unescapeHTML,
ExtractorError,
int_or_none,
+ mimetype2ext,
)
+from .nbc import NBCSportsVPlayerIE
+
class YahooIE(InfoExtractor):
IE_DESC = 'Yahoo screen and movies'
- _VALID_URL = r'(?P<url>(?P<host>https?://(?:[a-zA-Z]{2}\.)?[\da-zA-Z_-]+\.yahoo\.com)/(?:[^/]+/)*(?P<display_id>.+?)-(?P<id>[0-9]+)(?:-[a-z]+)?\.html)'
+ _VALID_URL = r'(?P<url>(?P<host>https?://(?:[a-zA-Z]{2}\.)?[\da-zA-Z_-]+\.yahoo\.com)/(?:[^/]+/)*(?P<display_id>.+)?-(?P<id>[0-9]+)(?:-[a-z]+)?\.html)'
_TESTS = [
{
'url': 'http://screen.yahoo.com/julian-smith-travis-legg-watch-214727115.html',
'description': 'md5:df390f70a9ba7c95ff1daace988f0d8d',
'title': 'Tyler Kalinoski hits buzzer-beater to lift Davidson',
}
+ }, {
+ 'url': 'https://tw.news.yahoo.com/-100120367.html',
+ 'only_matching': True,
}
]
def _real_extract(self, url):
mobj = re.match(self._VALID_URL, url)
- display_id = mobj.group('display_id')
+ display_id = mobj.group('display_id') or self._match_id(url)
page_id = mobj.group('id')
url = mobj.group('url')
host = mobj.group('host')
video_id = items[0]['id']
return self._get_info(video_id, display_id, webpage)
# Look for NBCSports iframes
- iframe_m = re.search(
- r'<iframe[^>]+src="(?P<url>https?://vplayer\.nbcsports\.com/[^"]+)"', webpage)
- if iframe_m:
- return self.url_result(iframe_m.group('url'), 'NBCSports')
+ nbc_sports_url = NBCSportsVPlayerIE._extract_url(webpage)
+ if nbc_sports_url:
+ return self.url_result(nbc_sports_url, 'NBCSportsVPlayer')
items_json = self._search_regex(
r'mediaItems: ({.*?})$', webpage, 'items', flags=re.MULTILINE,
self._sort_formats(formats)
+ closed_captions = self._html_search_regex(
+ r'"closedcaptions":(\[[^\]]+\])', webpage, 'closed captions',
+ default='[]')
+
+ cc_json = self._parse_json(closed_captions, video_id, fatal=False)
+ subtitles = {}
+ if cc_json:
+ for closed_caption in cc_json:
+ lang = closed_caption['lang']
+ if lang not in subtitles:
+ subtitles[lang] = []
+ subtitles[lang].append({
+ 'url': closed_caption['url'],
+ 'ext': mimetype2ext(closed_caption['content_type']),
+ })
+
return {
'id': video_id,
'display_id': display_id,
'description': clean_html(meta['description']),
'thumbnail': meta['thumbnail'] if meta.get('thumbnail') else self._og_search_thumbnail(webpage),
'duration': int_or_none(meta.get('duration')),
+ 'subtitles': subtitles,
}