Merge 1cae3bf46d into c699bafc50

Closes #11537

Authored by: bashonly

test/test_traversal.py

@@ -481,7 +481,7 @@ class TestTraversalHelpers:
             'id': 'name',
             'data': 'content',
             'url': 'url',
-        }, all, {subs_list_to_dict}]) == {
+        }, all, {subs_list_to_dict(lang=None)}]) == {
             'de': [{'url': 'https://example.com/subs/de.ass'}],
             'en': [{'data': 'content'}],
         }, 'subs with mandatory items missing should be filtered'
@@ -507,6 +507,54 @@ class TestTraversalHelpers:
             {'url': 'https://example.com/subs/en1', 'ext': 'ext'},
             {'url': 'https://example.com/subs/en2', 'ext': 'ext'},
         ]}, '`quality` key should sort subtitle list accordingly'
+        assert traverse_obj([
+            {'name': 'de', 'url': 'https://example.com/subs/de.ass'},
+            {'name': 'de'},
+            {'name': 'en', 'content': 'content'},
+            {'url': 'https://example.com/subs/en'},
+        ], [..., {
+            'id': 'name',
+            'url': 'url',
+            'data': 'content',
+        }, all, {subs_list_to_dict(lang='en')}]) == {
+            'de': [{'url': 'https://example.com/subs/de.ass'}],
+            'en': [
+                {'data': 'content'},
+                {'url': 'https://example.com/subs/en'},
+            ],
+        }, 'optionally provided lang should be used if no id available'
+        assert traverse_obj([
+            {'name': 1, 'url': 'https://example.com/subs/de1'},
+            {'name': {}, 'url': 'https://example.com/subs/de2'},
+            {'name': 'de', 'ext': 1, 'url': 'https://example.com/subs/de3'},
+            {'name': 'de', 'ext': {}, 'url': 'https://example.com/subs/de4'},
+        ], [..., {
+            'id': 'name',
+            'url': 'url',
+            'ext': 'ext',
+        }, all, {subs_list_to_dict(lang=None)}]) == {
+            'de': [
+                {'url': 'https://example.com/subs/de3'},
+                {'url': 'https://example.com/subs/de4'},
+            ],
+        }, 'non str types should be ignored for id and ext'
+        assert traverse_obj([
+            {'name': 1, 'url': 'https://example.com/subs/de1'},
+            {'name': {}, 'url': 'https://example.com/subs/de2'},
+            {'name': 'de', 'ext': 1, 'url': 'https://example.com/subs/de3'},
+            {'name': 'de', 'ext': {}, 'url': 'https://example.com/subs/de4'},
+        ], [..., {
+            'id': 'name',
+            'url': 'url',
+            'ext': 'ext',
+        }, all, {subs_list_to_dict(lang='de')}]) == {
+            'de': [
+                {'url': 'https://example.com/subs/de1'},
+                {'url': 'https://example.com/subs/de2'},
+                {'url': 'https://example.com/subs/de3'},
+                {'url': 'https://example.com/subs/de4'},
+            ],
+        }, 'non str types should be replaced by default id'
 
     def test_trim_str(self):
         with pytest.raises(TypeError):
@@ -525,7 +573,7 @@ class TestTraversalHelpers:
     def test_unpack(self):
         assert unpack(lambda *x: ''.join(map(str, x)))([1, 2, 3]) == '123'
         assert unpack(join_nonempty)([1, 2, 3]) == '1-2-3'
-        assert unpack(join_nonempty(delim=' '))([1, 2, 3]) == '1 2 3'
+        assert unpack(join_nonempty, delim=' ')([1, 2, 3]) == '1 2 3'
         with pytest.raises(TypeError):
             unpack(join_nonempty)()
         with pytest.raises(TypeError):

test/test_utils.py

@@ -72,7 +72,6 @@ from yt_dlp.utils import (
     intlist_to_bytes,
     iri_to_uri,
     is_html,
-    join_nonempty,
     js_to_json,
     limit_length,
     locked_file,
@@ -2158,10 +2157,6 @@ Line 1
         assert int_or_none(v=10) == 10, 'keyword passed positional should call function'
         assert int_or_none(scale=0.1)(10) == 100, 'call after partial application should call the function'
 
