fix: don't replace characters invalid on Windows when --no-windows-filenames is passed

test/test_utils.py

@@ -146,80 +146,93 @@ def test_timeconvert(self):
         self.assertTrue(timeconvert('bougrg') is None)
 
     def test_sanitize_filename(self):
-        self.assertEqual(sanitize_filename(''), '')
-        self.assertEqual(sanitize_filename('abc'), 'abc')
-        self.assertEqual(sanitize_filename('abc_d-e'), 'abc_d-e')
+        self.assertEqual(sanitize_filename('', use_win_filenames=True), '')
+        self.assertEqual(sanitize_filename('abc', use_win_filenames=True), 'abc')
+        self.assertEqual(sanitize_filename('abc_d-e', use_win_filenames=True), 'abc_d-e')
 
-        self.assertEqual(sanitize_filename('123'), '123')
+        self.assertEqual(sanitize_filename('123', use_win_filenames=True), '123')
 
-        self.assertEqual('abc⧸de', sanitize_filename('abc/de'))
-        self.assertFalse('/' in sanitize_filename('abc/de///'))
+        self.assertEqual('abc⧸de', sanitize_filename('abc/de', use_win_filenames=True))
+        self.assertFalse('/' in sanitize_filename('abc/de///', use_win_filenames=True))
 
-        self.assertEqual('abc_de', sanitize_filename('abc/<>\\*|de', is_id=False))
-        self.assertEqual('xxx', sanitize_filename('xxx/<>\\*|', is_id=False))
-        self.assertEqual('yes no', sanitize_filename('yes? no', is_id=False))
-        self.assertEqual('this - that', sanitize_filename('this: that', is_id=False))
+        self.assertEqual('abc_de', sanitize_filename('abc/<>\\*|de', use_win_filenames=True, is_id=False))
+        self.assertEqual('xxx', sanitize_filename('xxx/<>\\*|', use_win_filenames=True, is_id=False))
+        self.assertEqual('yes no', sanitize_filename('yes? no', use_win_filenames=True, is_id=False))
+        self.assertEqual('this - that', sanitize_filename('this: that', use_win_filenames=True, is_id=False))
 
-        self.assertEqual(sanitize_filename('AT&T'), 'AT&T')
+        self.assertEqual('abc_<>\\*|de', sanitize_filename('abc/<>\\*|de', use_win_filenames=False, is_id=False))
+        self.assertEqual('xxx_<>\\*|', sanitize_filename('xxx/<>\\*|', use_win_filenames=False, is_id=False))
+        self.assertEqual('yes? no', sanitize_filename('yes? no', use_win_filenames=False, is_id=False))
+        self.assertEqual('this: that', sanitize_filename('this: that', use_win_filenames=False, is_id=False))
+
+        self.assertEqual(sanitize_filename('AT&T', use_win_filenames=True), 'AT&T')
         aumlaut = 'ä'
-        self.assertEqual(sanitize_filename(aumlaut), aumlaut)
+        self.assertEqual(sanitize_filename(aumlaut, use_win_filenames=True), aumlaut)
         tests = '\u043a\u0438\u0440\u0438\u043b\u043b\u0438\u0446\u0430'
-        self.assertEqual(sanitize_filename(tests), tests)
+        self.assertEqual(sanitize_filename(tests, use_win_filenames=True), tests)
 
         self.assertEqual(
-            sanitize_filename('New World record at 0:12:34'),
+            sanitize_filename('New World record at 0:12:34', use_win_filenames=True),
             'New World record at 0_12_34')
 
-        self.assertEqual(sanitize_filename('--gasdgf'), '--gasdgf')
-        self.assertEqual(sanitize_filename('--gasdgf', is_id=True), '--gasdgf')
-        self.assertEqual(sanitize_filename('--gasdgf', is_id=False), '_-gasdgf')
-        self.assertEqual(sanitize_filename('.gasdgf'), '.gasdgf')
-        self.assertEqual(sanitize_filename('.gasdgf', is_id=True), '.gasdgf')
-        self.assertEqual(sanitize_filename('.gasdgf', is_id=False), 'gasdgf')
+        self.assertEqual(
+            sanitize_filename('New World record at 0:12:34', use_win_filenames=False),
+            'New World record at 0:12:34')
+
+        self.assertEqual(sanitize_filename('--gasdgf', use_win_filenames=True), '--gasdgf')
+        self.assertEqual(sanitize_filename('--gasdgf', use_win_filenames=True, is_id=True), '--gasdgf')
+        self.assertEqual(sanitize_filename('--gasdgf', use_win_filenames=True, is_id=False), '_-gasdgf')
+        self.assertEqual(sanitize_filename('.gasdgf', use_win_filenames=True), '.gasdgf')
+        self.assertEqual(sanitize_filename('.gasdgf', use_win_filenames=True, is_id=True), '.gasdgf')
+        self.assertEqual(sanitize_filename('.gasdgf', use_win_filenames=True, is_id=False), 'gasdgf')
 
         forbidden = '"\0\\/'
         for fc in forbidden:
             for fbc in forbidden:
-                self.assertTrue(fbc not in sanitize_filename(fc))
+                self.assertTrue(fbc not in sanitize_filename(fc, use_win_filenames=True))
 
     def test_sanitize_filename_restricted(self):
-        self.assertEqual(sanitize_filename('abc', restricted=True), 'abc')
-        self.assertEqual(sanitize_filename('abc_d-e', restricted=True), 'abc_d-e')
+        self.assertEqual(sanitize_filename('abc', use_win_filenames=True, restricted=True), 'abc')
+        self.assertEqual(sanitize_filename('abc_d-e', use_win_filenames=True, restricted=True), 'abc_d-e')
 
