from .common import InfoExtractor
from .adobepass import AdobePassIE
+from .theplatform import ThePlatformIE
from ..utils import (
smuggle_url,
url_basename,
}
-class NationalGeographicIE(AdobePassIE):
+class NationalGeographicIE(ThePlatformIE, AdobePassIE):
IE_NAME = 'natgeo'
_VALID_URL = r'https?://channel\.nationalgeographic\.com/(?:wild/)?[^/]+/(?:videos|episodes)/(?P<id>[^/?]+)'
release_url = self._search_regex(
r'video_auth_playlist_url\s*=\s*"([^"]+)"',
webpage, 'release url')
+ theplatform_path = self._search_regex(r'https?://link.theplatform.com/s/([^?]+)', release_url, 'theplatform path')
+ video_id = theplatform_path.split('/')[-1]
query = {
'mbr': 'true',
- 'switch': 'http',
}
is_auth = self._search_regex(r'video_is_auth\s*=\s*"([^"]+)"', webpage, 'is auth', fatal=False)
if is_auth == 'auth':
auth_resource_id = self._search_regex(
r"video_auth_resourceId\s*=\s*'([^']+)'",
webpage, 'auth resource id')
- query['auth'] = self._extract_mvpd_auth(url, display_id, 'natgeo', auth_resource_id)
-
- return {
- '_type': 'url_transparent',
- 'ie_key': 'ThePlatform',
- 'url': smuggle_url(
- update_url_query(release_url, query),
- {'force_smil_url': True}),
+ query['auth'] = self._extract_mvpd_auth(url, video_id, 'natgeo', auth_resource_id)
+
+ formats = []
+ subtitles = {}
+ for key, value in (('switch', 'http'), ('manifest', 'm3u')):
+ tp_query = query.copy()
+ tp_query.update({
+ key: value,
+ })
+ tp_formats, tp_subtitles = self._extract_theplatform_smil(
+ update_url_query(release_url, tp_query), video_id, 'Downloading %s SMIL data' % value)
+ formats.extend(tp_formats)
+ subtitles = self._merge_subtitles(subtitles, tp_subtitles)
+ self._sort_formats(formats)
+
+ info = self._extract_theplatform_metadata(theplatform_path, display_id)
+ info.update({
+ 'id': video_id,
+ 'formats': formats,
+ 'subtitles': subtitles,
'display_id': display_id,
- }
+ })
+ return info
class NationalGeographicEpisodeGuideIE(InfoExtractor):