mirror of
https://github.com/ytdl-org/youtube-dl.git
synced 2024-11-25 11:41:52 +00:00
[zdf] Add support for channels
This commit is contained in:
parent
a81bbebf44
commit
8560c61842
2 changed files with 83 additions and 4 deletions
|
@ -526,7 +526,7 @@ from .youtube import (
|
||||||
YoutubeUserIE,
|
YoutubeUserIE,
|
||||||
YoutubeWatchLaterIE,
|
YoutubeWatchLaterIE,
|
||||||
)
|
)
|
||||||
from .zdf import ZDFIE
|
from .zdf import ZDFIE, ZDFChannelIE
|
||||||
from .zingmp3 import (
|
from .zingmp3 import (
|
||||||
ZingMp3SongIE,
|
ZingMp3SongIE,
|
||||||
ZingMp3AlbumIE,
|
ZingMp3AlbumIE,
|
||||||
|
|
|
@ -86,6 +86,26 @@ def extract_from_xml_url(ie, video_id, xml_url):
|
||||||
}
|
}
|
||||||
|
|
||||||
|
|
||||||
|
def extract_channel_from_xml_url(ie, channel_id, xml_url):
|
||||||
|
doc = ie._download_xml(
|
||||||
|
xml_url, channel_id,
|
||||||
|
note='Downloading channel info',
|
||||||
|
errnote='Failed to download channel info')
|
||||||
|
|
||||||
|
title = doc.find('.//information/title').text
|
||||||
|
description = doc.find('.//information/detail').text
|
||||||
|
assets = [{'id': asset.find('./details/assetId').text,
|
||||||
|
'type': asset.find('./type').text,
|
||||||
|
} for asset in doc.findall('.//teasers/teaser')]
|
||||||
|
|
||||||
|
return {
|
||||||
|
'id': channel_id,
|
||||||
|
'title': title,
|
||||||
|
'description': description,
|
||||||
|
'assets': assets,
|
||||||
|
}
|
||||||
|
|
||||||
|
|
||||||
class ZDFIE(InfoExtractor):
|
class ZDFIE(InfoExtractor):
|
||||||
_VALID_URL = r'^https?://www\.zdf\.de/ZDFmediathek(?P<hash>#)?/(.*beitrag/(?:video/)?)(?P<id>[0-9]+)(?:/[^/?]+)?(?:\?.*)?'
|
_VALID_URL = r'^https?://www\.zdf\.de/ZDFmediathek(?P<hash>#)?/(.*beitrag/(?:video/)?)(?P<id>[0-9]+)(?:/[^/?]+)?(?:\?.*)?'
|
||||||
|
|
||||||
|
@ -104,8 +124,67 @@ class ZDFIE(InfoExtractor):
|
||||||
'skip': 'Videos on ZDF.de are depublicised in short order',
|
'skip': 'Videos on ZDF.de are depublicised in short order',
|
||||||
}
|
}
|
||||||
|
|
||||||
def _real_extract(self, url):
|
def _extract_video(self, video_id):
|
||||||
video_id = self._match_id(url)
|
|
||||||
|
|
||||||
xml_url = 'http://www.zdf.de/ZDFmediathek/xmlservice/web/beitragsDetails?ak=web&id=%s' % video_id
|
xml_url = 'http://www.zdf.de/ZDFmediathek/xmlservice/web/beitragsDetails?ak=web&id=%s' % video_id
|
||||||
return extract_from_xml_url(self, video_id, xml_url)
|
return extract_from_xml_url(self, video_id, xml_url)
|
||||||
|
|
||||||
|
def _real_extract(self, url):
|
||||||
|
return self._extract_video(self._match_id(url))
|
||||||
|
|
||||||
|
|
||||||
|
class ZDFChannelIE(ZDFIE):
|
||||||
|
_VALID_URL = r'^https?://www\.zdf\.de/ZDFmediathek(?P<hash>#)?/(.*kanaluebersicht/)(?P<id>[0-9]+)'
|
||||||
|
|
||||||
|
_TEST = {
|
||||||
|
'url': 'http://www.zdf.de/ZDFmediathek#/kanaluebersicht/1586442/sendung/Titanic',
|
||||||
|
'info_dict': {
|
||||||
|
'id': '1586442',
|
||||||
|
'title': 'Titanic',
|
||||||
|
'description': 'md5:444c048cfe3fdc2561be7de4bcbf1d04',
|
||||||
|
},
|
||||||
|
'playlist_count': 3,
|
||||||
|
}
|
||||||
|
|
||||||
|
def _extract_channel(self, channel_id):
|
||||||
|
def load_chunks(channel_id, chunk_length):
|
||||||
|
offset = 0
|
||||||
|
while True:
|
||||||
|
url = ('http://www.zdf.de/ZDFmediathek/xmlservice/web/aktuellste?ak=web&offset=%d&maxLength=%d&id=%s'
|
||||||
|
% (offset, chunk_length, channel_id))
|
||||||
|
result = extract_channel_from_xml_url(self, channel_id, url)
|
||||||
|
yield result
|
||||||
|
if len(result['assets']) < chunk_length:
|
||||||
|
return
|
||||||
|
offset += chunk_length
|
||||||
|
|
||||||
|
def load_channel(channel_id):
|
||||||
|
chunks = list(load_chunks(channel_id, 50)) # The server rejects higher values
|
||||||
|
assets = [asset for chunk in chunks for asset in chunk['assets']]
|
||||||
|
video_ids = [asset['id'] for asset in
|
||||||
|
filter(lambda asset: asset['type'] == 'video',
|
||||||
|
assets)]
|
||||||
|
topic_ids = [asset['id'] for asset in
|
||||||
|
filter(lambda asset: asset['type'] == 'thema',
|
||||||
|
assets)]
|
||||||
|
if topic_ids:
|
||||||
|
video_ids = reduce(list.__add__,
|
||||||
|
[load_channel(topic_id)['video_ids']
|
||||||
|
for topic_id in topic_ids],
|
||||||
|
video_ids)
|
||||||
|
|
||||||
|
result = chunks[0]
|
||||||
|
result['video_ids'] = video_ids
|
||||||
|
return result
|
||||||
|
|
||||||
|
channel = load_channel(channel_id)
|
||||||
|
return {
|
||||||
|
'_type': 'playlist',
|
||||||
|
'id': channel['id'],
|
||||||
|
'title': channel['title'],
|
||||||
|
'description': channel['description'],
|
||||||
|
'entries': [self._extract_video(video_id)
|
||||||
|
for video_id in channel['video_ids']],
|
||||||
|
}
|
||||||
|
|
||||||
|
def _real_extract(self, url):
|
||||||
|
return self._extract_channel(self._match_id(url))
|
||||||
|
|
Loading…
Reference in a new issue