1
0
Fork 0
mirror of https://github.com/ytdl-org/youtube-dl.git synced 2024-12-27 03:07:42 +00:00
youtube-dl/youtube_dl/extractor/facebook.py

707 lines
29 KiB
Python
Raw Normal View History

2017-02-11 09:18:45 +00:00
# coding: utf-8
2014-03-04 02:36:54 +00:00
from __future__ import unicode_literals
import json
2013-06-23 18:59:45 +00:00
import re
import socket
from .common import InfoExtractor
from ..compat import (
compat_etree_fromstring,
2013-06-23 18:59:45 +00:00
compat_http_client,
compat_str,
2013-06-23 18:59:45 +00:00
compat_urllib_error,
compat_urllib_parse_unquote,
compat_urllib_parse_unquote_plus,
)
from ..utils import (
clean_html,
error_to_compat_str,
2013-06-23 18:59:45 +00:00
ExtractorError,
float_or_none,
get_element_by_id,
int_or_none,
js_to_json,
limit_length,
parse_count,
qualities,
Switch codebase to use sanitized_Request instead of compat_urllib_request.Request [downloader/dash] Use sanitized_Request [downloader/http] Use sanitized_Request [atresplayer] Use sanitized_Request [bambuser] Use sanitized_Request [bliptv] Use sanitized_Request [brightcove] Use sanitized_Request [cbs] Use sanitized_Request [ceskatelevize] Use sanitized_Request [collegerama] Use sanitized_Request [extractor/common] Use sanitized_Request [crunchyroll] Use sanitized_Request [dailymotion] Use sanitized_Request [dcn] Use sanitized_Request [dramafever] Use sanitized_Request [dumpert] Use sanitized_Request [eitb] Use sanitized_Request [escapist] Use sanitized_Request [everyonesmixtape] Use sanitized_Request [extremetube] Use sanitized_Request [facebook] Use sanitized_Request [fc2] Use sanitized_Request [flickr] Use sanitized_Request [4tube] Use sanitized_Request [gdcvault] Use sanitized_Request [extractor/generic] Use sanitized_Request [hearthisat] Use sanitized_Request [hotnewhiphop] Use sanitized_Request [hypem] Use sanitized_Request [iprima] Use sanitized_Request [ivi] Use sanitized_Request [keezmovies] Use sanitized_Request [letv] Use sanitized_Request [lynda] Use sanitized_Request [metacafe] Use sanitized_Request [minhateca] Use sanitized_Request [miomio] Use sanitized_Request [meovideo] Use sanitized_Request [mofosex] Use sanitized_Request [moniker] Use sanitized_Request [mooshare] Use sanitized_Request [movieclips] Use sanitized_Request [mtv] Use sanitized_Request [myvideo] Use sanitized_Request [neteasemusic] Use sanitized_Request [nfb] Use sanitized_Request [niconico] Use sanitized_Request [noco] Use sanitized_Request [nosvideo] Use sanitized_Request [novamov] Use sanitized_Request [nowness] Use sanitized_Request [nuvid] Use sanitized_Request [played] Use sanitized_Request [pluralsight] Use sanitized_Request [pornhub] Use sanitized_Request [pornotube] Use sanitized_Request [primesharetv] Use sanitized_Request [promptfile] Use sanitized_Request [qqmusic] Use sanitized_Request [rtve] Use sanitized_Request [safari] Use sanitized_Request [sandia] Use sanitized_Request [shared] Use sanitized_Request [sharesix] Use sanitized_Request [sina] Use sanitized_Request [smotri] Use sanitized_Request [sohu] Use sanitized_Request [spankwire] Use sanitized_Request [sportdeutschland] Use sanitized_Request [streamcloud] Use sanitized_Request [streamcz] Use sanitized_Request [tapely] Use sanitized_Request [tube8] Use sanitized_Request [tubitv] Use sanitized_Request [twitch] Use sanitized_Request [twitter] Use sanitized_Request [udemy] Use sanitized_Request [vbox7] Use sanitized_Request [veoh] Use sanitized_Request [vessel] Use sanitized_Request [vevo] Use sanitized_Request [viddler] Use sanitized_Request [videomega] Use sanitized_Request [viewvster] Use sanitized_Request [viki] Use sanitized_Request [vk] Use sanitized_Request [vodlocker] Use sanitized_Request [voicerepublic] Use sanitized_Request [wistia] Use sanitized_Request [xfileshare] Use sanitized_Request [xtube] Use sanitized_Request [xvideos] Use sanitized_Request [yandexmusic] Use sanitized_Request [youku] Use sanitized_Request [youporn] Use sanitized_Request [youtube] Use sanitized_Request [patreon] Use sanitized_Request [extractor/common] Remove unused import [nfb] PEP 8
2015-11-21 16:18:17 +00:00
sanitized_Request,
try_get,
urlencode_postdata,
urljoin,
2013-06-23 18:59:45 +00:00
)
class FacebookIE(InfoExtractor):
_VALID_URL = r'''(?x)
(?:
https?://
(?:[\w-]+\.)?(?:facebook\.com|facebookcorewwwi\.onion)/
(?:[^#]*?\#!/)?
(?:
(?:
video/video\.php|
photo\.php|
video\.php|
video/embed|
story\.php|
watch(?:/live)?/?