-        assert callable(join_nonempty(delim=', ')), 'varargs positional should apply partially'
-        assert callable(join_nonempty()), 'varargs positional should apply partially'
-        assert join_nonempty(None, delim=', ') == '', 'passed varargs should call the function'
-
 
 if __name__ == '__main__':
     unittest.main()

yt_dlp/YoutubeDL.py

@@ -4381,7 +4381,9 @@ class YoutubeDL:
             return None
 
         for idx, t in list(enumerate(thumbnails))[::-1]:
-            thumb_ext = (f'{t["id"]}.' if multiple else '') + determine_ext(t['url'], 'jpg')
+            thumb_ext = t.get('ext') or determine_ext(t['url'], 'jpg')
+            if multiple:
+                thumb_ext = f'{t["id"]}.{thumb_ext}'
             thumb_display_id = f'{label} thumbnail {t["id"]}'
             thumb_filename = replace_extension(filename, thumb_ext, info_dict.get('ext'))
             thumb_filename_final = replace_extension(thumb_filename_base, thumb_ext, info_dict.get('ext'))

yt_dlp/extractor/afreecatv.py

@@ -66,6 +66,14 @@ class AfreecaTVBaseIE(InfoExtractor):
             extensions={'legacy_ssl': True}), display_id,
             'Downloading API JSON', 'Unable to download API JSON')
 
+    @staticmethod
+    def _fixup_thumb(thumb_url):
+        if not url_or_none(thumb_url):
+            return None
+        # Core would determine_ext as 'php' from the url, so we need to provide the real ext
+        # See: https://github.com/yt-dlp/yt-dlp/issues/11537
+        return [{'url': thumb_url, 'ext': 'jpg'}]
+
 
 class AfreecaTVIE(AfreecaTVBaseIE):
     IE_NAME = 'soop'
@@ -155,7 +163,7 @@ class AfreecaTVIE(AfreecaTVBaseIE):
             'uploader': ('writer_nick', {str}),
             'uploader_id': ('bj_id', {str}),
             'duration': ('total_file_duration', {int_or_none(scale=1000)}),
-            'thumbnail': ('thumb', {url_or_none}),
+            'thumbnails': ('thumb', {self._fixup_thumb}),
         })
 
         entries = []
@@ -226,8 +234,7 @@ class AfreecaTVCatchStoryIE(AfreecaTVBaseIE):
 
         return self.playlist_result(self._entries(data), video_id)
 
-    @staticmethod
-    def _entries(data):
+    def _entries(self, data):
         # 'files' is always a list with 1 element
         yield from traverse_obj(data, (
             'data', lambda _, v: v['story_type'] == 'catch',
@@ -238,7 +245,7 @@ class AfreecaTVCatchStoryIE(AfreecaTVBaseIE):
                 'title': ('title', {str}),
                 'uploader': ('writer_nick', {str}),
                 'uploader_id': ('writer_id', {str}),
-                'thumbnail': ('thumb', {url_or_none}),
+                'thumbnails': ('thumb', {self._fixup_thumb}),
                 'timestamp': ('write_timestamp', {int_or_none}),
             }))
 

yt_dlp/extractor/common.py

@@ -279,6 +279,7 @@ class InfoExtractor:
     thumbnails:     A list of dictionaries, with the following entries:
                         * "id" (optional, string) - Thumbnail format ID
                         * "url"
+                        * "ext" (optional, string) - actual image extension if not given in URL
                         * "preference" (optional, int) - quality of the image
                         * "width" (optional, int)
                         * "height" (optional, int)

yt_dlp/postprocessor/ffmpeg.py

@@ -220,9 +220,20 @@ class FFmpegPostProcessor(PostProcessor):
     @staticmethod
     def stream_copy_opts(copy=True, *, ext=None):
         yield from ('-map', '0')
+
+        if ext in ('mkv', 'mka'):
+            # Some streams, such as JSON attachments, are considered of unknown
+            # type by FFmpeg but we still want to copy them.
+            yield '-copy_unknown'
+        else:
+            # Most containers don't really like unknown streams. Let's make
+            # sure to get rid of them.
+            yield '-ignore_unknown'
+
         # Don't copy Apple TV chapters track, bin_data
         # See https://github.com/yt-dlp/yt-dlp/issues/2, #19042, #19024, https://trac.ffmpeg.org/ticket/6016
-        yield from ('-dn', '-ignore_unknown')
+        yield '-dn'
+
         if copy:
             yield from ('-c', 'copy')
         if ext in ('mp4', 'mov', 'm4a'):
@@ -557,7 +568,7 @@ class FFmpegVideoConvertorPP(FFmpegPostProcessor):
 
