mirror of
https://github.com/yt-dlp/yt-dlp.git
synced 2024-11-30 17:49:34 -05:00
145 lines
4.9 KiB
Python
145 lines
4.9 KiB
Python
|
from __future__ import unicode_literals
|
||
|
|
||
|
import re
|
||
|
|
||
|
from .common import InfoExtractor
|
||
|
from ..utils import (
|
||
|
decode_packed_codes,
|
||
|
ExtractorError,
|
||
|
urlencode_postdata,
|
||
|
)
|
||
|
|
||
|
|
||
|
class SCTEBaseIE(InfoExtractor):
|
||
|
_LOGIN_URL = 'https://www.scte.org/SCTE/Sign_In.aspx'
|
||
|
_NETRC_MACHINE = 'scte'
|
||
|
|
||
|
def _real_initialize(self):
|
||
|
self._login()
|
||
|
|
||
|
def _login(self):
|
||
|
username, password = self._get_login_info()
|
||
|
if username is None:
|
||
|
return
|
||
|
|
||
|
login_popup = self._download_webpage(
|
||
|
self._LOGIN_URL, None, 'Downloading login popup')
|
||
|
|
||
|
def is_logged(webpage):
|
||
|
return any(re.search(p, webpage) for p in (
|
||
|
r'class=["\']welcome\b', r'>Sign Out<'))
|
||
|
|
||
|
# already logged in
|
||
|
if is_logged(login_popup):
|
||
|
return
|
||
|
|
||
|
login_form = self._hidden_inputs(login_popup)
|
||
|
|
||
|
login_form.update({
|
||
|
'ctl01$TemplateBody$WebPartManager1$gwpciNewContactSignInCommon$ciNewContactSignInCommon$signInUserName': username,
|
||
|
'ctl01$TemplateBody$WebPartManager1$gwpciNewContactSignInCommon$ciNewContactSignInCommon$signInPassword': password,
|
||
|
'ctl01$TemplateBody$WebPartManager1$gwpciNewContactSignInCommon$ciNewContactSignInCommon$RememberMe': 'on',
|
||
|
})
|
||
|
|
||
|
response = self._download_webpage(
|
||
|
self._LOGIN_URL, None, 'Logging in',
|
||
|
data=urlencode_postdata(login_form))
|
||
|
|
||
|
if '|pageRedirect|' not in response and not is_logged(response):
|
||
|
error = self._html_search_regex(
|
||
|
r'(?s)<[^>]+class=["\']AsiError["\'][^>]*>(.+?)</',
|
||
|
response, 'error message', default=None)
|
||
|
if error:
|
||
|
raise ExtractorError('Unable to login: %s' % error, expected=True)
|
||
|
raise ExtractorError('Unable to log in')
|
||
|
|
||
|
|
||
|
class SCTEIE(SCTEBaseIE):
|
||
|
_VALID_URL = r'https?://learning\.scte\.org/mod/scorm/view\.php?.*?\bid=(?P<id>\d+)'
|
||
|
_TESTS = [{
|
||
|
'url': 'https://learning.scte.org/mod/scorm/view.php?id=31484',
|
||
|
'info_dict': {
|
||
|
'title': 'Introduction to DOCSIS Engineering Professional',
|
||
|
'id': '31484',
|
||
|
},
|
||
|
'playlist_count': 5,
|
||
|
'skip': 'Requires account credentials',
|
||
|
}]
|
||
|
|
||
|
def _real_extract(self, url):
|
||
|
video_id = self._match_id(url)
|
||
|
|
||
|
webpage = self._download_webpage(url, video_id)
|
||
|
|
||
|
title = self._search_regex(r'<h1>(.+?)</h1>', webpage, 'title')
|
||
|
|
||
|
context_id = self._search_regex(r'context-(\d+)', webpage, video_id)
|
||
|
content_base = 'https://learning.scte.org/pluginfile.php/%s/mod_scorm/content/8/' % context_id
|
||
|
context = decode_packed_codes(self._download_webpage(
|
||
|
'%smobile/data.js' % content_base, video_id))
|
||
|
|
||
|
data = self._parse_xml(
|
||
|
self._search_regex(
|
||
|
r'CreateData\(\s*"(.+?)"', context, 'data').replace(r"\'", "'"),
|
||
|
video_id)
|
||
|
|
||
|
entries = []
|
||
|
for asset in data.findall('.//asset'):
|
||
|
asset_url = asset.get('url')
|
||
|
if not asset_url or not asset_url.endswith('.mp4'):
|
||
|
continue
|
||
|
asset_id = self._search_regex(
|
||
|
r'video_([^_]+)_', asset_url, 'asset id', default=None)
|
||
|
if not asset_id:
|
||
|
continue
|
||
|
entries.append({
|
||
|
'id': asset_id,
|
||
|
'title': title,
|
||
|
'url': content_base + asset_url,
|
||
|
})
|
||
|
|
||
|
return self.playlist_result(entries, video_id, title)
|
||
|
|
||
|
|
||
|
class SCTECourseIE(SCTEBaseIE):
|
||
|
_VALID_URL = r'https?://learning\.scte\.org/(?:mod/sub)?course/view\.php?.*?\bid=(?P<id>\d+)'
|
||
|
_TESTS = [{
|
||
|
'url': 'https://learning.scte.org/mod/subcourse/view.php?id=31491',
|
||
|
'only_matching': True,
|
||
|
}, {
|
||
|
'url': 'https://learning.scte.org/course/view.php?id=3639',
|
||
|
'only_matching': True,
|
||
|
}, {
|
||
|
'url': 'https://learning.scte.org/course/view.php?id=3073',
|
||
|
'only_matching': True,
|
||
|
}]
|
||
|
|
||
|
def _real_extract(self, url):
|
||
|
course_id = self._match_id(url)
|
||
|
|
||
|
webpage = self._download_webpage(url, course_id)
|
||
|
|
||
|
title = self._search_regex(
|
||
|
r'<h1>(.+?)</h1>', webpage, 'title', default=None)
|
||
|
|
||
|
entries = []
|
||
|
for mobj in re.finditer(
|
||
|
r'''(?x)
|
||
|
<a[^>]+
|
||
|
href=(["\'])
|
||
|
(?P<url>
|
||
|
https?://learning\.scte\.org/mod/
|
||
|
(?P<kind>scorm|subcourse)/view\.php?(?:(?!\1).)*?
|
||
|
\bid=\d+
|
||
|
)
|
||
|
''',
|
||
|
webpage):
|
||
|
item_url = mobj.group('url')
|
||
|
if item_url == url:
|
||
|
continue
|
||
|
ie = (SCTEIE.ie_key() if mobj.group('kind') == 'scorm'
|
||
|
else SCTECourseIE.ie_key())
|
||
|
entries.append(self.url_result(item_url, ie=ie))
|
||
|
|
||
|
return self.playlist_result(entries, course_id, title)
|