url[:i] == 'http': # optimize the common case
             scheme = url[:i].lower()
             url = url[i+1:]
--- /dev/null
+++ b/Misc/NEWS.d/next/Library/2022-11-12-15-45-51.gh-issue-99418.FxfAXS.rst
@@ -0,0 +1,2 @@
+Fix bug in :func:`urllib.parse.urlparse` that causes URL schemes that begin
+with a digit, a plus sign, or a minus sign to be parsed incorrectly.