)\?(?:.*?)(?:v|video_id|story_fbid)=|
[^/]+/videos/(?:[^/]+/)?|
[^/]+/posts/|
groups/[^/]+/permalink/|
watchparty/
)|
facebook:
)
(?P<id>[0-9]+)
'''
_LOGIN_URL = 'https://www.facebook.com/login.php?next=http%3A%2F%2Ffacebook.com%2Fhome.php&login_attempt=1'
_CHECKPOINT_URL = 'https://www.facebook.com/checkpoint/?next=http%3A%2F%2Ffacebook.com%2Fhome.php&_fb_noscript=1'
2013-06-23 18:59:45 +00:00
_NETRC_MACHINE = 'facebook'
2014-03-04 02:36:54 +00:00
IE_NAME = 'facebook'
_VIDEO_PAGE_TEMPLATE = 'https://www.facebook.com/video/video.php?v=%s'
_VIDEO_PAGE_TAHOE_TEMPLATE = 'https://www.facebook.com/video/tahoe/async/%s/?chain=true&isvideo=true&payloadtype=primary'
2014-08-27 09:08:47 +00:00
_TESTS = [{
'url': 'https://www.facebook.com/video.php?v=637842556329505&fref=nf',
'md5': '6a40d33c0eccbb1af76cf0485a052659',
2014-03-04 02:36:54 +00:00
'info_dict': {
'id': '637842556329505',
2014-03-04 02:36:54 +00:00
'ext': 'mp4',
2014-09-29 03:19:56 +00:00
'title': 're:Did you know Kei Nishikori is the first Asian man to ever reach a Grand Slam',
'uploader': 'Tennis on Facebook',
'upload_date': '20140908',
'timestamp': 1410199200,
},
'skip': 'Requires logging in',
}, {
# data.video
'url': 'https://www.facebook.com/video.php?v=274175099429670',
'info_dict': {
'id': '274175099429670',
'ext': 'mp4',
'title': 're:^Asif Nawab Butt posted a video',
'uploader': 'Asif Nawab Butt',
'upload_date': '20140506',
'timestamp': 1399398998,
'thumbnail': r're:^https?://.*',
},
'expected_warnings': [
'title'
]
}, {
'note': 'Video with DASH manifest',
'url': 'https://www.facebook.com/video.php?v=957955867617029',
'md5': 'b2c28d528273b323abe5c6ab59f0f030',
'info_dict': {
'id': '957955867617029',
'ext': 'mp4',
'title': 'When you post epic content on instagram.com/433 8 million followers, this is ...',
'uploader': 'Demy de Zeeuw',
'upload_date': '20160110',
'timestamp': 1452431627,
},
'skip': 'Requires logging in',
}, {
'url': 'https://www.facebook.com/maxlayn/posts/10153807558977570',
'md5': '037b1fa7f3c2d02b7a0d7bc16031ecc6',
'info_dict': {
'id': '544765982287235',
'ext': 'mp4',
'title': '"What are you doing running in the snow?"',
'uploader': 'FailArmy',
},
'skip': 'Video gone',
2016-03-10 06:58:05 +00:00
}, {
'url': 'https://m.facebook.com/story.php?story_fbid=1035862816472149&id=116132035111903',
'md5': '1deb90b6ac27f7efcf6d747c8a27f5e3',
'info_dict': {
'id': '1035862816472149',
'ext': 'mp4',
'title': 'What the Flock Is Going On In New Zealand Credit: ViralHog',
'uploader': 'S. Saint',
},
'skip': 'Video gone',
}, {
'note': 'swf params escaped',
'url': 'https://www.facebook.com/barackobama/posts/10153664894881749',
'md5': '97ba073838964d12c70566e0085c2b91',
'info_dict': {
'id': '10153664894881749',
'ext': 'mp4',
'title': 'Average time to confirm recent Supreme Court nominees: 67 days Longest it\'s t...',
'thumbnail': r're:^https?://.*',
'timestamp': 1456259628,
'upload_date': '20160223',
'uploader': 'Barack Obama',
},
}, {
# have 1080P, but only up to 720p in swf params
# data.video.story.attachments[].media
'url': 'https://www.facebook.com/cnn/videos/10155529876156509/',
'md5': '9571fae53d4165bbbadb17a94651dcdc',
'info_dict': {
'id': '10155529876156509',
'ext': 'mp4',
'title': 'She survived the holocaust — and years later, shes getting her citizenship s...',
'timestamp': 1477818095,
'upload_date': '20161030',
'uploader': 'CNN',
'thumbnail': r're:^https?://.*',
'view_count': int,
},
}, {
# bigPipe.onPageletArrive ... onPageletArrive pagelet_group_mall
# data.node.comet_sections.content.story.attachments[].style_type_renderer.attachment.media
'url': 'https://www.facebook.com/yaroslav.korpan/videos/1417995061575415/',
'info_dict': {
'id': '1417995061575415',
'ext': 'mp4',
'title': 'md5:1db063d6a8c13faa8da727817339c857',
'timestamp': 1486648217,
'upload_date': '20170209',
'uploader': 'Yaroslav Korpan',
},
'params': {
'skip_download': True,
},
}, {
'url': 'https://www.facebook.com/LaGuiaDelVaron/posts/1072691702860471',
'info_dict': {
'id': '1072691702860471',
'ext': 'mp4',
'title': 'md5:ae2d22a93fbb12dad20dc393a869739d',
'timestamp': 1477305000,
'upload_date': '20161024',
'uploader': 'La Guía Del Varón',
'thumbnail': r're:^https?://.*',
},
'params': {
'skip_download': True,
},
}, {
