[extractor/common] Properly extract audio only formats in master m3u8 playlists

[youtube-dl] / youtube_dl / extractor / common.py

diff --git a/youtube_dl/extractor/common.py b/youtube_dl/extractor/common.py
index f411ea7633568915e4e97df61958d13c6a8aca80..51351fb57c95cc18637ed6c126d4ff894680784e 100644 (file)
--- a/youtube_dl/extractor/common.py
+++ b/youtube_dl/extractor/common.py
@@ -46,6 +46,7 @@ from ..utils import (
     xpath_with_ns,
     determine_protocol,
     parse_duration,
+    mimetype2ext,
 )
 
 
@@ -899,6 +900,16 @@ class InfoExtractor(object):
                     item='%s video format' % f.get('format_id') if f.get('format_id') else 'video'),
                 formats)
 
+    @staticmethod
+    def _remove_duplicate_formats(formats):
+        format_urls = set()
+        unique_formats = []
+        for f in formats:
+            if f['url'] not in format_urls:
+                format_urls.add(f['url'])
+                unique_formats.append(f)
+        formats[:] = unique_formats
+
     def _is_valid_url(self, url, video_id, item='video'):
         url = self._proto_relative_url(url, scheme='http:')
         # For now assume non HTTP(S) URLs always valid
@@ -1073,19 +1084,29 @@ class InfoExtractor(object):
                     'protocol': entry_protocol,
                     'preference': preference,
                 }
-                codecs = last_info.get('CODECS')
-                if codecs:
-                    # TODO: looks like video codec is not always necessarily goes first
-                    va_codecs = codecs.split(',')
-                    if va_codecs[0]:
-                        f['vcodec'] = va_codecs[0]
-                    if len(va_codecs) > 1 and va_codecs[1]:
-                        f['acodec'] = va_codecs[1]
                 resolution = last_info.get('RESOLUTION')
                 if resolution:
                     width_str, height_str = resolution.split('x')
                     f['width'] = int(width_str)
                     f['height'] = int(height_str)
+                codecs = last_info.get('CODECS')
+                if codecs:
+                    vcodec, acodec = [None] * 2
+                    va_codecs = codecs.split(',')
+                    if len(va_codecs) == 1:
+                        # Audio only entries usually come with single codec and
+                        # no resolution. For more robustness we also check it to
+                        # be mp4 audio.
+                        if not resolution and va_codecs[0].startswith('mp4a'):
+                            vcodec, acodec = 'none', va_codecs[0]
+                        else:
+                            vcodec = va_codecs[0]
+                    else:
+                        vcodec, acodec = va_codecs[:2]
+                    f.update({
+                        'acodec': acodec,
+                        'vcodec': vcodec,
+                    })
                 if last_media is not None:
                     f['m3u8_media'] = last_media
                     last_media = None
@@ -1277,16 +1298,7 @@ class InfoExtractor(object):
             if not src or src in urls:
                 continue
             urls.append(src)
-            ext = textstream.get('ext') or determine_ext(src)
-            if not ext:
-                type_ = textstream.get('type')
-                SUBTITLES_TYPES = {
-                    'text/vtt': 'vtt',
-                    'text/srt': 'srt',
-                    'application/smptett+xml': 'tt',
-                }
-                if type_ in SUBTITLES_TYPES:
-                    ext = SUBTITLES_TYPES[type_]
+            ext = textstream.get('ext') or determine_ext(src) or mimetype2ext(textstream.get('type'))
             lang = textstream.get('systemLanguage') or textstream.get('systemLanguageName') or textstream.get('lang') or subtitles_lang
             subtitles.setdefault(lang, []).append({
                 'url': src,