[theplatform] add basic support for Adobe Pass

pull/9988/head
Remita Amine 8 years ago
parent a0cfd82dda
commit bf83024826

@ -80,9 +80,6 @@ class CBSInteractiveIE(ThePlatformIE):
media_guid_path = 'media/guid/%d/%s' % (self.MPX_ACCOUNTS[site], vdata['mpxRefId'])
formats, subtitles = [], {}
if site == 'cnet':
formats, subtitles = self._extract_theplatform_smil(
self.TP_RELEASE_URL_TEMPLATE % media_guid_path, video_id)
for (fkey, vid) in vdata['files'].items():
if fkey == 'hls_phone' and 'hls_tablet' in vdata['files']:
continue
@ -94,7 +91,7 @@ class CBSInteractiveIE(ThePlatformIE):
subtitles = self._merge_subtitles(subtitles, tp_subtitles)
self._sort_formats(formats)
info = self.get_metadata('kYEXFC/%s' % media_guid_path, video_id)
info = self._extract_theplatform_metadata('kYEXFC/%s' % media_guid_path, video_id)
info.update({
'id': video_id,
'display_id': display_id,

@ -6,6 +6,7 @@ import time
import hmac
import binascii
import hashlib
import netrc
from .once import OnceIE
@ -24,6 +25,9 @@ from ..utils import (
xpath_with_ns,
mimetype2ext,
find_xpath_attr,
unescapeHTML,
urlencode_postdata,
unified_timestamp,
)
default_ns = 'http://www.w3.org/2005/SMIL21/Language'
@ -62,10 +66,11 @@ class ThePlatformBaseIE(OnceIE):
return formats, subtitles
def get_metadata(self, path, video_id):
def _download_theplatform_metadata(self, path, video_id):
info_url = 'http://link.theplatform.com/s/%s?format=preview' % path
info = self._download_json(info_url, video_id)
return self._download_json(info_url, video_id)
def _parse_theplatform_metadata(self, info):
subtitles = {}
captions = info.get('captions')
if isinstance(captions, list):
@ -86,6 +91,10 @@ class ThePlatformBaseIE(OnceIE):
'uploader': info.get('billingCode'),
}
def _extract_theplatform_metadata(self, path, video_id):
info = self._download_theplatform_metadata(path, video_id)
return self._parse_theplatform_metadata(info)
class ThePlatformIE(ThePlatformBaseIE):
_VALID_URL = r'''(?x)
@ -158,6 +167,7 @@ class ThePlatformIE(ThePlatformBaseIE):
'url': 'http://player.theplatform.com/p/NnzsPC/onsite_universal/select/media/guid/2410887629/2928790?fwsitesection=nbc_the_blacklist_video_library&autoPlay=true&carouselID=137781',
'only_matching': True,
}]
_SERVICE_PROVIDER_TEMPLATE = 'https://sp.auth.adobe.com/adobe-services/%s'
@classmethod
def _extract_urls(cls, webpage):
@ -192,6 +202,96 @@ class ThePlatformIE(ThePlatformBaseIE):
sig = flags + expiration_date + checksum + str_to_hex(sig_secret)
return '%s&sig=%s' % (url, sig)
def _extract_mvpd_auth(self, url, video_id, requestor_id, resource):
def xml_text(xml_str, tag):
return self._search_regex(
'<%s>(.+?)</%s>' % (tag, tag), xml_str, tag)
mvpd_headers = {
'ap_42': 'anonymous',
'ap_11': 'Linux i686',
'ap_z': 'Mozilla/5.0 (X11; Linux i686; rv:47.0) Gecko/20100101 Firefox/47.0',
'User-Agent': 'Mozilla/5.0 (X11; Linux i686; rv:47.0) Gecko/20100101 Firefox/47.0',
}
guid = xml_text(resource, 'guid')
requestor_info = self._downloader.cache.load('mvpd', requestor_id) or {}
authn_token = requestor_info.get('authn_token')
if authn_token:
token_expires = unified_timestamp(xml_text(authn_token, 'simpleTokenExpires').replace('_GMT', ''))
if token_expires and token_expires >= time.time():
authn_token = None
if not authn_token:
# TODO add support for other TV Providers
mso_id = 'DTV'