# data.node.comet_sections.content.story.attachments[].style_type_renderer.attachment.media
'url': 'https://www.facebook.com/groups/1024490957622648/permalink/1396382447100162/',
'info_dict': {
'id': '1396382447100162',
'ext': 'mp4',
'title': 'md5:19a428bbde91364e3de815383b54a235',
'timestamp': 1486035494,
'upload_date': '20170202',
'uploader': 'Elisabeth Ahtn',
},
'params': {
'skip_download': True,
},
2014-08-27 09:08:47 +00:00
}, {
'url': 'https://www.facebook.com/video.php?v=10204634152394104',
'only_matching': True,
2015-04-16 16:08:52 +00:00
}, {
'url': 'https://www.facebook.com/amogood/videos/1618742068337349/?fref=nf',
'only_matching': True,
}, {
# data.mediaset.currMedia.edges
'url': 'https://www.facebook.com/ChristyClarkForBC/videos/vb.22819070941/10153870694020942/?type=2&theater',
'only_matching': True,
}, {
# data.video.story.attachments[].media
'url': 'facebook:544765982287235',
'only_matching': True,
}, {
# data.node.comet_sections.content.story.attachments[].style_type_renderer.attachment.media
'url': 'https://www.facebook.com/groups/164828000315060/permalink/764967300301124/',
'only_matching': True,
}, {
# data.video.creation_story.attachments[].media
'url': 'https://zh-hk.facebook.com/peoplespower/videos/1135894589806027/',
'only_matching': True,
}, {
# data.video
'url': 'https://www.facebookcorewwwi.onion/video.php?v=274175099429670',
'only_matching': True,
}, {
# no title
'url': 'https://www.facebook.com/onlycleverentertainment/videos/1947995502095005/',
'only_matching': True,
}, {
# data.video
'url': 'https://www.facebook.com/WatchESLOne/videos/359649331226507/',
'info_dict': {
'id': '359649331226507',
'ext': 'mp4',
'title': '#ESLOne VoD - Birmingham Finals Day#1 Fnatic vs. @Evil Geniuses',
'uploader': 'ESL One Dota 2',
},
'params': {
'skip_download': True,
},
}, {
# data.node.comet_sections.content.story.attachments[].style_type_renderer.attachment.all_subattachments.nodes[].media
'url': 'https://www.facebook.com/100033620354545/videos/106560053808006/',
'info_dict': {
'id': '106560053808006',
},
'playlist_count': 2,
}, {
# data.video.story.attachments[].media
'url': 'https://www.facebook.com/watch/?v=647537299265662',
'only_matching': True,
}, {
# data.node.comet_sections.content.story.attachments[].style_type_renderer.attachment.all_subattachments.nodes[].media
'url': 'https://www.facebook.com/PankajShahLondon/posts/10157667649866271',
'info_dict': {
'id': '10157667649866271',
},
'playlist_count': 3,
}, {
# data.nodes[].comet_sections.content.story.attachments[].style_type_renderer.attachment.media
'url': 'https://m.facebook.com/Alliance.Police.Department/posts/4048563708499330',
'info_dict': {
'id': '117576630041613',
'ext': 'mp4',
# TODO: title can be extracted from video page
'title': 'Facebook video #117576630041613',
'uploader_id': '189393014416438',
'upload_date': '20201123',
'timestamp': 1606162592,
},
'skip': 'Requires logging in',
}, {
# node.comet_sections.content.story.attached_story.attachments.style_type_renderer.attachment.media
'url': 'https://www.facebook.com/groups/ateistiskselskab/permalink/10154930137678856/',
'info_dict': {
'id': '211567722618337',
'ext': 'mp4',
'title': 'Facebook video #211567722618337',
'uploader_id': '127875227654254',
'upload_date': '20161122',
'timestamp': 1479793574,
},
}, {
# data.video.creation_story.attachments[].media
'url': 'https://www.facebook.com/watch/live/?v=1823658634322275',
'only_matching': True,
}, {
'url': 'https://www.facebook.com/watchparty/211641140192478',
'info_dict': {
'id': '211641140192478',
},
'playlist_count': 1,
'skip': 'Requires logging in',
2014-08-27 09:08:47 +00:00
}]
_SUPPORTED_PAGLETS_REGEX = r'(?:pagelet_group_mall|permalink_video_pagelet|hyperfeed_story_id_[0-9a-f]+)'
_api_config = {
'graphURI': '/api/graphql/'
}
2013-06-23 18:59:45 +00:00
@staticmethod
def _extract_urls(webpage):
urls = []
for mobj in re.finditer(
r'<iframe[^>]+?src=(["\'])(?P<url>https?://www\.facebook\.com/(?:video/embed|plugins/video\.php).+?)\1',
webpage):
urls.append(mobj.group('url'))
# Facebook API embed
# see https://developers.facebook.com/docs/plugins/embedded-video-player
for mobj in re.finditer(r'''(?x)<div[^>]+
class=(?P<q1>[\'"])[^\'"]*\bfb-(?:video|post)\b[^\'"]*(?P=q1)[^>]+
data-href=(?P<q2>[\'"])(?P<url>(?:https?:)?//(?:www\.)?facebook.com/.+?)(?P=q2)''', webpage):
