Merge 34bf80c62d into be3579aaf0

test/test_utils.py

@@ -161,6 +161,11 @@ class TestUtil(unittest.TestCase):
         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', keep_bad_win_chars=True, is_id=False))
+        self.assertEqual('xxx_<>\\*|', sanitize_filename('xxx/<>\\*|', keep_bad_win_chars=True, is_id=False))
+        self.assertEqual('yes? no', sanitize_filename('yes? no', keep_bad_win_chars=True, is_id=False))
+        self.assertEqual('this: that', sanitize_filename('this: that', keep_bad_win_chars=True, is_id=False))
+
         self.assertEqual(sanitize_filename('AT&T'), 'AT&T')
         aumlaut = 'ä'
         self.assertEqual(sanitize_filename(aumlaut), aumlaut)
@@ -171,6 +176,10 @@ class TestUtil(unittest.TestCase):
             sanitize_filename('New World record at 0:12:34'),
             'New World record at 0_12_34')
 
+        self.assertEqual(
+            sanitize_filename('New World record at 0:12:34', keep_bad_win_chars=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')
@@ -222,6 +231,10 @@ class TestUtil(unittest.TestCase):
         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', keep_bad_win_chars=True, is_id=True), '_n_cd26wFpw')
+        self.assertEqual(sanitize_filename('_BD_eEpuzXw', keep_bad_win_chars=True, is_id=True), '_BD_eEpuzXw')
+        self.assertEqual(sanitize_filename('N0Y__7-UOdI', keep_bad_win_chars=True, is_id=True), 'N0Y__7-UOdI')
+
     def test_sanitize_path(self):
         with unittest.mock.patch('sys.platform', 'win32'):
             self._test_sanitize_path()

yt_dlp/YoutubeDL.py

@@ -1311,7 +1311,9 @@ class YoutubeDL:
         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.get('keep_bad_win_chars', False), restricted=restricted,
+                is_id=(
                     bool(re.search(r'(^|[_.])id(\.|$)', key))
                     if 'filename-sanitization' in self.params['compat_opts']
                     else NO_DEFAULT))

yt_dlp/__init__.py

@@ -833,6 +833,7 @@ def parse_options(argv=None):
         'autonumber_start': opts.autonumber_start,
         'restrictfilenames': opts.restrictfilenames,
         'windowsfilenames': opts.windowsfilenames,
+        'keep_bad_win_chars': opts.keep_bad_win_chars,
         'ignoreerrors': opts.ignoreerrors,
         'force_generic_extractor': opts.force_generic_extractor,
         'allowed_extractors': opts.allowed_extractors or ['default'],

yt_dlp/extractor/common.py

@@ -1025,7 +1025,7 @@ class InfoExtractor:
         if len(basen) > trim_length:
             h = '___' + hashlib.md5(basen.encode()).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('keep_bad_win_chars', False), 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/options.py

@@ -267,6 +267,10 @@ def create_parser():
             out_dict[key] = [*out_dict.get(key, []), val] if append else val
         setattr(parser.values, option.dest, out_dict)
 
+    def _store_multiple_callback(option, opt_str, value, parser, values):
+        for key, value in values.items():
+            setattr(parser.values, key, value)
+
     def when_prefix(default):
         return {
             'default': {},
@@ -1372,7 +1376,13 @@ def create_parser():
         help='Force filenames to be Windows-compatible')
     filesystem.add_option(
         '--no-windows-filenames',
-        action='store_false', dest='windowsfilenames',
+        action='callback', dest='keep_bad_win_chars', default=False, callback=_store_multiple_callback,
+        callback_kwargs={
+            'values': {
+                'windowsfilenames': False,
+                'keep_bad_win_chars': True
+            }
+        },
         help='Make filenames Windows-compatible only if using Windows (default)')
     filesystem.add_option(
         '--trim-filenames', '--trim-file-names', metavar='LENGTH',

yt_dlp/utils/_utils.py

@@ -626,8 +626,9 @@ def timeconvert(timestr):
     return timestamp
 
 
-def sanitize_filename(s, restricted=False, is_id=NO_DEFAULT):
+def sanitize_filename(s, keep_bad_win_chars=False, restricted=False, is_id=NO_DEFAULT):
     """Sanitizes a string so it could be used as part of a filename.
+    @param keep_bad_win_chars    Whether to keep characters invalid on Windows
     @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
@@ -640,16 +641,16 @@ def sanitize_filename(s, restricted=False, is_id=NO_DEFAULT):
             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 not keep_bad_win_chars:
             # 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 (not keep_bad_win_chars and char == '?') or ord(char) < 32 or ord(char) == 127:
             return ''
-        elif char == '"':
+        elif not keep_bad_win_chars and char == '"':
             return '' if restricted else '\''
-        elif char == ':':
+        elif not keep_bad_win_chars and char == ':':
             return '\0_\0-' if restricted else '\0 \0-'
-        elif char in '\\/|*<>':
+        elif (not keep_bad_win_chars and char in '\\|*<>') or char == '/':
             return '\0_'
         if restricted and (char in '!&\'()[]{}$;`^,#' or char.isspace() or ord(char) > 127):
             return '' if unicodedata.category(char)[0] in 'CM' else '\0_'
@@ -658,7 +659,8 @@ def sanitize_filename(s, restricted=False, is_id=NO_DEFAULT):
     # 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) if keep_bad_win_chars
+               else m.group(0).replace(':', '_'), 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