-        self.assertEqual(sanitize_filename('123', restricted=True), '123')
+        self.assertEqual(sanitize_filename('123', use_win_filenames=True, restricted=True), '123')
 
-        self.assertEqual('abc_de', sanitize_filename('abc/de', restricted=True))
-        self.assertFalse('/' in sanitize_filename('abc/de///', restricted=True))
+        self.assertEqual('abc_de', sanitize_filename('abc/de', use_win_filenames=True, restricted=True))
+        self.assertFalse('/' in sanitize_filename('abc/de///', use_win_filenames=True, restricted=True))
 
-        self.assertEqual('abc_de', sanitize_filename('abc/<>\\*|de', restricted=True))
-        self.assertEqual('xxx', sanitize_filename('xxx/<>\\*|', restricted=True))
-        self.assertEqual('yes_no', sanitize_filename('yes? no', restricted=True))
-        self.assertEqual('this_-_that', sanitize_filename('this: that', restricted=True))
+        self.assertEqual('abc_de', sanitize_filename('abc/<>\\*|de', use_win_filenames=True, restricted=True))
+        self.assertEqual('xxx', sanitize_filename('xxx/<>\\*|', use_win_filenames=True, restricted=True))
+        self.assertEqual('yes_no', sanitize_filename('yes? no', use_win_filenames=True, restricted=True))
+        self.assertEqual('this_-_that', sanitize_filename('this: that', use_win_filenames=True ,restricted=True))
 
         tests = 'aäb\u4e2d\u56fd\u7684c'
-        self.assertEqual(sanitize_filename(tests, restricted=True), 'aab_c')
-        self.assertTrue(sanitize_filename('\xf6', restricted=True) != '')  # No empty filename
+        self.assertEqual(sanitize_filename(tests, use_win_filenames=True, restricted=True), 'aab_c')
+        self.assertTrue(sanitize_filename('\xf6', use_win_filenames=True, restricted=True) != '')  # No empty filename
 
         forbidden = '"\0\\/&!: \'\t\n()[]{}$;`^,#'
         for fc in forbidden:
             for fbc in forbidden:
-                self.assertTrue(fbc not in sanitize_filename(fc, restricted=True))
+                self.assertTrue(fbc not in sanitize_filename(fc, use_win_filenames=True, restricted=True))
 
         # Handle a common case more neatly
-        self.assertEqual(sanitize_filename('\u5927\u58f0\u5e26 - Song', restricted=True), 'Song')
-        self.assertEqual(sanitize_filename('\u603b\u7edf: Speech', restricted=True), 'Speech')
+        self.assertEqual(sanitize_filename('\u5927\u58f0\u5e26 - Song', use_win_filenames=True, restricted=True), 'Song')
+        self.assertEqual(sanitize_filename('\u603b\u7edf: Speech', use_win_filenames=True, restricted=True), 'Speech')
         # .. but make sure the file name is never empty
-        self.assertTrue(sanitize_filename('-', restricted=True) != '')
-        self.assertTrue(sanitize_filename(':', restricted=True) != '')
+        self.assertTrue(sanitize_filename('-', use_win_filenames=True, restricted=True) != '')
+        self.assertTrue(sanitize_filename(':', use_win_filenames=True, restricted=True) != '')
 
         self.assertEqual(sanitize_filename(
-            'ÂÃÄÀÁÅÆÇÈÉÊËÌÍÎÏÐÑÒÓÔÕÖŐØŒÙÚÛÜŰÝÞßàáâãäåæçèéêëìíîïðñòóôõöőøœùúûüűýþÿ', restricted=True),
+            'ÂÃÄÀÁÅÆÇÈÉÊËÌÍÎÏÐÑÒÓÔÕÖŐØŒÙÚÛÜŰÝÞßàáâãäåæçèéêëìíîïðñòóôõöőøœùúûüűýþÿ', use_win_filenames=True ,restricted=True),
             'AAAAAAAECEEEEIIIIDNOOOOOOOOEUUUUUYTHssaaaaaaaeceeeeiiiionooooooooeuuuuuythy')
 