urls.append(mobj.group('url'))
return urls
def _login(self):
useremail, password = self._get_login_info()
2013-06-23 18:59:45 +00:00
if useremail is None:
return
Switch codebase to use sanitized_Request instead of compat_urllib_request.Request [downloader/dash] Use sanitized_Request [downloader/http] Use sanitized_Request [atresplayer] Use sanitized_Request [bambuser] Use sanitized_Request [bliptv] Use sanitized_Request [brightcove] Use sanitized_Request [cbs] Use sanitized_Request [ceskatelevize] Use sanitized_Request [collegerama] Use sanitized_Request [extractor/common] Use sanitized_Request [crunchyroll] Use sanitized_Request [dailymotion] Use sanitized_Request [dcn] Use sanitized_Request [dramafever] Use sanitized_Request [dumpert] Use sanitized_Request [eitb] Use sanitized_Request [escapist] Use sanitized_Request [everyonesmixtape] Use sanitized_Request [extremetube] Use sanitized_Request [facebook] Use sanitized_Request [fc2] Use sanitized_Request [flickr] Use sanitized_Request [4tube] Use sanitized_Request [gdcvault] Use sanitized_Request [extractor/generic] Use sanitized_Request [hearthisat] Use sanitized_Request [hotnewhiphop] Use sanitized_Request [hypem] Use sanitized_Request [iprima] Use sanitized_Request [ivi] Use sanitized_Request [keezmovies] Use sanitized_Request [letv] Use sanitized_Request [lynda] Use sanitized_Request [metacafe] Use sanitized_Request [minhateca] Use sanitized_Request [miomio] Use sanitized_Request [meovideo] Use sanitized_Request [mofosex] Use sanitized_Request [moniker] Use sanitized_Request [mooshare] Use sanitized_Request [movieclips] Use sanitized_Request [mtv] Use sanitized_Request [myvideo] Use sanitized_Request [neteasemusic] Use sanitized_Request [nfb] Use sanitized_Request [niconico] Use sanitized_Request [noco] Use sanitized_Request [nosvideo] Use sanitized_Request [novamov] Use sanitized_Request [nowness] Use sanitized_Request [nuvid] Use sanitized_Request [played] Use sanitized_Request [pluralsight] Use sanitized_Request [pornhub] Use sanitized_Request [pornotube] Use sanitized_Request [primesharetv] Use sanitized_Request [promptfile] Use sanitized_Request [qqmusic] Use sanitized_Request [rtve] Use sanitized_Request [safari] Use sanitized_Request [sandia] Use sanitized_Request [shared] Use sanitized_Request [sharesix] Use sanitized_Request [sina] Use sanitized_Request [smotri] Use sanitized_Request [sohu] Use sanitized_Request [spankwire] Use sanitized_Request [sportdeutschland] Use sanitized_Request [streamcloud] Use sanitized_Request [streamcz] Use sanitized_Request [tapely] Use sanitized_Request [tube8] Use sanitized_Request [tubitv] Use sanitized_Request [twitch] Use sanitized_Request [twitter] Use sanitized_Request [udemy] Use sanitized_Request [vbox7] Use sanitized_Request [veoh] Use sanitized_Request [vessel] Use sanitized_Request [vevo] Use sanitized_Request [viddler] Use sanitized_Request [videomega] Use sanitized_Request [viewvster] Use sanitized_Request [viki] Use sanitized_Request [vk] Use sanitized_Request [vodlocker] Use sanitized_Request [voicerepublic] Use sanitized_Request [wistia] Use sanitized_Request [xfileshare] Use sanitized_Request [xtube] Use sanitized_Request [xvideos] Use sanitized_Request [yandexmusic] Use sanitized_Request [youku] Use sanitized_Request [youporn] Use sanitized_Request [youtube] Use sanitized_Request [patreon] Use sanitized_Request [extractor/common] Remove unused import [nfb] PEP 8
2015-11-21 16:18:17 +00:00
login_page_req = sanitized_Request(self._LOGIN_URL)
2015-12-28 15:37:02 +00:00
self._set_cookie('facebook.com', 'locale', 'en_US')
login_page = self._download_webpage(login_page_req, None,
2014-11-23 20:39:15 +00:00
note='Downloading login page',
errnote='Unable to download login page')
2014-03-04 02:39:04 +00:00
lsd = self._search_regex(
2014-03-04 02:39:45 +00:00
r'<input type="hidden" name="lsd" value="([^"]*)"',
2014-03-04 02:39:04 +00:00
login_page, 'lsd')
2014-03-04 02:36:54 +00:00
lgnrnd = self._search_regex(r'name="lgnrnd" value="([^"]*?)"', login_page, 'lgnrnd')
2013-06-23 18:59:45 +00:00
login_form = {
'email': useremail,
'pass': password,
'lsd': lsd,
'lgnrnd': lgnrnd,
'next': 'http://facebook.com/home.php',
'default_persistent': '0',
'legacy_return': '1',
