[yahoo] fix single video extraction

This commit is contained in:
Remita Amine 2021-01-19 14:47:39 +01:00
parent 015e19b350
commit bc7c8f3d4e
1 changed files with 40 additions and 40 deletions

View File

@ -177,46 +177,9 @@ class YahooIE(InfoExtractor):
'only_matching': True, 'only_matching': True,
}] }]
def _real_extract(self, url): def _extract_yahoo_video(self, video_id, country):
url, country, display_id = re.match(self._VALID_URL, url).groups()
if not country:
country = 'us'
else:
country = country.split('-')[0]
api_base = 'https://%s.yahoo.com/_td/api/resource/' % country
for i, uuid in enumerate(['url=' + url, 'ymedia-alias=' + display_id]):
content = self._download_json(
api_base + 'content;getDetailView=true;uuids=["%s"]' % uuid,
display_id, 'Downloading content JSON metadata', fatal=i == 1)
if content:
item = content['items'][0]
break
if item.get('type') != 'video':
entries = []
cover = item.get('cover') or {}
if cover.get('type') == 'yvideo':
cover_url = cover.get('url')
if cover_url:
entries.append(self.url_result(
cover_url, 'Yahoo', cover.get('uuid')))
for e in item.get('body', []):
if e.get('type') == 'videoIframe':
iframe_url = e.get('url')
if not iframe_url:
continue
entries.append(self.url_result(iframe_url))
return self.playlist_result(
entries, item.get('uuid'),
item.get('title'), item.get('summary'))
video_id = item['uuid']
video = self._download_json( video = self._download_json(
api_base + 'VideoService.videos;view=full;video_ids=["%s"]' % video_id, 'https://%s.yahoo.com/_td/api/resource/VideoService.videos;view=full;video_ids=["%s"]' % (country, video_id),
video_id, 'Downloading video JSON metadata')[0] video_id, 'Downloading video JSON metadata')[0]
title = video['title'] title = video['title']
@ -298,7 +261,6 @@ class YahooIE(InfoExtractor):
'id': video_id, 'id': video_id,
'title': self._live_title(title) if is_live else title, 'title': self._live_title(title) if is_live else title,
'formats': formats, 'formats': formats,
'display_id': display_id,
'thumbnails': thumbnails, 'thumbnails': thumbnails,
'description': clean_html(video.get('description')), 'description': clean_html(video.get('description')),
'timestamp': parse_iso8601(video.get('publish_time')), 'timestamp': parse_iso8601(video.get('publish_time')),
@ -311,6 +273,44 @@ class YahooIE(InfoExtractor):
'episode_number': int_or_none(series_info.get('episode_number')), 'episode_number': int_or_none(series_info.get('episode_number')),
} }
def _real_extract(self, url):
url, country, display_id = re.match(self._VALID_URL, url).groups()
if not country:
country = 'us'
else:
country = country.split('-')[0]
item = self._download_json(
'https://%s.yahoo.com/caas/content/article' % country, display_id,
'Downloading content JSON metadata', query={
'url': url
})['items'][0]['data']['partnerData']
if item.get('type') != 'video':
entries = []
cover = item.get('cover') or {}
if cover.get('type') == 'yvideo':
cover_url = cover.get('url')
if cover_url:
entries.append(self.url_result(
cover_url, 'Yahoo', cover.get('uuid')))
for e in (item.get('body') or []):
if e.get('type') == 'videoIframe':
iframe_url = e.get('url')
if not iframe_url:
continue
entries.append(self.url_result(iframe_url))
return self.playlist_result(
entries, item.get('uuid'),
item.get('title'), item.get('summary'))
info = self._extract_yahoo_video(item['uuid'], country)
info['display_id'] = display_id
return info
class YahooSearchIE(SearchInfoExtractor): class YahooSearchIE(SearchInfoExtractor):
IE_DESC = 'Yahoo screen search' IE_DESC = 'Yahoo screen search'