1
0
Fork 0
mirror of https://github.com/ytdl-org/youtube-dl.git synced 2024-11-26 04:02:11 +00:00

[dailymotion] improve extraction

- extract http formats included in m3u8 manifest
- fix user extraction(closes #3553)(closes #21415)
- add suport for User Authentication(closes #11491)
- fix password protected videos extraction(closes #23176)
- respect age limit option and family filter cookie value(closes #18437)
- handle video url playlist query param
- report alowed countries for geo-restricted videos
This commit is contained in:
Remita Amine 2019-11-26 22:01:34 +01:00
parent df65a4a1ed
commit 5ef62fc4ce
3 changed files with 234 additions and 341 deletions

View file

@ -1766,6 +1766,19 @@ class InfoExtractor(object):
# the same GROUP-ID # the same GROUP-ID
f['acodec'] = 'none' f['acodec'] = 'none'
formats.append(f) formats.append(f)
# for DailyMotion
progressive_uri = last_stream_inf.get('PROGRESSIVE-URI')
if progressive_uri:
http_f = f.copy()
del http_f['manifest_url']
http_f.update({
'format_id': f['format_id'].replace('hls-', 'http-'),
'protocol': 'http',
'url': progressive_uri,
})
formats.append(http_f)
last_stream_inf = {} last_stream_inf = {}
return formats return formats

View file

@ -1,50 +1,93 @@
# coding: utf-8 # coding: utf-8
from __future__ import unicode_literals from __future__ import unicode_literals
import base64
import functools import functools
import hashlib
import itertools
import json import json
import random
import re import re
import string
from .common import InfoExtractor from .common import InfoExtractor
from ..compat import compat_struct_pack from ..compat import compat_HTTPError
from ..utils import ( from ..utils import (
determine_ext, age_restricted,
error_to_compat_str, clean_html,
ExtractorError, ExtractorError,
int_or_none, int_or_none,
mimetype2ext,
OnDemandPagedList, OnDemandPagedList,
parse_iso8601,
sanitized_Request,
str_to_int,
try_get, try_get,
unescapeHTML, unescapeHTML,
update_url_query,
url_or_none,
urlencode_postdata, urlencode_postdata,
) )
class DailymotionBaseInfoExtractor(InfoExtractor): class DailymotionBaseInfoExtractor(InfoExtractor):
_FAMILY_FILTER = None
_HEADERS = {
'Content-Type': 'application/json',
'Origin': 'https://www.dailymotion.com',
}
_NETRC_MACHINE = 'dailymotion'
def _get_dailymotion_cookies(self):
return self._get_cookies('https://www.dailymotion.com/')
@staticmethod @staticmethod
def _build_request(url): def _get_cookie_value(cookies, name):
"""Build a request with the family filter disabled""" cookie = cookies.get('name')
request = sanitized_Request(url) if cookie:
request.add_header('Cookie', 'family_filter=off; ff=off') return cookie.value
return request
def _download_webpage_handle_no_ff(self, url, *args, **kwargs): def _set_dailymotion_cookie(self, name, value):
request = self._build_request(url) self._set_cookie('www.dailymotion.com', name, value)
return self._download_webpage_handle(request, *args, **kwargs)
def _download_webpage_no_ff(self, url, *args, **kwargs): def _real_initialize(self):
request = self._build_request(url) cookies = self._get_dailymotion_cookies()
return self._download_webpage(request, *args, **kwargs) ff = self._get_cookie_value(cookies, 'ff')
self._FAMILY_FILTER = ff == 'on' if ff else age_restricted(18, self._downloader.params.get('age_limit'))
self._set_dailymotion_cookie('ff', 'on' if self._FAMILY_FILTER else 'off')
def _call_api(self, object_type, xid, object_fields, note, filter_extra=None):