'timezone': '-60',
'trynum': '1',
2014-11-23 21:21:46 +00:00
}
Switch codebase to use sanitized_Request instead of compat_urllib_request.Request [downloader/dash] Use sanitized_Request [downloader/http] Use sanitized_Request [atresplayer] Use sanitized_Request [bambuser] Use sanitized_Request [bliptv] Use sanitized_Request [brightcove] Use sanitized_Request [cbs] Use sanitized_Request [ceskatelevize] Use sanitized_Request [collegerama] Use sanitized_Request [extractor/common] Use sanitized_Request [crunchyroll] Use sanitized_Request [dailymotion] Use sanitized_Request [dcn] Use sanitized_Request [dramafever] Use sanitized_Request [dumpert] Use sanitized_Request [eitb] Use sanitized_Request [escapist] Use sanitized_Request [everyonesmixtape] Use sanitized_Request [extremetube] Use sanitized_Request [facebook] Use sanitized_Request [fc2] Use sanitized_Request [flickr] Use sanitized_Request [4tube] Use sanitized_Request [gdcvault] Use sanitized_Request [extractor/generic] Use sanitized_Request [hearthisat] Use sanitized_Request [hotnewhiphop] Use sanitized_Request [hypem] Use sanitized_Request [iprima] Use sanitized_Request [ivi] Use sanitized_Request [keezmovies] Use sanitized_Request [letv] Use sanitized_Request [lynda] Use sanitized_Request [metacafe] Use sanitized_Request [minhateca] Use sanitized_Request [miomio] Use sanitized_Request [meovideo] Use sanitized_Request [mofosex] Use sanitized_Request [moniker] Use sanitized_Request [mooshare] Use sanitized_Request [movieclips] Use sanitized_Request [mtv] Use sanitized_Request [myvideo] Use sanitized_Request [neteasemusic] Use sanitized_Request [nfb] Use sanitized_Request [niconico] Use sanitized_Request [noco] Use sanitized_Request [nosvideo] Use sanitized_Request [novamov] Use sanitized_Request [nowness] Use sanitized_Request [nuvid] Use sanitized_Request [played] Use sanitized_Request [pluralsight] Use sanitized_Request [pornhub] Use sanitized_Request [pornotube] Use sanitized_Request [primesharetv] Use sanitized_Request [promptfile] Use sanitized_Request [qqmusic] Use sanitized_Request [rtve] Use sanitized_Request [safari] Use sanitized_Request [sandia] Use sanitized_Request [shared] Use sanitized_Request [sharesix] Use sanitized_Request [sina] Use sanitized_Request [smotri] Use sanitized_Request [sohu] Use sanitized_Request [spankwire] Use sanitized_Request [sportdeutschland] Use sanitized_Request [streamcloud] Use sanitized_Request [streamcz] Use sanitized_Request [tapely] Use sanitized_Request [tube8] Use sanitized_Request [tubitv] Use sanitized_Request [twitch] Use sanitized_Request [twitter] Use sanitized_Request [udemy] Use sanitized_Request [vbox7] Use sanitized_Request [veoh] Use sanitized_Request [vessel] Use sanitized_Request [vevo] Use sanitized_Request [viddler] Use sanitized_Request [videomega] Use sanitized_Request [viewvster] Use sanitized_Request [viki] Use sanitized_Request [vk] Use sanitized_Request [vodlocker] Use sanitized_Request [voicerepublic] Use sanitized_Request [wistia] Use sanitized_Request [xfileshare] Use sanitized_Request [xtube] Use sanitized_Request [xvideos] Use sanitized_Request [yandexmusic] Use sanitized_Request [youku] Use sanitized_Request [youporn] Use sanitized_Request [youtube] Use sanitized_Request [patreon] Use sanitized_Request [extractor/common] Remove unused import [nfb] PEP 8
2015-11-21 16:18:17 +00:00
request = sanitized_Request(self._LOGIN_URL, urlencode_postdata(login_form))
request.add_header('Content-Type', 'application/x-www-form-urlencoded')
2013-06-23 18:59:45 +00:00
try:
login_results = self._download_webpage(request, None,
2014-11-23 20:39:15 +00:00
note='Logging in', errnote='unable to fetch login page')
2013-06-23 18:59:45 +00:00
if re.search(r'<form(.*)name="login"(.*)</form>', login_results) is not None:
2015-12-28 15:20:09 +00:00
error = self._html_search_regex(
r'(?s)<div[^>]+class=(["\']).*?login_error_box.*?\1[^>]*><div[^>]*>.*?</div><div[^>]*>(?P<error>.+?)</div>',
login_results, 'login error', default=None, group='error')
if error:
raise ExtractorError('Unable to login: %s' % error, expected=True)
2016-01-10 15:17:47 +00:00
self._downloader.report_warning('unable to log in: bad username/password, or exceeded login rate limit (~3/min). Check credentials or wait.')