     def test_sanitize_ids(self):
-        self.assertEqual(sanitize_filename('_n_cd26wFpw', is_id=True), '_n_cd26wFpw')
-        self.assertEqual(sanitize_filename('_BD_eEpuzXw', is_id=True), '_BD_eEpuzXw')
-        self.assertEqual(sanitize_filename('N0Y__7-UOdI', is_id=True), 'N0Y__7-UOdI')
+        self.assertEqual(sanitize_filename('_n_cd26wFpw', use_win_filenames=True, is_id=True), '_n_cd26wFpw')
+        self.assertEqual(sanitize_filename('_BD_eEpuzXw', use_win_filenames=True, is_id=True), '_BD_eEpuzXw')
+        self.assertEqual(sanitize_filename('N0Y__7-UOdI', use_win_filenames=True, is_id=True), 'N0Y__7-UOdI')
+
+        self.assertEqual(sanitize_filename('_n_cd26wFpw', use_win_filenames=False, is_id=True), '_n_cd26wFpw')
+        self.assertEqual(sanitize_filename('_BD_eEpuzXw', use_win_filenames=False, is_id=True), '_BD_eEpuzXw')
+        self.assertEqual(sanitize_filename('N0Y__7-UOdI', use_win_filenames=False, is_id=True), 'N0Y__7-UOdI')
 
     def test_sanitize_path(self):
         if sys.platform != 'win32':

yt_dlp/YoutubeDL.py

@@ -1254,7 +1254,8 @@ def get_value(mdict):
         na = self.params.get('outtmpl_na_placeholder', 'NA')
 
         def filename_sanitizer(key, value, restricted=self.params.get('restrictfilenames')):
-            return sanitize_filename(str(value), restricted=restricted, is_id=(
+            return sanitize_filename(
+                str(value), self.params["windowsfilenames"], restricted=restricted, is_id=(
                 bool(re.search(r'(^|[_.])id(\.|$)', key))
                 if 'filename-sanitization' in self.params['compat_opts']
                 else NO_DEFAULT))

yt_dlp/extractor/common.py

@@ -958,7 +958,7 @@ def _request_dump_filename(self, url, video_id):
         if len(basen) > trim_length:
             h = '___' + hashlib.md5(basen.encode('utf-8')).hexdigest()
             basen = basen[:trim_length - len(h)] + h
-        filename = sanitize_filename(f'{basen}.dump', restricted=True)
+        filename = sanitize_filename(f'{basen}.dump', self.get_param("windowsfilenames"), restricted=True)
         # Working around MAX_PATH limitation on Windows (see
         # http://msdn.microsoft.com/en-us/library/windows/desktop/aa365247(v=vs.85).aspx)
         if compat_os_name == 'nt':

yt_dlp/utils/_utils.py

@@ -612,11 +612,12 @@ def timeconvert(timestr):
     return timestamp
 
 
-def sanitize_filename(s, restricted=False, is_id=NO_DEFAULT):
+def sanitize_filename(s, use_win_filenames, restricted=False, is_id=NO_DEFAULT):
     """Sanitizes a string so it could be used as part of a filename.
-    @param restricted   Use a stricter subset of allowed characters
-    @param is_id        Whether this is an ID that should be kept unchanged if possible.
-                        If unset, yt-dlp's new sanitization rules are in effect
+    @param use_win_filenames    Whether to use Windows filename restrictions
+    @param restricted           Use a stricter subset of allowed characters
+    @param is_id                Whether this is an ID that should be kept unchanged if possible.
+                                If unset, yt-dlp's new sanitization rules are in effect
     """
     if s == '':
         return ''
@@ -626,16 +627,16 @@ def replace_insane(char):
             return ACCENT_CHARS[char]
         elif not restricted and char == '\n':
             return '\0 '
-        elif is_id is NO_DEFAULT and not restricted and char in '"*:<>?|/\\':
+        elif is_id is NO_DEFAULT and not restricted and char in '"*:<>?|/\\' and use_win_filenames:
             # Replace with their full-width unicode counterparts
             return {'/': '\u29F8', '\\': '\u29f9'}.get(char, chr(ord(char) + 0xfee0))
-        elif char == '?' or ord(char) < 32 or ord(char) == 127:
+        elif (use_win_filenames and char == '?') or ord(char) < 32 or ord(char) == 127:
             return ''
-        elif char == '"':
+        elif use_win_filenames and char == '"':
             return '' if restricted else '\''
-        elif char == ':':
+        elif use_win_filenames and char == ':':
             return '\0_\0-' if restricted else '\0 \0-'
-        elif char in '\\/|*<>':
+        elif (use_win_filenames and char in '\\|*<>') or char == "/":
             return '\0_'
         if restricted and (char in '!&\'()[]{}$;`^,#' or char.isspace() or ord(char) > 127):
             return '\0_'
@@ -644,7 +645,8 @@ def replace_insane(char):
     # Replace look-alike Unicode glyphs
     if restricted and (is_id is NO_DEFAULT or not is_id):
         s = unicodedata.normalize('NFKC', s)
-    s = re.sub(r'[0-9]+(?::[0-9]+)+', lambda m: m.group(0).replace(':', '_'), s)  # Handle timestamps
+    s = re.sub(r'[0-9]+(?::[0-9]+)+', lambda m: m.group(0).replace(':', '_')
+               if use_win_filenames else m.group(0), s)  # Handle timestamps
     result = ''.join(map(replace_insane, s))
     if is_id is NO_DEFAULT:
         result = re.sub(r'(\0.)(?:(?=\1)..)+', r'\1', result)  # Remove repeated substitute chars