     @staticmethod
     def _options(target_ext):
-        yield from FFmpegPostProcessor.stream_copy_opts(False)
+        yield from FFmpegPostProcessor.stream_copy_opts(False, ext=target_ext)
         if target_ext == 'avi':
             yield from ('-c:v', 'libxvid', '-vtag', 'XVID')
 
@@ -583,7 +594,7 @@ class FFmpegVideoRemuxerPP(FFmpegVideoConvertorPP):
 
     @staticmethod
     def _options(target_ext):
-        return FFmpegPostProcessor.stream_copy_opts()
+        return FFmpegPostProcessor.stream_copy_opts(ext=target_ext)
 
 
 class FFmpegEmbedSubtitlePP(FFmpegPostProcessor):
@@ -620,13 +631,18 @@ class FFmpegEmbedSubtitlePP(FFmpegPostProcessor):
         webm_vtt_warn = False
         mp4_ass_warn = False
 
+        json_subs = {}
+
         for lang, sub_info in subtitles.items():
             if not os.path.exists(sub_info.get('filepath', '')):
                 self.report_warning(f'Skipping embedding {lang} subtitle because the file is missing')
                 continue
             sub_ext = sub_info['ext']
             if sub_ext == 'json':
-                self.report_warning('JSON subtitles cannot be embedded')
+                if info['ext'] in ('mkv', 'mka'):
+                    json_subs[lang] = sub_info['filepath']
+                else:
+                    self.report_warning('JSON subtitles can only be embedded in mkv/mka files.')
             elif ext != 'webm' or ext == 'webm' and sub_ext == 'vtt':
                 sub_langs.append(lang)
                 sub_names.append(sub_info.get('name'))
@@ -639,31 +655,48 @@ class FFmpegEmbedSubtitlePP(FFmpegPostProcessor):
                 mp4_ass_warn = True
                 self.report_warning('ASS subtitles cannot be properly embedded in mp4 files; expect issues')
 
-        if not sub_langs:
+        if not sub_langs and not json_subs:
             return [], info
 
         input_files = [filename, *sub_filenames]
 
-        opts = [
-            *self.stream_copy_opts(ext=info['ext']),
-            # Don't copy the existing subtitles, we may be running the
-            # postprocessor a second time
-            '-map', '-0:s',
-        ]
+        opts = [*self.stream_copy_opts(ext=info['ext'])]
+
+        if sub_langs and sub_names:
+            # We have regular subtitles available to embed. Don't copy the
+            # existing subtitles, we may be running the postprocessor a second
+            # time.
+            opts.extend([
+                '-map', '-0:s',
+            ])
+
         for i, (lang, name) in enumerate(zip(sub_langs, sub_names)):
-            opts.extend(['-map', f'{i + 1}:0'])
             lang_code = ISO639Utils.short2long(lang) or lang
-            opts.extend([f'-metadata:s:s:{i}', f'language={lang_code}'])
+            opts.extend([
+                '-map', f'{i + 1}:0',
+                f'-metadata:s:s:{i}', f'language={lang_code}',
+            ])
+
             if name:
                 opts.extend([f'-metadata:s:s:{i}', f'handler_name={name}',
                              f'-metadata:s:s:{i}', f'title={name}'])
 
+        for json_lang, json_filename in json_subs.items():
+            escaped_json_filename = self._ffmpeg_filename_argument(json_filename)
+            json_basename = os.path.basename(json_filename)
+            opts.extend([
+                '-map', f'-0:m:filename:{json_lang}.json?',
+                '-attach', escaped_json_filename,
+                f'-metadata:s:m:filename:{json_basename}', 'mimetype=application/json',
+                f'-metadata:s:m:filename:{json_basename}', f'filename={json_lang}.json',
+            ])
+
         temp_filename = prepend_extension(filename, 'temp')
         self.to_screen(f'Embedding subtitles in "{filename}"')
         self.run_ffmpeg_multiple_files(input_files, temp_filename, opts)
         os.replace(temp_filename, filename)
 
-        files_to_delete = [] if self._already_have_subtitle else sub_filenames
+        files_to_delete = [] if self._already_have_subtitle else [*sub_filenames, *json_subs.values()]
         return files_to_delete, info
 