2013-06-23 18:59:45 +00:00
return
2015-12-28 15:37:02 +00:00
fb_dtsg = self._search_regex(
r'name="fb_dtsg" value="(.+?)"', login_results, 'fb_dtsg', default=None)
h = self._search_regex(
r'name="h"\s+(?:\w+="[^"]+"\s+)*?value="([^"]+)"', login_results, 'h', default=None)
if not fb_dtsg or not h:
return
check_form = {
2015-12-28 15:37:02 +00:00
'fb_dtsg': fb_dtsg,
'h': h,
'name_action_selected': 'dont_save',
}
Switch codebase to use sanitized_Request instead of compat_urllib_request.Request [downloader/dash] Use sanitized_Request [downloader/http] Use sanitized_Request [atresplayer] Use sanitized_Request [bambuser] Use sanitized_Request [bliptv] Use sanitized_Request [brightcove] Use sanitized_Request [cbs] Use sanitized_Request [ceskatelevize] Use sanitized_Request [collegerama] Use sanitized_Request [extractor/common] Use sanitized_Request [crunchyroll] Use sanitized_Request [dailymotion] Use sanitized_Request [dcn] Use sanitized_Request [dramafever] Use sanitized_Request [dumpert] Use sanitized_Request [eitb] Use sanitized_Request [escapist] Use sanitized_Request [everyonesmixtape] Use sanitized_Request [extremetube] Use sanitized_Request [facebook] Use sanitized_Request [fc2] Use sanitized_Request [flickr] Use sanitized_Request [4tube] Use sanitized_Request [gdcvault] Use sanitized_Request [extractor/generic] Use sanitized_Request [hearthisat] Use sanitized_Request [hotnewhiphop] Use sanitized_Request [hypem] Use sanitized_Request [iprima] Use sanitized_Request [ivi] Use sanitized_Request [keezmovies] Use sanitized_Request [letv] Use sanitized_Request [lynda] Use sanitized_Request [metacafe] Use sanitized_Request [minhateca] Use sanitized_Request [miomio] Use sanitized_Request [meovideo] Use sanitized_Request [mofosex] Use sanitized_Request [moniker] Use sanitized_Request [mooshare] Use sanitized_Request [movieclips] Use sanitized_Request [mtv] Use sanitized_Request [myvideo] Use sanitized_Request [neteasemusic] Use sanitized_Request [nfb] Use sanitized_Request [niconico] Use sanitized_Request [noco] Use sanitized_Request [nosvideo] Use sanitized_Request [novamov] Use sanitized_Request [nowness] Use sanitized_Request [nuvid] Use sanitized_Request [played] Use sanitized_Request [pluralsight] Use sanitized_Request [pornhub] Use sanitized_Request [pornotube] Use sanitized_Request [primesharetv] Use sanitized_Request [promptfile] Use sanitized_Request [qqmusic] Use sanitized_Request [rtve] Use sanitized_Request [safari] Use sanitized_Request [sandia] Use sanitized_Request [shared] Use sanitized_Request [sharesix] Use sanitized_Request [sina] Use sanitized_Request [smotri] Use sanitized_Request [sohu] Use sanitized_Request [spankwire] Use sanitized_Request [sportdeutschland] Use sanitized_Request [streamcloud] Use sanitized_Request [streamcz] Use sanitized_Request [tapely] Use sanitized_Request [tube8] Use sanitized_Request [tubitv] Use sanitized_Request [twitch] Use sanitized_Request [twitter] Use sanitized_Request [udemy] Use sanitized_Request [vbox7] Use sanitized_Request [veoh] Use sanitized_Request [vessel] Use sanitized_Request [vevo] Use sanitized_Request [viddler] Use sanitized_Request [videomega] Use sanitized_Request [viewvster] Use sanitized_Request [viki] Use sanitized_Request [vk] Use sanitized_Request [vodlocker] Use sanitized_Request [voicerepublic] Use sanitized_Request [wistia] Use sanitized_Request [xfileshare] Use sanitized_Request [xtube] Use sanitized_Request [xvideos] Use sanitized_Request [yandexmusic] Use sanitized_Request [youku] Use sanitized_Request [youporn] Use sanitized_Request [youtube] Use sanitized_Request [patreon] Use sanitized_Request [extractor/common] Remove unused import [nfb] PEP 8
2015-11-21 16:18:17 +00:00
check_req = sanitized_Request(self._CHECKPOINT_URL, urlencode_postdata(check_form))
check_req.add_header('Content-Type', 'application/x-www-form-urlencoded')
check_response = self._download_webpage(check_req, None,
2014-11-23 20:39:15 +00:00
note='Confirming login')
if re.search(r'id="checkpointSubmitButton"', check_response) is not None:
2016-01-10 15:17:47 +00:00
self._downloader.report_warning('Unable to confirm login, you have to login in your browser and authorize the login.')
2013-06-23 18:59:45 +00:00
except (compat_urllib_error.URLError, compat_http_client.HTTPException, socket.error) as err:
self._downloader.report_warning('unable to log in: %s' % error_to_compat_str(err))
2013-06-23 18:59:45 +00:00
return
def _real_initialize(self):
self._login()
def _extract_from_url(self, url, video_id):
webpage = self._download_webpage(
url.replace('://m.facebook.com/', '://www.facebook.com/'), video_id)
2013-06-23 18:59:45 +00:00
video_data = None
def extract_video_data(instances):
video_data = []
for item in instances:
2020-12-11 20:24:08 +00:00
if try_get(item, lambda x: x[1][0]) == 'VideoConfig':
video_item = item[2][0]
if video_item.get('video_id'):
video_data.append(video_item['videoData'])
return video_data
server_js_data = self._parse_json(self._search_regex(
[r'handleServerJS\(({.+})(?:\);|,")', r'\bs\.handle\(({.+?})\);'],
webpage, 'server js data', default='{}'), video_id, fatal=False)
if server_js_data:
video_data = extract_video_data(server_js_data.get('instances', []))
def extract_from_jsmods_instances(js_data):
if js_data:
return extract_video_data(try_get(
js_data, lambda x: x['jsmods']['instances'], list) or [])
def extract_dash_manifest(video, formats):
dash_manifest = video.get('dash_manifest')
if dash_manifest:
formats.extend(self._parse_mpd_formats(
compat_etree_fromstring(compat_urllib_parse_unquote_plus(dash_manifest))))
def process_formats(formats):