if not self._HEADERS.get('Authorization'):
cookies = self._get_dailymotion_cookies()
token = self._get_cookie_value(cookies, 'access_token') or self._get_cookie_value(cookies, 'client_token')
if not token:
data = {
'client_id': 'f1a362d288c1b98099c7',
'client_secret': 'eea605b96e01c796ff369935357eca920c5da4c5',
}
username, password = self._get_login_info()
if username:
data.update({
'grant_type': 'password',
'password': password,
'username': username,
})
else:
data['grant_type'] = 'client_credentials'
try:
token = self._download_json(
'https://graphql.api.dailymotion.com/oauth/token',
None, 'Downloading Access Token',
data=urlencode_postdata(data))['access_token']
except ExtractorError as e:
if isinstance(e.cause, compat_HTTPError) and e.cause.code == 400:
raise ExtractorError(self._parse_json(
e.cause.read().decode(), xid)['error_description'], expected=True)
raise
self._set_dailymotion_cookie('access_token' if username else 'client_token', token)
self._HEADERS['Authorization'] = 'Bearer ' + token
resp = self._download_json(
'https://graphql.api.dailymotion.com/', xid, note, data=json.dumps({
'query': '''{
%s(xid: "%s"%s) {
%s
}
}''' % (object_type, xid, ', ' + filter_extra if filter_extra else '', object_fields),
}).encode(), headers=self._HEADERS)
obj = resp['data'][object_type]
if not obj:
raise ExtractorError(resp['errors'][0]['message'], expected=True)
return obj
class DailymotionIE(DailymotionBaseInfoExtractor): class DailymotionIE(DailymotionBaseInfoExtractor):
@ -54,18 +97,9 @@ class DailymotionIE(DailymotionBaseInfoExtractor):
(?:(?:www|touch)\.)?dailymotion\.[a-z]{2,3}/(?:(?:(?:embed|swf|\#)/)?video|swf)| (?:(?:www|touch)\.)?dailymotion\.[a-z]{2,3}/(?:(?:(?:embed|swf|\#)/)?video|swf)|
(?:www\.)?lequipe\.fr/video (?:www\.)?lequipe\.fr/video
) )
/(?P<id>[^/?_]+) /(?P<id>[^/?_]+)(?:.+?\bplaylist=(?P<playlist_id>x[0-9a-z]+))?
''' '''
IE_NAME = 'dailymotion' IE_NAME = 'dailymotion'
_FORMATS = [
('stream_h264_ld_url', 'ld'),
('stream_h264_url', 'standard'),
('stream_h264_hq_url', 'hq'),
('stream_h264_hd_url', 'hd'),
('stream_h264_hd1080_url', 'hd180'),
]
_TESTS = [{ _TESTS = [{
'url': 'http://www.dailymotion.com/video/x5kesuj_office-christmas-party-review-jason-bateman-olivia-munn-t-j-miller_news', 'url': 'http://www.dailymotion.com/video/x5kesuj_office-christmas-party-review-jason-bateman-olivia-munn-t-j-miller_news',
'md5': '074b95bdee76b9e3654137aee9c79dfe', 'md5': '074b95bdee76b9e3654137aee9c79dfe',
@ -74,7 +108,6 @@ class DailymotionIE(DailymotionBaseInfoExtractor):
'ext': 'mp4', 'ext': 'mp4',
'title': 'Office Christmas Party Review Jason Bateman, Olivia Munn, T.J. Miller', 'title': 'Office Christmas Party Review Jason Bateman, Olivia Munn, T.J. Miller',
'description': 'Office Christmas Party Review - Jason Bateman, Olivia Munn, T.J. Miller', 'description': 'Office Christmas Party Review - Jason Bateman, Olivia Munn, T.J. Miller',
'thumbnail': r're:^https?:.*\.(?:jpg|png)$',
'duration': 187, 'duration': 187,
'timestamp': 1493651285, 'timestamp': 1493651285,
'upload_date': '20170501', 'upload_date': '20170501',
@ -146,7 +179,16 @@ class DailymotionIE(DailymotionBaseInfoExtractor):
}, { }, {
'url': 'https://www.lequipe.fr/video/k7MtHciueyTcrFtFKA2', 'url': 'https://www.lequipe.fr/video/k7MtHciueyTcrFtFKA2',
'only_matching': True, 'only_matching': True,
}, {
'url': 'https://www.dailymotion.com/video/x3z49k?playlist=xv4bw',
'only_matching': True,
}] }]
_GEO_BYPASS = False