login_info = netrc.netrc().authenticators(mso_id)
if not login_info:
return None
def post_form(form_page, note, data={}):
post_url = self._html_search_regex(r'<form[^>]+action=(["\'])(?P<url>.+?)\1', form_page, 'post url', group='url')
return self._download_webpage(
post_url, video_id, note, data=urlencode_postdata(data or self._hidden_inputs(form_page)), headers={
'Content-Type': 'application/x-www-form-urlencoded',
})
provider_redirect_page = self._download_webpage(
self._SERVICE_PROVIDER_TEMPLATE % 'authenticate/saml', video_id,
'Downloading Provider Redirect Page', query={
'noflash': 'true',
'mso_id': mso_id,
'requestor_id': requestor_id,
'no_iframe': 'false',
'domain_name': 'adobe.com',
'redirect_url': url,
})
provider_login_page = post_form(
provider_redirect_page, 'Downloading Provider Login Page')
mvpd_confirm_page = post_form(provider_login_page, 'Logging in', {
'username': login_info[0],
'password': login_info[2],
})
post_form(mvpd_confirm_page, 'Confirming Login')
session = self._download_webpage(
self._SERVICE_PROVIDER_TEMPLATE % 'session', video_id,
'Retrieving Session', data=urlencode_postdata({
'_method': 'GET',
'requestor_id': requestor_id,
}), headers=mvpd_headers)
authn_token = unescapeHTML(xml_text(session, 'authnToken'))
requestor_info['authn_token'] = authn_token
self._downloader.cache.store('mvpd', requestor_id, requestor_info)
authz_token = requestor_info.get(guid)
if not authz_token:
authorize = self._download_webpage(
self._SERVICE_PROVIDER_TEMPLATE % 'authorize', video_id,
'Retrieving Authorization Token', data=urlencode_postdata({
'resource_id': resource,
'requestor_id': requestor_id,
'authentication_token': authn_token,
'mso_id': xml_text(authn_token, 'simpleTokenMsoID'),
'userMeta': '1',
}), headers=mvpd_headers)
authz_token = unescapeHTML(xml_text(authorize, 'authzToken'))
requestor_info[guid] = authz_token
self._downloader.cache.store('mvpd', requestor_id, requestor_info)
mvpd_headers.update({
'ap_19': xml_text(authn_token, 'simpleSamlNameID'),
'ap_23': xml_text(authn_token, 'simpleSamlSessionIndex'),
})
return self._download_webpage(
self._SERVICE_PROVIDER_TEMPLATE % 'shortAuthorize',
video_id, 'Retrieving Media Token', data=urlencode_postdata({
'authz_token': authz_token,
'requestor_id': requestor_id,
'session_guid': xml_text(authn_token, 'simpleTokenAuthenticationGuid'),
'hashed_guid': 'false',
}), headers=mvpd_headers)
def _real_extract(self, url):
url, smuggled_data = unsmuggle_url(url, {})
@ -265,7 +365,7 @@ class ThePlatformIE(ThePlatformBaseIE):
formats, subtitles = self._extract_theplatform_smil(smil_url, video_id)
self._sort_formats(formats)
ret = self.get_metadata(path, video_id)
ret = self._extract_theplatform_metadata(path, video_id)
combined_subtitles = self._merge_subtitles(ret.get('subtitles', {}), subtitles)
ret.update({
'id': video_id,
@ -339,7 +439,7 @@ class ThePlatformFeedIE(ThePlatformBaseIE):
timestamp = int_or_none(entry.get('media$availableDate'), scale=1000)
categories = [item['media$name'] for item in entry.get('media$categories', [])]
ret = self.get_metadata('%s/%s' % (provider_id, first_video_id), video_id)
ret = self._extract_theplatform_metadata('%s/%s' % (provider_id, first_video_id), video_id)
subtitles = self._merge_subtitles(subtitles, ret['subtitles'])
ret.update({
'id': video_id,

Loading…
Cancel
Save