# Downloads with browser's User-Agent are rate limited. Working around
# with non-browser User-Agent.
for f in formats:
f.setdefault('http_headers', {})['User-Agent'] = 'facebookexternalhit/1.1'
self._sort_formats(formats)
def extract_relay_data(_filter):
return self._parse_json(self._search_regex(
r'handleWithCustomApplyEach\([^,]+,\s*({.*?%s.*?})\);' % _filter,
webpage, 'replay data', default='{}'), video_id, fatal=False) or {}
def extract_relay_prefetched_data(_filter):
replay_data = extract_relay_data(_filter)
for require in (replay_data.get('require') or []):
if require[0] == 'RelayPrefetchedStreamCache':
return try_get(require, lambda x: x[3][1]['__bbox']['result']['data'], dict) or {}
if not video_data:
server_js_data = self._parse_json(self._search_regex([
r'bigPipe\.onPageletArrive\(({.+?})\)\s*;\s*}\s*\)\s*,\s*["\']onPageletArrive\s+' + self._SUPPORTED_PAGLETS_REGEX,
r'bigPipe\.onPageletArrive\(({.*?id\s*:\s*"%s".*?})\);' % self._SUPPORTED_PAGLETS_REGEX
], webpage, 'js data', default='{}'), video_id, js_to_json, False)
video_data = extract_from_jsmods_instances(server_js_data)
if not video_data:
data = extract_relay_prefetched_data(
r'"(?:dash_manifest|playable_url(?:_quality_hd)?)"\s*:\s*"[^"]+"')
if data:
entries = []
def parse_graphql_video(video):
formats = []
q = qualities(['sd', 'hd'])
for (suffix, format_id) in [('', 'sd'), ('_quality_hd', 'hd')]:
playable_url = video.get('playable_url' + suffix)
if not playable_url:
continue
formats.append({
'format_id': format_id,
'quality': q(format_id),
'url': playable_url,
})
extract_dash_manifest(video, formats)
process_formats(formats)
v_id = video.get('videoId') or video.get('id') or video_id
info = {
'id': v_id,
'formats': formats,
'thumbnail': try_get(video, lambda x: x['thumbnailImage']['uri']),
'uploader_id': try_get(video, lambda x: x['owner']['id']),
'timestamp': int_or_none(video.get('publish_time')),
'duration': float_or_none(video.get('playable_duration_in_ms'), 1000),
}
description = try_get(video, lambda x: x['savable_description']['text'])
title = video.get('name')
if title:
info.update({
'title': title,
'description': description,
})
else:
info['title'] = description or 'Facebook video #%s' % v_id
entries.append(info)
def parse_attachment(attachment, key='media'):
media = attachment.get(key) or {}
if media.get('__typename') == 'Video':
return parse_graphql_video(media)
nodes = data.get('nodes') or []
node = data.get('node') or {}
if not nodes and node:
nodes.append(node)
for node in nodes:
story = try_get(node, lambda x: x['comet_sections']['content']['story'], dict) or {}
attachments = try_get(story, [
lambda x: x['attached_story']['attachments'],
lambda x: x['attachments']
], list) or []
for attachment in attachments:
attachment = try_get(attachment, lambda x: x['style_type_renderer']['attachment'], dict)
ns = try_get(attachment, lambda x: x['all_subattachments']['nodes'], list) or []
for n in ns:
parse_attachment(n)
parse_attachment(attachment)
edges = try_get(data, lambda x: x['mediaset']['currMedia']['edges'], list) or []
for edge in edges:
parse_attachment(edge, key='node')
video = data.get('video') or {}
if video:
attachments = try_get(video, [
lambda x: x['story']['attachments'],
lambda x: x['creation_story']['attachments']
], list) or []
for attachment in attachments:
parse_attachment(attachment)
if not entries:
parse_graphql_video(video)
return self.playlist_result(entries, video_id)
if not video_data:
m_msg = re.search(r'class="[^"]*uiInterstitialContent[^"]*"><div>(.*?)</div>', webpage)
if m_msg is not None:
raise ExtractorError(
'The video is not available, Facebook said: "%s"' % m_msg.group(1),
expected=True)
elif '>You must log in to continue' in webpage:
self.raise_login_required()
if not video_data and '/watchparty/' in url:
post_data = {
'doc_id': 3731964053542869,
'variables': json.dumps({
'livingRoomID': video_id,
}),
}
prefetched_data = extract_relay_prefetched_data(r'"login_data"\s*:\s*{')
if prefetched_data:
lsd = try_get(prefetched_data, lambda x: x['login_data']['lsd'], dict)
if lsd:
post_data[lsd['name']] = lsd['value']
relay_data = extract_relay_data(r'\[\s*"RelayAPIConfigDefaults"\s*,')
for define in (relay_data.get('define') or []):
if define[0] == 'RelayAPIConfigDefaults':
self._api_config = define[2]
living_room = self._download_json(
urljoin(url, self._api_config['graphURI']), video_id,
data=urlencode_postdata(post_data))['data']['living_room']
entries = []
for edge in (try_get(living_room, lambda x: x['recap']['watched_content']['edges']) or []):
video = try_get(edge, lambda x: x['node']['video']) or {}
v_id = video.get('id')
if not v_id:
continue
v_id = compat_str(v_id)
entries.append(self.url_result(
self._VIDEO_PAGE_TEMPLATE % v_id,