 
@@ -678,7 +711,7 @@ class FFmpegMetadataPP(FFmpegPostProcessor):
     @staticmethod
     def _options(target_ext):
         audio_only = target_ext == 'm4a'
-        yield from FFmpegPostProcessor.stream_copy_opts(not audio_only)
+        yield from FFmpegPostProcessor.stream_copy_opts(not audio_only, ext=target_ext)
         if audio_only:
             yield from ('-vn', '-acodec', 'copy')
 
@@ -806,15 +839,20 @@ class FFmpegMetadataPP(FFmpegPostProcessor):
             write_json_file(self._downloader.sanitize_info(info, self.get_param('clean_infojson', True)), infofn)
             info['infojson_filename'] = infofn
 
-        old_stream, new_stream = self.get_stream_number(info['filepath'], ('tags', 'mimetype'), 'application/json')
-        if old_stream is not None:
-            yield ('-map', f'-0:{old_stream}')
-            new_stream -= 1
+        escaped_name = self._ffmpeg_filename_argument(infofn)
+        info_basename = os.path.basename(infofn)
 
         yield (
-            '-attach', self._ffmpeg_filename_argument(infofn),
-            f'-metadata:s:{new_stream}', 'mimetype=application/json',
-            f'-metadata:s:{new_stream}', 'filename=info.json',
+            # In order to override any old info.json reliably we need to
+            # instruct FFmpeg to consider valid tracks without a codec id, like
+            # JSON attachments.
+            '-copy_unknown',
+            # This map operation allows us to actually replace any previous
+            # info.json data.
+            '-map', '-0:m:filename:info.json?',
+            '-attach', escaped_name,
+            f'-metadata:s:m:filename:{info_basename}', 'mimetype=application/json',
+            f'-metadata:s:m:filename:{info_basename}', 'filename=info.json',
         )
 
 
@@ -873,7 +911,7 @@ class FFmpegFixupStretchedPP(FFmpegFixupPostProcessor):
         stretched_ratio = info.get('stretched_ratio')
         if stretched_ratio not in (None, 1):
             self._fixup('Fixing aspect ratio', info['filepath'], [
-                *self.stream_copy_opts(), '-aspect', f'{stretched_ratio:f}'])
+                *self.stream_copy_opts(ext=info['ext']), '-aspect', f'{stretched_ratio:f}'])
         return [], info
 
 
@@ -881,7 +919,7 @@ class FFmpegFixupM4aPP(FFmpegFixupPostProcessor):
     @PostProcessor._restrict_to(images=False, video=False)
     def run(self, info):
         if info.get('container') == 'm4a_dash':
-            self._fixup('Correcting container', info['filepath'], [*self.stream_copy_opts(), '-f', 'mp4'])
+            self._fixup('Correcting container', info['filepath'], [*self.stream_copy_opts(ext=info['ext']), '-f', 'mp4'])
         return [], info
 
 
@@ -904,7 +942,7 @@ class FFmpegFixupM3u8PP(FFmpegFixupPostProcessor):
             if self.get_audio_codec(info['filepath']) == 'aac':
                 args.extend(['-bsf:a', 'aac_adtstoasc'])
             self._fixup('Fixing MPEG-TS in MP4 container', info['filepath'], [
-                *self.stream_copy_opts(), *args])
+                *self.stream_copy_opts(ext=info['ext']), *args])
         return [], info
 
 
@@ -925,7 +963,7 @@ class FFmpegFixupTimestampPP(FFmpegFixupPostProcessor):
             opts = ['-vf', 'setpts=PTS-STARTPTS']
         else:
             opts = ['-c', 'copy', '-bsf', 'setts=ts=TS-STARTPTS']
-        self._fixup('Fixing frame timestamp', info['filepath'], [*opts, *self.stream_copy_opts(False), '-ss', self.trim])
+        self._fixup('Fixing frame timestamp', info['filepath'], [*opts, *self.stream_copy_opts(False, ext=info['ext']), '-ss', self.trim])
         return [], info
 
 
@@ -934,7 +972,7 @@ class FFmpegCopyStreamPP(FFmpegFixupPostProcessor):
 
