Make URL parsing in mimetypes.guess_type() more reliable.

python · serhiy-storchaka · Mar 26, 2024 · Sep 5, 2019 · Mar 25, 2024 · Mar 25, 2024
commit caa728db554a91ca13739c1b4a0cdd52671243a7
@@ -121,7 +121,12 @@ def guess_type(self, url, strict=True):
         """
         url = os.fspath(url)
         p = urllib.parse.urlparse(url)
-        scheme, url = p.scheme, p.path
+        if p.scheme and len(p.scheme) > 1:
+            scheme = p.scheme
+            url = p.path
+        else:
+            scheme = None
+            url = os.path.splitdrive(url)[1]
         if scheme == 'data':
             # syntax of data URLs:
             # dataurl   := "data:" [ mediatype ] [ ";base64" ] "," data

@@ -1,5 +1,6 @@
 import io
 import mimetypes
+import os
 import pathlib
 import sys
 import unittest.mock
@@ -109,13 +110,26 @@ def test_filename_with_url_delimiters(self):
         # compared to when interpreted as filename because of the semicolon.
         eq = self.assertEqual
         gzip_expected = ('application/x-tar', 'gzip')
-        eq(self.db.guess_type(";1.tar.gz"), gzip_expected)
-        eq(self.db.guess_type("?1.tar.gz"), gzip_expected)
-        eq(self.db.guess_type("#1.tar.gz"), gzip_expected)
-        eq(self.db.guess_type("#1#.tar.gz"), gzip_expected)
-        eq(self.db.guess_type(";1#.tar.gz"), gzip_expected)
-        eq(self.db.guess_type(";&1=123;?.tar.gz"), gzip_expected)
-        eq(self.db.guess_type("?k1=v1&k2=v2.tar.gz"), gzip_expected)
+        for name in (
+                ';1.tar.gz',
+                '?1.tar.gz',
+                '#1.tar.gz',
+                '#1#.tar.gz',
+                ';1#.tar.gz',
+                ';&1=123;?.tar.gz',
+                '?k1=v1&k2=v2.tar.gz',
+            ):
+            for prefix in ('', '/', '\\',
+                           'c:', 'c:/', 'c:\\', 'c:/d/', 'c:\\d\\',
+                           '//share/server/', '\\\\share\\server\\'):
+                path = prefix + name
+                with self.subTest(path=path):
+                    eq(self.db.guess_type(path), gzip_expected)
+            expected = (None, None) if os.name == 'nt' else gzip_expected
+            for prefix in ('//', '\\\\', '//share/', '\\\\share\\'):
+                path = prefix + name
+                with self.subTest(path=path):
+                    eq(self.db.guess_type(path), expected)
         eq(self.db.guess_type(r" \"\`;b&b&c |.tar.gz"), gzip_expected)
 
     def test_url(self):
@@ -125,6 +139,10 @@ def test_url(self):
         result = self.db.guess_type('http://example.com/host.html')
         msg = 'Should be text/html'
         self.assertSequenceEqual(result, ('text/html', None), msg)
+        result = self.db.guess_type('http://example.com/host.html#x.tar')
+        self.assertSequenceEqual(result, ('text/html', None))
+        result = self.db.guess_type('http://example.com/host.html?q=x.tar')
+        self.assertSequenceEqual(result, ('text/html', None))
 
     def test_guess_all_types(self):
         # First try strict.  Use a set here for testing the results because if

diff --git a/Misc/NEWS.d/next/Library/2019-08-27-01-03-26.bpo-22347._TRpYr.rst b/Misc/NEWS.d/next/Library/2019-08-27-01-03-26.bpo-22347._TRpYr.rst
diff --git a/Misc/NEWS.d/next/Library/2019-08-27-01-03-26.gh-66543._TRpYr.rst b/Misc/NEWS.d/next/Library/2019-08-27-01-03-26.gh-66543._TRpYr.rst
@@ -0,0 +1,4 @@
+Make :func:`mimetypes.guess_type` properly parsing of URLs with only a host
+name, URLs containing fragment or query, and filenames with only a UNC
+sharepoint on Windows.
+Based on patch by Dong-hee Na.