self.ie_key(), v_id, video.get('name')))
return self.playlist_result(entries, video_id)
if not video_data:
# Video info not in first request, do a secondary request using
# tahoe player specific URL
tahoe_data = self._download_webpage(
self._VIDEO_PAGE_TAHOE_TEMPLATE % video_id, video_id,
data=urlencode_postdata({
'__a': 1,
'__pc': self._search_regex(
r'pkg_cohort["\']\s*:\s*["\'](.+?)["\']', webpage,
'pkg cohort', default='PHASED:DEFAULT'),
'__rev': self._search_regex(
r'client_revision["\']\s*:\s*(\d+),', webpage,
'client revision', default='3944515'),
'fb_dtsg': self._search_regex(
r'"DTSGInitialData"\s*,\s*\[\]\s*,\s*{\s*"token"\s*:\s*"([^"]+)"',
webpage, 'dtsg token', default=''),
}),
headers={
'Content-Type': 'application/x-www-form-urlencoded',
})
tahoe_js_data = self._parse_json(
self._search_regex(
r'for\s+\(\s*;\s*;\s*\)\s*;(.+)', tahoe_data,
'tahoe js data', default='{}'),
video_id, fatal=False)
video_data = extract_from_jsmods_instances(tahoe_js_data)
if not video_data:
raise ExtractorError('Cannot parse data')
if len(video_data) > 1:
entries = []
for v in video_data:
video_url = v[0].get('video_url')
if not video_url:
continue
entries.append(self.url_result(urljoin(
url, video_url), self.ie_key(), v[0].get('video_id')))
return self.playlist_result(entries, video_id)
video_data = video_data[0]
formats = []
subtitles = {}
for f in video_data:
format_id = f['stream_type']
if f and isinstance(f, dict):
f = [f]
if not f or not isinstance(f, list):
continue
for quality in ('sd', 'hd'):
for src_type in ('src', 'src_no_ratelimit'):
src = f[0].get('%s_%s' % (quality, src_type))
if src:
preference = -10 if format_id == 'progressive' else 0
if quality == 'hd':
preference += 5
formats.append({
'format_id': '%s_%s_%s' % (format_id, quality, src_type),
'url': src,
'preference': preference,
})
extract_dash_manifest(f[0], formats)
subtitles_src = f[0].get('subtitles_src')
if subtitles_src:
subtitles.setdefault('en', []).append({'url': subtitles_src})
if not formats:
raise ExtractorError('Cannot find video formats')
2013-06-23 18:59:45 +00:00
process_formats(formats)
video_title = self._html_search_regex(
r'<h2\s+[^>]*class="uiHeaderTitle"[^>]*>([^<]*)</h2>', webpage,
'title', default=None)
if not video_title:
video_title = self._html_search_regex(
r'(?s)<span class="fbPhotosPhotoCaption".*?id="fbPhotoPageCaption"><span class="hasCaption">(.*?)</span>',
webpage, 'alternative title', default=None)
if not video_title:
video_title = self._html_search_meta(
'description', webpage, 'title', default=None)
if video_title:
video_title = limit_length(video_title, 80)
else:
video_title = 'Facebook video #%s' % video_id
uploader = clean_html(get_element_by_id(
'fbPhotoPageAuthorName', webpage)) or self._search_regex(
r'ownerName\s*:\s*"([^"]+)"', webpage, 'uploader',
default=None) or self._og_search_title(webpage, fatal=False)
timestamp = int_or_none(self._search_regex(
r'<abbr[^>]+data-utime=["\'](\d+)', webpage,
'timestamp', default=None))
thumbnail = self._html_search_meta(['og:image', 'twitter:image'], webpage)
2013-06-23 18:59:45 +00:00
view_count = parse_count(self._search_regex(
r'\bviewCount\s*:\s*["\']([\d,.]+)', webpage, 'view count',
default=None))
info_dict = {
2013-06-23 18:59:45 +00:00
'id': video_id,
'title': video_title,
'formats': formats,
2015-08-02 21:52:12 +00:00
'uploader': uploader,
'timestamp': timestamp,
'thumbnail': thumbnail,
'view_count': view_count,
'subtitles': subtitles,
2013-06-23 18:59:45 +00:00
}
return info_dict
def _real_extract(self, url):
video_id = self._match_id(url)
real_url = self._VIDEO_PAGE_TEMPLATE % video_id if url.startswith('facebook:') else url
return self._extract_from_url(real_url, video_id)
class FacebookPluginsVideoIE(InfoExtractor):
_VALID_URL = r'https?://(?:[\w-]+\.)?facebook\.com/plugins/video\.php\?.*?\bhref=(?P<id>https.+)'
_TESTS = [{
'url': 'https://www.facebook.com/plugins/video.php?href=https%3A%2F%2Fwww.facebook.com%2Fgov.sg%2Fvideos%2F10154383743583686%2F&show_text=0&width=560',
'md5': '5954e92cdfe51fe5782ae9bda7058a07',
'info_dict': {
'id': '10154383743583686',
'ext': 'mp4',
'title': 'What to do during the haze?',
'uploader': 'Gov.sg',
'upload_date': '20160826',
'timestamp': 1472184808,
},
'add_ie': [FacebookIE.ie_key()],
}, {
'url': 'https://www.facebook.com/plugins/video.php?href=https%3A%2F%2Fwww.facebook.com%2Fvideo.php%3Fv%3D10204634152394104',
'only_matching': True,
}, {
'url': 'https://www.facebook.com/plugins/video.php?href=https://www.facebook.com/gov.sg/videos/10154383743583686/&show_text=0&width=560',
'only_matching': True,
}]
def _real_extract(self, url):
return self.url_result(
compat_urllib_parse_unquote(self._match_id(url)),
FacebookIE.ie_key())