_COMMON_MEDIA_FIELDS = '''description
geoblockedCountries {
allowed
}
xid'''
@staticmethod @staticmethod
def _extract_urls(webpage): def _extract_urls(webpage):
@ -162,264 +204,140 @@ class DailymotionIE(DailymotionBaseInfoExtractor):
return urls return urls
def _real_extract(self, url): def _real_extract(self, url):
video_id = self._match_id(url) video_id, playlist_id = re.match(self._VALID_URL, url).groups()
webpage = self._download_webpage_no_ff( if playlist_id:
'https://www.dailymotion.com/video/%s' % video_id, video_id) if not self._downloader.params.get('noplaylist'):
self.to_screen('Downloading playlist %s - add --no-playlist to just download video' % playlist_id)
return self.url_result(
'http://www.dailymotion.com/playlist/' + playlist_id,
'DailymotionPlaylist', playlist_id)
self.to_screen('Downloading just video %s because of --no-playlist' % video_id)
age_limit = self._rta_search(webpage)
description = self._og_search_description(
webpage, default=None) or self._html_search_meta(
'description', webpage, 'description')
view_count_str = self._search_regex(
(r'<meta[^>]+itemprop="interactionCount"[^>]+content="UserPlays:([\s\d,.]+)"',
r'video_views_count[^>]+>\s+([\s\d\,.]+)'),
webpage, 'view count', default=None)
if view_count_str:
view_count_str = re.sub(r'\s', '', view_count_str)
view_count = str_to_int(view_count_str)
comment_count = int_or_none(self._search_regex(
r'<meta[^>]+itemprop="interactionCount"[^>]+content="UserComments:(\d+)"',
webpage, 'comment count', default=None))
player_v5 = self._search_regex(
[r'buildPlayer\(({.+?})\);\n', # See https://github.com/ytdl-org/youtube-dl/issues/7826
r'playerV5\s*=\s*dmp\.create\([^,]+?,\s*({.+?})\);',
r'buildPlayer\(({.+?})\);',
r'var\s+config\s*=\s*({.+?});',
# New layout regex (see https://github.com/ytdl-org/youtube-dl/issues/13580)
r'__PLAYER_CONFIG__\s*=\s*({.+?});'],
webpage, 'player v5', default=None)
if player_v5:
player = self._parse_json(player_v5, video_id, fatal=False) or {}
metadata = try_get(player, lambda x: x['metadata'], dict)
if not metadata:
metadata_url = url_or_none(try_get(
player, lambda x: x['context']['metadata_template_url1']))
if metadata_url:
metadata_url = metadata_url.replace(':videoId', video_id)
else:
metadata_url = update_url_query(
'https://www.dailymotion.com/player/metadata/video/%s'
% video_id, {
'embedder': url,
'integration': 'inline',
'GK_PV5_NEON': '1',
})
metadata = self._download_json(
metadata_url, video_id, 'Downloading metadata JSON')
if try_get(metadata, lambda x: x['error']['type']) == 'password_protected':
password = self._downloader.params.get('videopassword') password = self._downloader.params.get('videopassword')
if password: media = self._call_api(
r = int(metadata['id'][1:], 36) 'media', video_id, '''... on Video {
us64e = lambda x: base64.urlsafe_b64encode(x).decode().strip('=') %s
t = ''.join(random.choice(string.ascii_letters) for i in range(10)) stats {
n = us64e(compat_struct_pack('I', r)) likes {
i = us64e(hashlib.md5(('%s%d%s' % (password, r, t)).encode()).digest()) total
}
views {
total
}
}
}
... on Live {
%s
audienceCount
isOnAir
}''' % (self._COMMON_MEDIA_FIELDS, self._COMMON_MEDIA_FIELDS), 'Downloading media JSON metadata',
'password: "%s"' % self._downloader.params.get('videopassword') if password else None)
xid = media['xid']
metadata = self._download_json( metadata = self._download_json(
'http://www.dailymotion.com/player/metadata/video/p' + i + t + n, video_id) 'https://www.dailymotion.com/player/metadata/video/' + xid,
