mirror of
https://github.com/ytdl-org/youtube-dl.git
synced 2024-11-15 03:27:46 +01:00
280 lines
9.9 KiB
Python
280 lines
9.9 KiB
Python
# coding: utf-8
|
|
from __future__ import unicode_literals
|
|
|
|
import re
|
|
import itertools
|
|
|
|
from .common import InfoExtractor
|
|
from ..compat import (
|
|
compat_str,
|
|
compat_parse_qs,
|
|
compat_urllib_parse_urlparse,
|
|
)
|
|
from ..utils import (
|
|
determine_ext,
|
|
unified_strdate,
|
|
try_get,
|
|
int_or_none,
|
|
)
|
|
|
|
|
|
class RutubeIE(InfoExtractor):
|
|
IE_NAME = 'rutube'
|
|
IE_DESC = 'Rutube videos'
|
|
_VALID_URL = r'https?://rutube\.ru/(?:video|(?:play/)?embed)/(?P<id>[\da-z]{32})'
|
|
|
|
_TESTS = [{
|
|
'url': 'http://rutube.ru/video/3eac3b4561676c17df9132a9a1e62e3e/',
|
|
'info_dict': {
|
|
'id': '3eac3b4561676c17df9132a9a1e62e3e',
|
|
'ext': 'mp4',
|
|
'title': 'Раненный кенгуру забежал в аптеку',
|
|
'description': 'http://www.ntdtv.ru ',
|
|
'duration': 80,
|
|
'uploader': 'NTDRussian',
|
|
'uploader_id': '29790',
|
|
'upload_date': '20131016',
|
|
'age_limit': 0,
|
|
},
|
|
'params': {
|
|
# It requires ffmpeg (m3u8 download)
|
|
'skip_download': True,
|
|
},
|
|
}, {
|
|
'url': 'http://rutube.ru/play/embed/a10e53b86e8f349080f718582ce4c661',
|
|
'only_matching': True,
|
|
}, {
|
|
'url': 'http://rutube.ru/embed/a10e53b86e8f349080f718582ce4c661',
|
|
'only_matching': True,
|
|
}, {
|
|
'url': 'http://rutube.ru/video/3eac3b4561676c17df9132a9a1e62e3e/?pl_id=4252',
|
|
'only_matching': True,
|
|
}]
|
|
|
|
@classmethod
|
|
def suitable(cls, url):
|
|
parts = compat_urllib_parse_urlparse(url)
|
|
params = compat_parse_qs(parts.query)
|
|
|
|
# see if URL without parameters is OK
|
|
res = super(RutubeIE, cls).suitable(url)
|
|
|
|
if params: # we only allow pl_id parameter in the url
|
|
res = res and 'pl_id' in params and len(params) == 1
|
|
|
|
return res
|
|
|
|
@staticmethod
|
|
def _extract_urls(webpage):
|
|
return [mobj.group('url') for mobj in re.finditer(
|
|
r'<iframe[^>]+?src=(["\'])(?P<url>(?:https?:)?//rutube\.ru/embed/[\da-z]{32}.*?)\1',
|
|
webpage)]
|
|
|
|
def _real_extract(self, url):
|
|
video_id = self._match_id(url)
|
|
video = self._download_json(
|
|
'http://rutube.ru/api/video/%s/?format=json' % video_id,
|
|
video_id, 'Downloading video JSON')
|
|
|
|
# Some videos don't have the author field
|
|
author = video.get('author') or {}
|
|
|
|
options = self._download_json(
|
|
'http://rutube.ru/api/play/options/%s/?format=json' % video_id,
|
|
video_id, 'Downloading options JSON')
|
|
|
|
formats = []
|
|
for format_id, format_url in options['video_balancer'].items():
|
|
ext = determine_ext(format_url)
|
|
if ext == 'm3u8':
|
|
formats.extend(self._extract_m3u8_formats(
|
|
format_url, video_id, 'mp4', m3u8_id=format_id, fatal=False))
|
|
elif ext == 'f4m':
|
|
formats.extend(self._extract_f4m_formats(
|
|
format_url, video_id, f4m_id=format_id, fatal=False))
|
|
else:
|
|
formats.append({
|
|
'url': format_url,
|
|
'format_id': format_id,
|
|
})
|
|
