mirror of
https://github.com/ytdl-org/youtube-dl.git
synced 2024-11-25 08:10:31 +01:00
[discoverygo:playlist] Add extractor (closes #12424)
This commit is contained in:
parent
4605c94d1a
commit
7049799470
@ -1,17 +1,21 @@
|
||||
from __future__ import unicode_literals
|
||||
|
||||
import re
|
||||
|
||||
from .common import InfoExtractor
|
||||
from ..compat import compat_str
|
||||
from ..utils import (
|
||||
extract_attributes,
|
||||
ExtractorError,
|
||||
int_or_none,
|
||||
parse_age_limit,
|
||||
ExtractorError,
|
||||
remove_end,
|
||||
unescapeHTML,
|
||||
)
|
||||
|
||||
|
||||
class DiscoveryGoIE(InfoExtractor):
|
||||
_VALID_URL = r'''(?x)https?://(?:www\.)?(?:
|
||||
class DiscoveryGoBaseIE(InfoExtractor):
|
||||
_VALID_URL_TEMPLATE = r'''(?x)https?://(?:www\.)?(?:
|
||||
discovery|
|
||||
investigationdiscovery|
|
||||
discoverylife|
|
||||
@ -21,7 +25,11 @@ class DiscoveryGoIE(InfoExtractor):
|
||||
sciencechannel|
|
||||
tlc|
|
||||
velocitychannel
|
||||
)go\.com/(?:[^/]+/)*(?P<id>[^/?#&]+)'''
|
||||
)go\.com/%s(?P<id>[^/?#&]+)'''
|
||||
|
||||
|
||||
class DiscoveryGoIE(DiscoveryGoBaseIE):
|
||||
_VALID_URL = DiscoveryGoBaseIE._VALID_URL_TEMPLATE % r'(?:[^/]+/)+'
|
||||
_TEST = {
|
||||
'url': 'https://www.discoverygo.com/love-at-first-kiss/kiss-first-ask-questions-later/',
|
||||
'info_dict': {
|
||||
@ -113,3 +121,46 @@ class DiscoveryGoIE(InfoExtractor):
|
||||
'formats': formats,
|
||||
'subtitles': subtitles,
|
||||
}
|
||||
|
||||
|
||||
class DiscoveryGoPlaylistIE(DiscoveryGoBaseIE):
|
||||
_VALID_URL = DiscoveryGoBaseIE._VALID_URL_TEMPLATE % ''
|
||||
_TEST = {
|
||||
'url': 'https://www.discoverygo.com/bering-sea-gold/',
|
||||
'info_dict': {
|
||||
'id': 'bering-sea-gold',
|
||||
'title': 'Bering Sea Gold',
|
||||
'description': 'md5:cc5c6489835949043c0cc3ad66c2fa0e',
|
||||
},
|
||||
'playlist_mincount': 6,
|
||||
}
|
||||
|
||||
@classmethod
|
||||
def suitable(cls, url):
|
||||
return False if DiscoveryGoIE.suitable(url) else super(
|
||||
DiscoveryGoPlaylistIE, cls).suitable(url)
|
||||
|
||||
def _real_extract(self, url):
|
||||
display_id = self._match_id(url)
|
||||
|
||||
webpage = self._download_webpage(url, display_id)
|
||||
|
||||
entries = []
|
||||
for mobj in re.finditer(r'data-json=(["\'])(?P<json>{.+?})\1', webpage):
|
||||
data = self._parse_json(
|
||||
mobj.group('json'), display_id,
|
||||
transform_source=unescapeHTML, fatal=False)
|
||||
if not isinstance(data, dict) or data.get('type') != 'episode':
|
||||
continue
|
||||
episode_url = data.get('socialUrl')
|
||||
if not episode_url:
|
||||
continue
|
||||
entries.append(self.url_result(
|
||||
episode_url, ie=DiscoveryGoIE.ie_key(),
|
||||
video_id=data.get('id')))
|
||||
|
||||
return self.playlist_result(
|
||||
entries, display_id,
|
||||
remove_end(self._og_search_title(
|
||||
webpage, fatal=False), ' | Discovery GO'),
|
||||
self._og_search_description(webpage))
|
||||
|
@ -265,7 +265,10 @@ from .dvtv import DVTVIE
|
||||
from .dumpert import DumpertIE
|
||||
from .defense import DefenseGouvFrIE
|
||||
from .discovery import DiscoveryIE
|
||||
from .discoverygo import DiscoveryGoIE
|
||||
from .discoverygo import (
|
||||
DiscoveryGoIE,
|
||||
DiscoveryGoPlaylistIE,
|
||||
)
|
||||
from .disney import DisneyIE
|
||||
from .dispeak import DigitallySpeakingIE
|
||||
from .dropbox import DropboxIE
|
||||
|
Loading…
Reference in New Issue
Block a user