xid, 'Downloading metadata JSON',
query={'app': 'com.dailymotion.neon'})
self._check_error(metadata) error = metadata.get('error')
formats = []
for quality, media_list in metadata['qualities'].items():
for media in media_list:
media_url = media.get('url')
if not media_url:
continue
type_ = media.get('type')
if type_ == 'application/vnd.lumberjack.manifest':
continue
ext = mimetype2ext(type_) or determine_ext(media_url)
if ext == 'm3u8':
m3u8_formats = self._extract_m3u8_formats(
media_url, video_id, 'mp4', preference=-1,
m3u8_id='hls', fatal=False)
for f in m3u8_formats:
f['url'] = f['url'].split('#')[0]
formats.append(f)
elif ext == 'f4m':
formats.extend(self._extract_f4m_formats(
media_url, video_id, preference=-1, f4m_id='hds', fatal=False))
else:
f = {
'url': media_url,
'format_id': 'http-%s' % quality,
'ext': ext,
}
m = re.search(r'H264-(?P<width>\d+)x(?P<height>\d+)', media_url)
if m:
f.update({
'width': int(m.group('width')),
'height': int(m.group('height')),
})
formats.append(f)
self._sort_formats(formats)
title = metadata['title']
duration = int_or_none(metadata.get('duration'))
timestamp = int_or_none(metadata.get('created_time'))
thumbnail = metadata.get('poster_url')
uploader = metadata.get('owner', {}).get('screenname')
uploader_id = metadata.get('owner', {}).get('id')
subtitles = {}
subtitles_data = metadata.get('subtitles', {}).get('data', {})
if subtitles_data and isinstance(subtitles_data, dict):
for subtitle_lang, subtitle in subtitles_data.items():
subtitles[subtitle_lang] = [{
'ext': determine_ext(subtitle_url),
'url': subtitle_url,
} for subtitle_url in subtitle.get('urls', [])]
return {
'id': video_id,
'title': title,
'description': description,
'thumbnail': thumbnail,
'duration': duration,
'timestamp': timestamp,
'uploader': uploader,
'uploader_id': uploader_id,
'age_limit': age_limit,
'view_count': view_count,
'comment_count': comment_count,
'formats': formats,
'subtitles': subtitles,
}
# vevo embed
vevo_id = self._search_regex(
r'<link rel="video_src" href="[^"]*?vevo\.com[^"]*?video=(?P<id>[\w]*)',
webpage, 'vevo embed', default=None)
if vevo_id:
return self.url_result('vevo:%s' % vevo_id, 'Vevo')
# fallback old player
embed_page = self._download_webpage_no_ff(
'https://www.dailymotion.com/embed/video/%s' % video_id,
video_id, 'Downloading embed page')
timestamp = parse_iso8601(self._html_search_meta(
'video:release_date', webpage, 'upload date'))
info = self._parse_json(
self._search_regex(
r'var info = ({.*?}),$', embed_page,
'video info', flags=re.MULTILINE),
video_id)
self._check_error(info)
formats = []
for (key, format_id) in self._FORMATS:
video_url = info.get(key)
if video_url is not None:
m_size = re.search(r'H264-(\d+)x(\d+)', video_url)
if m_size is not None:
width, height = map(int_or_none, (m_size.group(1), m_size.group(2)))
else:
width, height = None, None
formats.append({
'url': video_url,
'ext': 'mp4',
'format_id': format_id,
'width': width,
'height': height,
})
self._sort_formats(formats)
# subtitles
video_subtitles = self.extract_subtitles(video_id, webpage)
title = self._og_search_title(webpage, default=None)
if title is None:
title = self._html_search_regex(
r'(?s)<span\s+id="video_title"[^>]*>(.*?)</span>', webpage,
'title')
return {
'id': video_id,
'formats': formats,
'uploader': info['owner.screenname'],
'timestamp': timestamp,
'title': title,
'description': description,