self._sort_formats(formats)
|
|
|
|
return {
|
|
'id': video['id'],
|
|
'title': video['title'],
|
|
'description': video['description'],
|
|
'duration': video['duration'],
|
|
'view_count': video['hits'],
|
|
'formats': formats,
|
|
'thumbnail': video['thumbnail_url'],
|
|
'uploader': author.get('name'),
|
|
'uploader_id': compat_str(author['id']) if author else None,
|
|
'upload_date': unified_strdate(video['created_ts']),
|
|
'age_limit': 18 if video['is_adult'] else 0,
|
|
}
|
|
|
|
|
|
class RutubeEmbedIE(InfoExtractor):
|
|
IE_NAME = 'rutube:embed'
|
|
IE_DESC = 'Rutube embedded videos'
|
|
_VALID_URL = r'https?://rutube\.ru/(?:video|play)/embed/(?P<id>[0-9]+)'
|
|
|
|
_TESTS = [{
|
|
'url': 'http://rutube.ru/video/embed/6722881?vk_puid37=&vk_puid38=',
|
|
'info_dict': {
|
|
'id': 'a10e53b86e8f349080f718582ce4c661',
|
|
'ext': 'mp4',
|
|
'upload_date': '20131223',
|
|
'uploader_id': '297833',
|
|
'description': 'Видео группы ★http://vk.com/foxkidsreset★ музей Fox Kids и Jetix<br/><br/> восстановлено и сделано в шикоформате subziro89 http://vk.com/subziro89',
|
|
'uploader': 'subziro89 ILya',
|
|
'title': 'Мистический городок Эйри в Индиан 5 серия озвучка subziro89',
|
|
},
|
|
'params': {
|
|
'skip_download': 'Requires ffmpeg',
|
|
},
|
|
}, {
|
|
'url': 'http://rutube.ru/play/embed/8083783',
|
|
'only_matching': True,
|
|
}]
|
|
|
|
def _real_extract(self, url):
|
|
embed_id = self._match_id(url)
|
|
webpage = self._download_webpage(url, embed_id)
|
|
|
|
canonical_url = self._html_search_regex(
|
|
r'<link\s+rel="canonical"\s+href="([^"]+?)"', webpage,
|
|
'Canonical URL')
|
|
return self.url_result(canonical_url, 'Rutube')
|
|
|
|
|
|
class RutubeChannelIE(InfoExtractor):
|
|
IE_NAME = 'rutube:channel'
|
|
IE_DESC = 'Rutube channels'
|
|
_VALID_URL = r'https?://rutube\.ru/tags/video/(?P<id>\d+)'
|
|
_TESTS = [{
|
|
'url': 'http://rutube.ru/tags/video/1800/',
|
|
'info_dict': {
|
|
'id': '1800',
|
|
},
|
|
'playlist_mincount': 68,
|
|
}]
|
|
|
|
_PAGE_TEMPLATE = 'http://rutube.ru/api/tags/video/%s/?page=%s&format=json'
|
|
|
|
def _extract_videos(self, channel_id, channel_title=None):
|
|
entries = []
|
|
for pagenum in itertools.count(1):
|
|
page = self._download_json(
|
|
self._PAGE_TEMPLATE % (channel_id, pagenum),
|
|
channel_id, 'Downloading page %s' % pagenum)
|
|
results = page['results']
|
|
if not results:
|
|
break
|
|
entries.extend(self.url_result(result['video_url'], 'Rutube') for result in results)
|
|
if not page['has_next']:
|
|
break
|
|
return self.playlist_result(entries, channel_id, channel_title)
|
|
|
|
def _real_extract(self, url):
|
|
mobj = re.match(self._VALID_URL, url)
|
|
channel_id = mobj.group('id')
|
|
return self._extract_videos(channel_id)
|
|
|
|
|
|
class RutubeMovieIE(RutubeChannelIE):
|
|
IE_NAME = 'rutube:movie'
|
|
IE_DESC = 'Rutube movies'
|
|
_VALID_URL = r'https?://rutube\.ru/metainfo/tv/(?P<id>\d+)'
|
|
_TESTS = []
|
|
|
|
_MOVIE_TEMPLATE = 'http://rutube.ru/api/metainfo/tv/%s/?format=json'
|
|
