[ie/CBCPlayer] Extract HLS formats and subtitles (#7484)

Authored by: trainman261

yt_dlp/extractor/cbc.py

@@ -161,7 +161,7 @@ class CBCPlayerIE(InfoExtractor):
             'upload_date': '20160210',
             'uploader': 'CBCC-NEW',
         },
-        'skip': 'Geo-restricted to Canada',
+        'skip': 'Geo-restricted to Canada and no longer available',
     }, {
         # Redirected from http://www.cbc.ca/player/AudioMobile/All%20in%20a%20Weekend%20Montreal/ID/2657632011/
         'url': 'http://www.cbc.ca/player/play/2657631896',
@@ -174,6 +174,9 @@ class CBCPlayerIE(InfoExtractor):
             'timestamp': 1425704400,
             'upload_date': '20150307',
             'uploader': 'CBCC-NEW',
+            'thumbnail': 'http://thumbnails.cbc.ca/maven_legacy/thumbnails/sonali-karnick-220.jpg',
+            'chapters': [],
+            'duration': 494.811,
         },
     }, {
         'url': 'http://www.cbc.ca/player/play/2164402062',
@@ -186,6 +189,28 @@ class CBCPlayerIE(InfoExtractor):
             'timestamp': 1320410746,
             'upload_date': '20111104',
             'uploader': 'CBCC-NEW',
+            'thumbnail': 'https://thumbnails.cbc.ca/maven_legacy/thumbnails/277/67/cancer_852x480_2164412612.jpg',
+            'chapters': [],
+            'duration': 186.867,
+        },
+    }, {
+        # Has subtitles
+        # These broadcasts expire after ~1 month, can find new test URL here:
+        # https://www.cbc.ca/player/news/TV%20Shows/The%20National/Latest%20Broadcast
+        'url': 'http://www.cbc.ca/player/play/2249992771553',
+        'md5': '2f2fb675dd4f0f8a5bb7588d1b13bacd',
+        'info_dict': {
+            'id': '2249992771553',
+            'ext': 'mp4',
+            'title': 'The National | Women’s soccer pay, Florida seawater, Swift quake',
+            'description': 'md5:adba28011a56cfa47a080ff198dad27a',
+            'timestamp': 1690596000,
+            'duration': 2716.333,
+            'subtitles': {'eng': [{'ext': 'vtt', 'protocol': 'm3u8_native'}]},
+            'thumbnail': 'https://thumbnails.cbc.ca/maven_legacy/thumbnails/481/326/thumbnail.jpeg',
+            'uploader': 'CBCC-NEW',
+            'chapters': 'count:5',
+            'upload_date': '20230729',
         },
     }]
 
@@ -199,6 +224,7 @@ def _real_extract(self, url):
                     'force_smil_url': True
                 }),
             'id': video_id,
+            '_format_sort_fields': ('res', 'proto')  # Prioritize direct http formats over HLS
         }
 
 

yt_dlp/extractor/scrippsnetworks.py

@@ -115,6 +115,7 @@ class ScrippsNetworksIE(InfoExtractor):
             'uploader': 'SCNI-SCND',
         },
         'add_ie': ['ThePlatform'],
+        'expected_warnings': ['No HLS formats found'],
     }, {
         'url': 'https://www.diynetwork.com/videos/diy-barnwood-tablet-stand-0265790',
         'only_matching': True,

yt_dlp/extractor/theplatform.py

@@ -19,7 +19,11 @@
     xpath_with_ns,
     mimetype2ext,
     find_xpath_attr,
+    traverse_obj,
+    update_url,
+    urlhandle_detect_ext,
 )
+from ..networking import HEADRequest
 
 default_ns = 'http://www.w3.org/2005/SMIL21/Language'
 _x = lambda p: xpath_with_ns(p, {'smil': default_ns})
@@ -162,7 +166,8 @@ class ThePlatformIE(ThePlatformBaseIE, AdobePassIE):
         'params': {
             # rtmp download
             'skip_download': True,
-        }
+        },
+        'skip': '404 Not Found',
     }, {
         'url': 'https://player.theplatform.com/p/D6x-PC/pulse_preview/embed/select/media/yMBg9E8KFxZD',
         'info_dict': {
@@ -171,7 +176,8 @@ class ThePlatformIE(ThePlatformBaseIE, AdobePassIE):
             'description': 'md5:644ad9188d655b742f942bf2e06b002d',
             'title': 'HIGHLIGHTS: USA bag first ever series Cup win',
             'uploader': 'EGSM',
-        }
+        },
+        'skip': '404 Not Found',
     }, {
         'url': 'http://player.theplatform.com/p/NnzsPC/widget/select/media/4Y0TlYUr_ZT7',
         'only_matching': True,
@@ -189,6 +195,7 @@ class ThePlatformIE(ThePlatformBaseIE, AdobePassIE):
             'upload_date': '20150701',
             'uploader': 'NBCU-NEWS',
         },
+        'skip': '404 Not Found',
     }, {
         # From http://www.nbc.com/the-blacklist/video/sir-crispin-crandall/2928790?onid=137781#vc137781=1
         # geo-restricted (US), HLS encrypted with AES-128
@@ -295,6 +302,17 @@ def _real_extract(self, url):
 
         formats, subtitles = self._extract_theplatform_smil(smil_url, video_id)
 
+        # With some sites, manifest URL must be forced to extract HLS formats
+        if not traverse_obj(formats, lambda _, v: v['format_id'].startswith('hls')):
+            m3u8_url = update_url(url, query='mbr=true&manifest=m3u', fragment=None)
+            urlh = self._request_webpage(
+                HEADRequest(m3u8_url), video_id, 'Checking for HLS formats', 'No HLS formats found', fatal=False)
+            if urlh and urlhandle_detect_ext(urlh) == 'm3u8':
+                m3u8_fmts, m3u8_subs = self._extract_m3u8_formats_and_subtitles(
+                    m3u8_url, video_id, m3u8_id='hls', fatal=False)
+                formats.extend(m3u8_fmts)
+                self._merge_subtitles(m3u8_subs, target=subtitles)
+
         ret = self._extract_theplatform_metadata(path, video_id)
         combined_subtitles = self._merge_subtitles(ret.get('subtitles', {}), subtitles)
         ret.update({