'subtitles': video_subtitles,
'thumbnail': info['thumbnail_url'],
'age_limit': age_limit,
'view_count': view_count,
'duration': info['duration']
}
def _check_error(self, info):
error = info.get('error')
if error: if error:
title = error.get('title') or error['message'] title = error.get('title') or error['raw_message']
# See https://developer.dailymotion.com/api#access-error # See https://developer.dailymotion.com/api#access-error
if error.get('code') == 'DM007': if error.get('code') == 'DM007':
self.raise_geo_restricted(msg=title) allowed_countries = try_get(media, lambda x: x['geoblockedCountries']['allowed'], list)
self.raise_geo_restricted(msg=title, countries=allowed_countries)
raise ExtractorError( raise ExtractorError(
'%s said: %s' % (self.IE_NAME, title), expected=True) '%s said: %s' % (self.IE_NAME, title), expected=True)
def _get_subtitles(self, video_id, webpage): title = metadata['title']
try: is_live = media.get('isOnAir')
sub_list = self._download_webpage( formats = []
'https://api.dailymotion.com/video/%s/subtitles?fields=id,language,url' % video_id, for quality, media_list in metadata['qualities'].items():
video_id, note=False) for m in media_list:
except ExtractorError as err: media_url = m.get('url')
self._downloader.report_warning('unable to download video subtitles: %s' % error_to_compat_str(err)) media_type = m.get('type')
return {} if not media_url or media_type == 'application/vnd.lumberjack.manifest':
info = json.loads(sub_list) continue
if (info['total'] > 0): if media_type == 'application/x-mpegURL':
sub_lang_list = dict((l['language'], [{'url': l['url'], 'ext': 'srt'}]) for l in info['list']) formats.extend(self._extract_m3u8_formats(
return sub_lang_list media_url, video_id, 'mp4',
self._downloader.report_warning('video doesn\'t have subtitles') 'm3u8' if is_live else 'm3u8_native',
return {} m3u8_id='hls', fatal=False))
else:
f = {
'url': media_url,
'format_id': 'http-' + quality,
}
m = re.search(r'/H264-(\d+)x(\d+)(?:-(60)/)?', media_url)
if m:
width, height, fps = map(int_or_none, m.groups())
f.update({
'fps': fps,
'height': height,
'width': width,
})
formats.append(f)
for f in formats:
f['url'] = f['url'].split('#')[0]
if not f.get('fps') and f['format_id'].endswith('@60'):
f['fps'] = 60
self._sort_formats(formats)
subtitles = {}
subtitles_data = try_get(metadata, lambda x: x['subtitles']['data'], dict) or {}
for subtitle_lang, subtitle in subtitles_data.items():
subtitles[subtitle_lang] = [{
'url': subtitle_url,
} for subtitle_url in subtitle.get('urls', [])]
thumbnails = []
for height, poster_url in metadata.get('posters', {}).items():
thumbnails.append({
'height': int_or_none(height),
'id': height,
'url': poster_url,
})
owner = metadata.get('owner') or {}
stats = media.get('stats') or {}
get_count = lambda x: int_or_none(try_get(stats, lambda y: y[x + 's']['total']))
return {
'id': video_id,
'title': self._live_title(title) if is_live else title,
'description': clean_html(media.get('description')),
'thumbnails': thumbnails,
'duration': int_or_none(metadata.get('duration')) or None,
'timestamp': int_or_none(metadata.get('created_time')),
'uploader': owner.get('screenname'),
'uploader_id': owner.get('id') or metadata.get('screenname'),
'age_limit': 18 if metadata.get('explicit') else 0,
'tags': metadata.get('tags'),
'view_count': get_count('view') or int_or_none(media.get('audienceCount')),
'like_count': get_count('like'),
'formats': formats,
'subtitles': subtitles,