_PAGE_TEMPLATE = 'http://rutube.ru/api/metainfo/tv/%s/video?page=%s&format=json'
|
|
|
|
def _real_extract(self, url):
|
|
movie_id = self._match_id(url)
|
|
movie = self._download_json(
|
|
self._MOVIE_TEMPLATE % movie_id, movie_id,
|
|
'Downloading movie JSON')
|
|
movie_name = movie['name']
|
|
return self._extract_videos(movie_id, movie_name)
|
|
|
|
|
|
class RutubePersonIE(RutubeChannelIE):
|
|
IE_NAME = 'rutube:person'
|
|
IE_DESC = 'Rutube person videos'
|
|
_VALID_URL = r'https?://rutube\.ru/video/person/(?P<id>\d+)'
|
|
_TESTS = [{
|
|
'url': 'http://rutube.ru/video/person/313878/',
|
|
'info_dict': {
|
|
'id': '313878',
|
|
},
|
|
'playlist_mincount': 37,
|
|
}]
|
|
|
|
_PAGE_TEMPLATE = 'http://rutube.ru/api/video/person/%s/?page=%s&format=json'
|
|
|
|
|
|
class RutubePlaylistIE(InfoExtractor):
|
|
IE_NAME = 'rutube:playlist'
|
|
IE_DESC = 'Rutube playlists'
|
|
_TESTS = [{
|
|
'url': 'https://rutube.ru/video/10b3a03fc01d5bbcc632a2f3514e8aab/?pl_id=4252&pl_type=source',
|
|
'info_dict': {
|
|
'id': '4252',
|
|
},
|
|
'playlist_count': 25,
|
|
}]
|
|
|
|
_VALID_URL = r'https?://rutube\.ru/(?:video|(?:play/)?embed)/[\da-z]{32}/\?(?:.+)?pl_id=(?P<id>\d+)'
|
|
_PAGE_TEMPLATE = 'http://rutube.ru/api/playlist/source/%s/?page=%s'
|
|
|
|
@staticmethod
|
|
def suitable(url):
|
|
params = compat_parse_qs(compat_urllib_parse_urlparse(url).query)
|
|
return params.get('pl_id') and int_or_none(params['pl_id'][0]) \
|
|
and params.get('pl_type')
|
|
|
|
def _real_extract(self, url):
|
|
playlist_id = self._match_id(url)
|
|
return self._extract_playlist(playlist_id)
|
|
|
|
def _extract_playlist(self, playlist_id):
|
|
entries = []
|
|
for pagenum in itertools.count(1):
|
|
page_url = self._PAGE_TEMPLATE % (playlist_id, pagenum)
|
|
|
|
# download_json will sent an accept: application/xml header
|
|
page = self._download_json(page_url, playlist_id,
|
|
"Downloading metadata for page %s" % pagenum,
|
|
headers={'Accept': 'application/json'})
|
|
|
|
if not page['results']:
|
|
break
|
|
|
|
results = page['results']
|
|
for result in results:
|
|
entry = self.url_result(result.get('video_url'), 'Rutube')
|
|
category = try_get(result, lambda x: x['category']['name'])
|
|
entry.update({
|
|
'id': result.get('id'),
|
|
'uploader': try_get(result, lambda x: x['author']['name']),
|
|
'uploader_id': try_get(result, lambda x: x['author']['id']),
|
|
'upload_date': unified_strdate(result.get('created_ts')),
|
|
'title': result.get('title'),
|
|
'description': result.get('description'),
|
|
'thumbnail': result.get('thumbnail_url'),
|
|
'duration': int_or_none(result.get('duration')),
|
|
'category': [category] if category else None,
|
|
'age_limit': 18 if result.get('is_adult') else 0,
|
|
'view_count': int_or_none(result.get('hits')),
|
|
'is_live': result.get('is_livestream'),
|
|
'webpage_url': result.get('video_url'),
|
|
})
|
|
entries.append(entry)
|
|
|
|
if page['has_next'] is False:
|
|
break
|
|
|
|
return self.playlist_result(entries, playlist_id, page['name'])
|