aboutsummaryrefslogtreecommitdiff
path: root/youtube_dl/extractor/lynda.py
diff options
context:
space:
mode:
Diffstat (limited to 'youtube_dl/extractor/lynda.py')
-rw-r--r--youtube_dl/extractor/lynda.py86
1 files changed, 47 insertions, 39 deletions
diff --git a/youtube_dl/extractor/lynda.py b/youtube_dl/extractor/lynda.py
index 378117270..d4e1ae99d 100644
--- a/youtube_dl/extractor/lynda.py
+++ b/youtube_dl/extractor/lynda.py
@@ -7,12 +7,12 @@ from .common import InfoExtractor
from ..compat import (
compat_str,
compat_urllib_parse,
- compat_urllib_request,
)
from ..utils import (
ExtractorError,
clean_html,
int_or_none,
+ sanitized_Request,
)
@@ -25,7 +25,7 @@ class LyndaBaseIE(InfoExtractor):
self._login()
def _login(self):
- (username, password) = self._get_login_info()
+ username, password = self._get_login_info()
if username is None:
return
@@ -35,7 +35,7 @@ class LyndaBaseIE(InfoExtractor):
'remember': 'false',
'stayPut': 'false'
}
- request = compat_urllib_request.Request(
+ request = sanitized_Request(
self._LOGIN_URL, compat_urllib_parse.urlencode(login_form).encode('utf-8'))
login_page = self._download_webpage(
request, None, 'Logging in as %s' % username)
@@ -64,7 +64,7 @@ class LyndaBaseIE(InfoExtractor):
'remember': 'false',
'stayPut': 'false',
}
- request = compat_urllib_request.Request(
+ request = sanitized_Request(
self._LOGIN_URL, compat_urllib_parse.urlencode(confirm_form).encode('utf-8'))
login_page = self._download_webpage(
request, None,
@@ -82,6 +82,15 @@ class LyndaBaseIE(InfoExtractor):
expected=True)
raise ExtractorError('Unable to log in')
+ def _logout(self):
+ username, _ = self._get_login_info()
+ if username is None:
+ return
+
+ self._download_webpage(
+ 'http://www.lynda.com/ajax/logout.aspx', None,
+ 'Logging out', 'Unable to log out', fatal=False)
+
class LyndaIE(LyndaBaseIE):
IE_NAME = 'lynda'
@@ -108,50 +117,47 @@ class LyndaIE(LyndaBaseIE):
def _real_extract(self, url):
video_id = self._match_id(url)
- page = self._download_webpage(
+ video = self._download_json(
'http://www.lynda.com/ajax/player?videoId=%s&type=video' % video_id,
video_id, 'Downloading video JSON')
- video_json = json.loads(page)
- if 'Status' in video_json:
+ if 'Status' in video:
raise ExtractorError(
- 'lynda returned error: %s' % video_json['Message'], expected=True)
+ 'lynda returned error: %s' % video['Message'], expected=True)
- if video_json['HasAccess'] is False:
+ if video.get('HasAccess') is False:
self.raise_login_required('Video %s is only available for members' % video_id)
- video_id = compat_str(video_json['ID'])
- duration = video_json['DurationInSeconds']
- title = video_json['Title']
+ video_id = compat_str(video.get('ID') or video_id)
+ duration = int_or_none(video.get('DurationInSeconds'))
+ title = video['Title']
formats = []
- fmts = video_json.get('Formats')
+ fmts = video.get('Formats')
if fmts:
- formats.extend([
- {
- 'url': fmt['Url'],
- 'ext': fmt['Extension'],
- 'width': fmt['Width'],
- 'height': fmt['Height'],
- 'filesize': fmt['FileSize'],
- 'format_id': str(fmt['Resolution'])
- } for fmt in fmts])
-
- prioritized_streams = video_json.get('PrioritizedStreams')
+ formats.extend([{
+ 'url': f['Url'],
+ 'ext': f.get('Extension'),
+ 'width': int_or_none(f.get('Width')),
+ 'height': int_or_none(f.get('Height')),
+ 'filesize': int_or_none(f.get('FileSize')),
+ 'format_id': compat_str(f.get('Resolution')) if f.get('Resolution') else None,
+ } for f in fmts if f.get('Url')])
+
+ prioritized_streams = video.get('PrioritizedStreams')
if prioritized_streams:
- formats.extend([
- {
+ for prioritized_stream_id, prioritized_stream in prioritized_streams.items():
+ formats.extend([{
'url': video_url,
'width': int_or_none(format_id),
- 'format_id': format_id,
- } for format_id, video_url in prioritized_streams['0'].items()
- ])
+ 'format_id': '%s-%s' % (prioritized_stream_id, format_id),
+ } for format_id, video_url in prioritized_stream.items()])
self._check_formats(formats, video_id)
self._sort_formats(formats)
- subtitles = self.extract_subtitles(video_id, page)
+ subtitles = self.extract_subtitles(video_id)
return {
'id': video_id,
@@ -182,7 +188,7 @@ class LyndaIE(LyndaBaseIE):
if srt:
return srt
- def _get_subtitles(self, video_id, webpage):
+ def _get_subtitles(self, video_id):
url = 'http://www.lynda.com/ajax/player?videoId=%s&type=transcript' % video_id
subs = self._download_json(url, None, False)
if subs:
@@ -204,12 +210,13 @@ class LyndaCourseIE(LyndaBaseIE):
course_path = mobj.group('coursepath')
course_id = mobj.group('courseid')
- page = self._download_webpage(
+ course = self._download_json(
'http://www.lynda.com/ajax/player?courseId=%s&type=course' % course_id,
course_id, 'Downloading course JSON')
- course_json = json.loads(page)
- if 'Status' in course_json and course_json['Status'] == 'NotFound':
+ self._logout()
+
+ if course.get('Status') == 'NotFound':
raise ExtractorError(
'Course %s does not exist' % course_id, expected=True)
@@ -219,12 +226,13 @@ class LyndaCourseIE(LyndaBaseIE):
# Might want to extract videos right here from video['Formats'] as it seems 'Formats' is not provided
# by single video API anymore
- for chapter in course_json['Chapters']:
- for video in chapter['Videos']:
- if video['HasAccess'] is False:
+ for chapter in course['Chapters']:
+ for video in chapter.get('Videos', []):
+ if video.get('HasAccess') is False:
unaccessible_videos += 1
continue
- videos.append(video['ID'])
+ if video.get('ID'):
+ videos.append(video['ID'])
if unaccessible_videos > 0:
self._downloader.report_warning(
@@ -237,6 +245,6 @@ class LyndaCourseIE(LyndaBaseIE):
'Lynda')
for video_id in videos]
- course_title = course_json['Title']
+ course_title = course.get('Title')
return self.playlist_result(entries, course_id, course_title)