'is_live': is_live,
}
class DailymotionPlaylistIE(DailymotionBaseInfoExtractor): class DailymotionPlaylistBaseIE(DailymotionBaseInfoExtractor):
IE_NAME = 'dailymotion:playlist'
_VALID_URL = r'(?:https?://)?(?:www\.)?dailymotion\.[a-z]{2,3}/playlist/(?P<id>x[0-9a-z]+)'
_TESTS = [{
'url': 'http://www.dailymotion.com/playlist/xv4bw_nqtv_sport/1#video=xl8v3q',
'info_dict': {
'title': 'SPORT',
'id': 'xv4bw',
},
'playlist_mincount': 20,
}]
_PAGE_SIZE = 100 _PAGE_SIZE = 100
def _fetch_page(self, playlist_id, authorizaion, page): def _fetch_page(self, playlist_id, page):
page += 1 page += 1
videos = self._download_json( videos = self._call_api(
'https://graphql.api.dailymotion.com', self._OBJECT_TYPE, playlist_id,
playlist_id, 'Downloading page %d' % page, '''videos(allowExplicit: %s, first: %d, page: %d) {
data=json.dumps({
'query': '''{
collection(xid: "%s") {
videos(first: %d, page: %d) {
pageInfo {
hasNextPage
nextPage
}
edges { edges {
node { node {
xid xid
url url
} }
} }
} }''' % ('false' if self._FAMILY_FILTER else 'true', self._PAGE_SIZE, page),
} 'Downloading page %d' % page)['videos']
}''' % (playlist_id, self._PAGE_SIZE, page)
}).encode(), headers={
'Authorization': authorizaion,
'Origin': 'https://www.dailymotion.com',
})['data']['collection']['videos']
for edge in videos['edges']: for edge in videos['edges']:
node = edge['node'] node = edge['node']
yield self.url_result( yield self.url_result(
@ -427,86 +345,49 @@ class DailymotionPlaylistIE(DailymotionBaseInfoExtractor):
def _real_extract(self, url): def _real_extract(self, url):
playlist_id = self._match_id(url) playlist_id = self._match_id(url)
webpage = self._download_webpage(url, playlist_id)
api = self._parse_json(self._search_regex(
r'__PLAYER_CONFIG__\s*=\s*({.+?});',
webpage, 'player config'), playlist_id)['context']['api']
auth = self._download_json(
api.get('auth_url', 'https://graphql.api.dailymotion.com/oauth/token'),
playlist_id, data=urlencode_postdata({
'client_id': api.get('client_id', 'f1a362d288c1b98099c7'),
'client_secret': api.get('client_secret', 'eea605b96e01c796ff369935357eca920c5da4c5'),
'grant_type': 'client_credentials',
}))
authorizaion = '%s %s' % (auth.get('token_type', 'Bearer'), auth['access_token'])
entries = OnDemandPagedList(functools.partial( entries = OnDemandPagedList(functools.partial(
self._fetch_page, playlist_id, authorizaion), self._PAGE_SIZE) self._fetch_page, playlist_id), self._PAGE_SIZE)
return self.playlist_result( return self.playlist_result(
entries, playlist_id, entries, playlist_id)
self._og_search_title(webpage))
class DailymotionUserIE(DailymotionBaseInfoExtractor): class DailymotionPlaylistIE(DailymotionPlaylistBaseIE):
IE_NAME = 'dailymotion:playlist'
_VALID_URL = r'(?:https?://)?(?:www\.)?dailymotion\.[a-z]{2,3}/playlist/(?P<id>x[0-9a-z]+)'
_TESTS = [{
'url': 'http://www.dailymotion.com/playlist/xv4bw_nqtv_sport/1#video=xl8v3q',
'info_dict': {
'id': 'xv4bw',
},
'playlist_mincount': 20,
}]
_OBJECT_TYPE = 'collection'
class DailymotionUserIE(DailymotionPlaylistBaseIE):
IE_NAME = 'dailymotion:user' IE_NAME = 'dailymotion:user'
_VALID_URL = r'https?://(?:www\.)?dailymotion\.[a-z]{2,3}/(?!(?:embed|swf|#|video|playlist)/)(?:(?:old/)?user/)?(?P<user>[^/]+)' _VALID_URL = r'https?://(?:www\.)?dailymotion\.[a-z]{2,3}/(?!(?:embed|swf|#|video|playlist)/)(?:(?:old/)?user/)?(?P<id>[^/]+)'
_MORE_PAGES_INDICATOR = r'(?s)<div class="pages[^"]*">.*?<a\s+class="[^"]*?icon-arrow_right[^"]*?"'
