mirror of
https://source.netsyms.com/Mirrors/youtube-dl
synced 2024-11-03 03:40:20 +00:00
[lynda] Add support for member accounts and paid videos (Closes #2125)
This commit is contained in:
parent
db22af36ec
commit
7ee40b5d1c
@ -5,13 +5,24 @@ import json
|
|||||||
|
|
||||||
from .subtitles import SubtitlesInfoExtractor
|
from .subtitles import SubtitlesInfoExtractor
|
||||||
from .common import InfoExtractor
|
from .common import InfoExtractor
|
||||||
from ..utils import ExtractorError
|
from ..utils import (
|
||||||
|
compat_urllib_parse,
|
||||||
|
compat_urllib_request,
|
||||||
|
ExtractorError
|
||||||
|
)
|
||||||
|
|
||||||
|
|
||||||
class LyndaIE(SubtitlesInfoExtractor):
|
class LyndaIE(SubtitlesInfoExtractor):
|
||||||
IE_NAME = 'lynda'
|
IE_NAME = 'lynda'
|
||||||
IE_DESC = 'lynda.com videos'
|
IE_DESC = 'lynda.com videos'
|
||||||
_VALID_URL = r'https?://www\.lynda\.com/[^/]+/[^/]+/\d+/(\d+)-\d\.html'
|
_VALID_URL = r'https?://www\.lynda\.com/[^/]+/[^/]+/\d+/(\d+)-\d\.html'
|
||||||
|
_LOGIN_URL = 'https://www.lynda.com/login/login.aspx'
|
||||||
|
_NETRC_MACHINE = 'lynda'
|
||||||
|
|
||||||
|
_SUCCESSFUL_LOGIN_REGEX = r'<a href="https://www.lynda.com/home/userAccount/ChangeContactInfo.aspx" data-qa="eyebrow_account_menu">My account'
|
||||||
|
_TIMECODE_REGEX = r'\[(?P<timecode>\d+:\d+:\d+[\.,]\d+)\]'
|
||||||
|
|
||||||
|
ACCOUNT_CREDENTIALS_HINT = 'Use --username and --password options to provide lynda.com account credentials.'
|
||||||
|
|
||||||
_TEST = {
|
_TEST = {
|
||||||
'url': 'http://www.lynda.com/Bootstrap-tutorials/Using-exercise-files/110885/114408-4.html',
|
'url': 'http://www.lynda.com/Bootstrap-tutorials/Using-exercise-files/110885/114408-4.html',
|
||||||
@ -22,7 +33,10 @@ class LyndaIE(SubtitlesInfoExtractor):
|
|||||||
'duration': 68
|
'duration': 68
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
|
def _real_initialize(self):
|
||||||
|
self._login()
|
||||||
|
|
||||||
def _real_extract(self, url):
|
def _real_extract(self, url):
|
||||||
mobj = re.match(self._VALID_URL, url)
|
mobj = re.match(self._VALID_URL, url)
|
||||||
video_id = mobj.group(1)
|
video_id = mobj.group(1)
|
||||||
@ -31,11 +45,11 @@ class LyndaIE(SubtitlesInfoExtractor):
|
|||||||
video_id, 'Downloading video JSON')
|
video_id, 'Downloading video JSON')
|
||||||
video_json = json.loads(page)
|
video_json = json.loads(page)
|
||||||
|
|
||||||
if 'Status' in video_json and video_json['Status'] == 'NotFound':
|
if 'Status' in video_json:
|
||||||
raise ExtractorError('Video %s does not exist' % video_id, expected=True)
|
raise ExtractorError('lynda returned error: %s' % video_json['Message'], expected=True)
|
||||||
|
|
||||||
if video_json['HasAccess'] is False:
|
if video_json['HasAccess'] is False:
|
||||||
raise ExtractorError('Video %s is only available for members' % video_id, expected=True)
|
raise ExtractorError('Video %s is only available for members. ' % video_id + self.ACCOUNT_CREDENTIALS_HINT, expected=True)
|
||||||
|
|
||||||
video_id = video_json['ID']
|
video_id = video_json['ID']
|
||||||
duration = video_json['DurationInSeconds']
|
duration = video_json['DurationInSeconds']
|
||||||
@ -50,13 +64,13 @@ class LyndaIE(SubtitlesInfoExtractor):
|
|||||||
} for fmt in video_json['Formats']]
|
} for fmt in video_json['Formats']]
|
||||||
|
|
||||||
self._sort_formats(formats)
|
self._sort_formats(formats)
|
||||||
|
|
||||||
if self._downloader.params.get('listsubtitles', False):
|
if self._downloader.params.get('listsubtitles', False):
|
||||||
self._list_available_subtitles(video_id, page)
|
self._list_available_subtitles(video_id, page)
|
||||||
return
|
return
|
||||||
|
|
||||||
subtitles = self._fix_subtitles(self.extract_subtitles(video_id, page))
|
subtitles = self._fix_subtitles(self.extract_subtitles(video_id, page))
|
||||||
|
|
||||||
return {
|
return {
|
||||||
'id': video_id,
|
'id': video_id,
|
||||||
'title': title,
|
'title': title,
|
||||||
@ -64,8 +78,48 @@ class LyndaIE(SubtitlesInfoExtractor):
|
|||||||
'subtitles': subtitles,
|
'subtitles': subtitles,
|
||||||
'formats': formats
|
'formats': formats
|
||||||
}
|
}
|
||||||
|
|
||||||
_TIMECODE_REGEX = r'\[(?P<timecode>\d+:\d+:\d+[\.,]\d+)\]'
|
def _login(self):
|
||||||
|
(username, password) = self._get_login_info()
|
||||||
|
if username is None:
|
||||||
|
return
|
||||||
|
|
||||||
|
login_form = {
|
||||||
|
'username': username,
|
||||||
|
'password': password,
|
||||||
|
'remember': 'false',
|
||||||
|
'stayPut': 'false'
|
||||||
|
}
|
||||||
|
request = compat_urllib_request.Request(self._LOGIN_URL, compat_urllib_parse.urlencode(login_form))
|
||||||
|
login_page = self._download_webpage(request, None, note='Logging in as %s' % username)
|
||||||
|
|
||||||
|
# Not (yet) logged in
|
||||||
|
m = re.search(r'loginResultJson = \'(?P<json>[^\']+)\';', login_page)
|
||||||
|
if m is not None:
|
||||||
|
response = m.group('json')
|
||||||
|
response_json = json.loads(response)
|
||||||
|
state = response_json['state']
|
||||||
|
|
||||||
|
if state == 'notlogged':
|
||||||
|
raise ExtractorError('Unable to login, incorrect username and/or password', expected=True)