     @PostProcessor._restrict_to(images=False)
     def run(self, info):
-        self._fixup(self.MESSAGE, info['filepath'], self.stream_copy_opts())
+        self._fixup(self.MESSAGE, info['filepath'], self.stream_copy_opts(ext=info['ext']))
         return [], info
 
 
@@ -1063,7 +1101,7 @@ class FFmpegSplitChaptersPP(FFmpegPostProcessor):
         self.to_screen(f'Splitting video by chapters; {len(chapters)} chapters found')
         for idx, chapter in enumerate(chapters):
             destination, opts = self._ffmpeg_args_for_chapter(idx + 1, chapter, info)
-            self.real_run_ffmpeg([(in_file, opts)], [(destination, self.stream_copy_opts())])
+            self.real_run_ffmpeg([(in_file, opts)], [(destination, self.stream_copy_opts(ext=info['ext']))])
         if in_file != info['filepath']:
             self._delete_downloaded_files(in_file, msg=None)
         return [], info

yt_dlp/utils/_utils.py

@@ -216,7 +216,7 @@ def partial_application(func):
     sig = inspect.signature(func)
     required_args = [
         param.name for param in sig.parameters.values()
-        if param.kind in (inspect.Parameter.POSITIONAL_ONLY, inspect.Parameter.POSITIONAL_OR_KEYWORD, inspect.Parameter.VAR_POSITIONAL)
+        if param.kind in (inspect.Parameter.POSITIONAL_ONLY, inspect.Parameter.POSITIONAL_OR_KEYWORD)
         if param.default is inspect.Parameter.empty
     ]
 
@@ -4837,7 +4837,6 @@ def number_of_digits(number):
     return len('%d' % number)
 
 
-@partial_application
 def join_nonempty(*values, delim='-', from_dict=None):
     if from_dict is not None:
         values = (traversal.traverse_obj(from_dict, variadic(v)) for v in values)

yt_dlp/utils/traversal.py

@@ -332,14 +332,14 @@ class _RequiredError(ExtractorError):
 
 
 @typing.overload
-def subs_list_to_dict(*, ext: str | None = None) -> collections.abc.Callable[[list[dict]], dict[str, list[dict]]]: ...
+def subs_list_to_dict(*, lang: str | None = 'und', ext: str | None = None) -> collections.abc.Callable[[list[dict]], dict[str, list[dict]]]: ...
 
 
 @typing.overload
-def subs_list_to_dict(subs: list[dict] | None, /, *, ext: str | None = None) -> dict[str, list[dict]]: ...
+def subs_list_to_dict(subs: list[dict] | None, /, *, lang: str | None = 'und', ext: str | None = None) -> dict[str, list[dict]]: ...
 
 
-def subs_list_to_dict(subs: list[dict] | None = None, /, *, ext=None):
+def subs_list_to_dict(subs: list[dict] | None = None, /, *, lang='und', ext=None):
     """
     Convert subtitles from a traversal into a subtitle dict.
     The path should have an `all` immediately before this function.
@@ -352,7 +352,7 @@ def subs_list_to_dict(subs: list[dict] | None = None, /, *, ext=None):
     `quality`  The sort order for each subtitle
     """
     if subs is None:
-        return functools.partial(subs_list_to_dict, ext=ext)
+        return functools.partial(subs_list_to_dict, lang=lang, ext=ext)
 
     result = collections.defaultdict(list)
 
@@ -360,10 +360,16 @@ def subs_list_to_dict(subs: list[dict] | None = None, /, *, ext=None):
         if not url_or_none(sub.get('url')) and not sub.get('data'):
             continue
         sub_id = sub.pop('id', None)
-        if sub_id is None:
-            continue
-        if ext is not None and not sub.get('ext'):
-            sub['ext'] = ext
+        if not isinstance(sub_id, str):
+            if not lang:
+                continue
+            sub_id = lang
+        sub_ext = sub.get('ext')
+        if not isinstance(sub_ext, str):
+            if not ext:
+                sub.pop('ext', None)
+            else:
+                sub['ext'] = ext
         result[sub_id].append(sub)
     result = dict(result)
 
@@ -452,9 +458,9 @@ def trim_str(*, start=None, end=None):
     return trim
 
 
-def unpack(func):
+def unpack(func, **kwargs):
     @functools.wraps(func)
-    def inner(items, **kwargs):
+    def inner(items):
         return func(*items, **kwargs)
 
     return inner