_PAGE_TEMPLATE = 'http://www.dailymotion.com/user/%s/%s'
_TESTS = [{ _TESTS = [{
'url': 'https://www.dailymotion.com/user/nqtv', 'url': 'https://www.dailymotion.com/user/nqtv',
'info_dict': { 'info_dict': {
'id': 'nqtv', 'id': 'nqtv',
'title': 'Rémi Gaillard',
}, },
'playlist_mincount': 100, 'playlist_mincount': 152,
}, { }, {
'url': 'http://www.dailymotion.com/user/UnderProject', 'url': 'http://www.dailymotion.com/user/UnderProject',
'info_dict': { 'info_dict': {
'id': 'UnderProject', 'id': 'UnderProject',
'title': 'UnderProject',
}, },
'playlist_mincount': 1800, 'playlist_mincount': 1000,
'expected_warnings': [
'Stopped at duplicated page',
],
'skip': 'Takes too long time', 'skip': 'Takes too long time',
}, {
'url': 'https://www.dailymotion.com/user/nqtv',
'info_dict': {
'id': 'nqtv',
},
'playlist_mincount': 148,
'params': {
'age_limit': 0,
},
}] }]
_OBJECT_TYPE = 'channel'
def _extract_entries(self, id):
video_ids = set()
processed_urls = set()
for pagenum in itertools.count(1):
page_url = self._PAGE_TEMPLATE % (id, pagenum)
webpage, urlh = self._download_webpage_handle_no_ff(
page_url, id, 'Downloading page %s' % pagenum)
if urlh.geturl() in processed_urls:
self.report_warning('Stopped at duplicated page %s, which is the same as %s' % (
page_url, urlh.geturl()), id)
break
processed_urls.add(urlh.geturl())
for video_id in re.findall(r'data-xid="(.+?)"', webpage):
if video_id not in video_ids:
yield self.url_result(
'http://www.dailymotion.com/video/%s' % video_id,
DailymotionIE.ie_key(), video_id)
video_ids.add(video_id)
if re.search(self._MORE_PAGES_INDICATOR, webpage) is None:
break
def _real_extract(self, url):
mobj = re.match(self._VALID_URL, url)
user = mobj.group('user')
webpage = self._download_webpage(
'https://www.dailymotion.com/user/%s' % user, user)
full_user = unescapeHTML(self._html_search_regex(
r'<a class="nav-image" title="([^"]+)" href="/%s">' % re.escape(user),
webpage, 'user'))
return {
'_type': 'playlist',
'id': user,
'title': full_user,
'entries': self._extract_entries(user),
}

View file

@ -216,8 +216,7 @@ class VKIE(VKBaseIE):
'id': 'k3lz2cmXyRuJQSjGHUv', 'id': 'k3lz2cmXyRuJQSjGHUv',
'ext': 'mp4', 'ext': 'mp4',
'title': 'md5:d52606645c20b0ddbb21655adaa4f56f', 'title': 'md5:d52606645c20b0ddbb21655adaa4f56f',
# TODO: fix test by fixing dailymotion description extraction 'description': 'md5:424b8e88cc873217f520e582ba28bb36',
'description': 'md5:c651358f03c56f1150b555c26d90a0fd',
'uploader': 'AniLibria.Tv', 'uploader': 'AniLibria.Tv',
'upload_date': '20160914', 'upload_date': '20160914',
'uploader_id': 'x1p5vl5', 'uploader_id': 'x1p5vl5',