|
||||||
|
|
||||||
|
# This is when we get popup:
|
||||||
|
# > You're already logged in to lynda.com on two devices.
|
||||||
|
# > If you log in here, we'll log you out of another device.
|
||||||
|
# So, we need to confirm this.
|
||||||
|
if state == 'conflicted':
|
||||||
|
confirm_form = {
|
||||||
|
'username': '',
|
||||||
|
'password': '',
|
||||||
|
'resolve': 'true',
|
||||||
|
'remember': 'false',
|
||||||
|
'stayPut': 'false',
|
||||||
|
}
|
||||||
|
request = compat_urllib_request.Request(self._LOGIN_URL, compat_urllib_parse.urlencode(confirm_form))
|
||||||
|
login_page = self._download_webpage(request, None, note='Confirming log in and log out from another device')
|
||||||
|
|
||||||
|
if re.search(self._SUCCESSFUL_LOGIN_REGEX, login_page) is None:
|
||||||
|
raise ExtractorError('Unable to log in')
|
||||||
|
|
||||||
def _fix_subtitles(self, subtitles):
|
def _fix_subtitles(self, subtitles):
|
||||||
if subtitles is None:
|
if subtitles is None:
|
||||||
@ -93,7 +147,7 @@ class LyndaIE(SubtitlesInfoExtractor):
|
|||||||
if srt:
|
if srt:
|
||||||
fixed_subtitles[k] = srt
|
fixed_subtitles[k] = srt
|
||||||
return fixed_subtitles
|
return fixed_subtitles
|
||||||
|
|
||||||
def _get_available_subtitles(self, video_id, webpage):
|
def _get_available_subtitles(self, video_id, webpage):
|
||||||
url = 'http://www.lynda.com/ajax/player?videoId=%s&type=transcript' % video_id
|
url = 'http://www.lynda.com/ajax/player?videoId=%s&type=transcript' % video_id
|
||||||
sub = self._download_webpage(url, None, note=False)
|
sub = self._download_webpage(url, None, note=False)
|
||||||
@ -113,7 +167,7 @@ class LyndaCourseIE(InfoExtractor):
|
|||||||
mobj = re.match(self._VALID_URL, url)
|
mobj = re.match(self._VALID_URL, url)
|
||||||
course_path = mobj.group('coursepath')
|
course_path = mobj.group('coursepath')
|
||||||
course_id = mobj.group('courseid')
|
course_id = mobj.group('courseid')
|
||||||
|
|
||||||
page = self._download_webpage('http://www.lynda.com/ajax/player?courseId=%s&type=course' % course_id,
|
page = self._download_webpage('http://www.lynda.com/ajax/player?courseId=%s&type=course' % course_id,
|
||||||
course_id, 'Downloading course JSON')
|
course_id, 'Downloading course JSON')
|
||||||
course_json = json.loads(page)
|
course_json = json.loads(page)
|
||||||
@ -123,16 +177,18 @@ class LyndaCourseIE(InfoExtractor):
|
|||||||
|
|
||||||
unaccessible_videos = 0
|
unaccessible_videos = 0
|
||||||
videos = []
|
videos = []
|
||||||
|
(username, _) = self._get_login_info()
|
||||||
|
|
||||||
for chapter in course_json['Chapters']:
|
for chapter in course_json['Chapters']:
|
||||||
for video in chapter['Videos']:
|
for video in chapter['Videos']:
|
||||||
if video['HasAccess'] is not True:
|
if username is None and video['HasAccess'] is False:
|
||||||
unaccessible_videos += 1
|
unaccessible_videos += 1
|
||||||
continue
|
continue
|
||||||
videos.append(video['ID'])
|
videos.append(video['ID'])
|
||||||
|
|
||||||
if unaccessible_videos > 0:
|
if unaccessible_videos > 0:
|
||||||
self._downloader.report_warning('%s videos are only available for members and will not be downloaded' % unaccessible_videos)
|
self._downloader.report_warning('%s videos are only available for members and will not be downloaded. '
|
||||||
|
% unaccessible_videos + LyndaIE.ACCOUNT_CREDENTIALS_HINT)
|
||||||
|
|
||||||
entries = [
|
entries = [
|
||||||
self.url_result('http://www.lynda.com/%s/%s-4.html' %
|
self.url_result('http://www.lynda.com/%s/%s-4.html' %
|
||||||
@ -142,4 +198,4 @@ class LyndaCourseIE(InfoExtractor):
|
|||||||
|
|
||||||
course_title = course_json['Title']
|
course_title = course_json['Title']
|
||||||
|
|
||||||
return self.playlist_result(entries, course_id, course_title)
|
return self.playlist_result(entries, course_id, course_title)
|
Loading…
Reference in New Issue
Block a user