aboutsummaryrefslogtreecommitdiff
path: root/youtube_dl/extractor/channel9.py
diff options
context:
space:
mode:
Diffstat (limited to 'youtube_dl/extractor/channel9.py')
-rw-r--r--youtube_dl/extractor/channel9.py82
1 files changed, 43 insertions, 39 deletions
diff --git a/youtube_dl/extractor/channel9.py b/youtube_dl/extractor/channel9.py
index 4f000292b..3dfc24f5b 100644
--- a/youtube_dl/extractor/channel9.py
+++ b/youtube_dl/extractor/channel9.py
@@ -5,6 +5,7 @@ import re
from .common import InfoExtractor
from ..utils import ExtractorError
+
class Channel9IE(InfoExtractor):
'''
Common extractor for channel9.msdn.com.
@@ -27,11 +28,11 @@ class Channel9IE(InfoExtractor):
'title': 'Developer Kick-Off Session: Stuff We Love',
'description': 'md5:c08d72240b7c87fcecafe2692f80e35f',
'duration': 4576,
- 'thumbnail': 'http://media.ch9.ms/ch9/9d51/03902f2d-fc97-4d3c-b195-0bfe15a19d51/KOS002_220.jpg',
+ 'thumbnail': 'http://video.ch9.ms/ch9/9d51/03902f2d-fc97-4d3c-b195-0bfe15a19d51/KOS002_220.jpg',
'session_code': 'KOS002',
'session_day': 'Day 1',
'session_room': 'Arena 1A',
- 'session_speakers': [ 'Ed Blankenship', 'Andrew Coates', 'Brady Gaster', 'Patrick Klug', 'Mads Kristensen' ],
+ 'session_speakers': ['Ed Blankenship', 'Andrew Coates', 'Brady Gaster', 'Patrick Klug', 'Mads Kristensen'],
},
},
{
@@ -43,8 +44,8 @@ class Channel9IE(InfoExtractor):
'title': 'Self-service BI with Power BI - nuclear testing',
'description': 'md5:d1e6ecaafa7fb52a2cacdf9599829f5b',
'duration': 1540,
- 'thumbnail': 'http://media.ch9.ms/ch9/87e1/0300391f-a455-4c72-bec3-4422f19287e1/selfservicenuk_512.jpg',
- 'authors': [ 'Mike Wilmot' ],
+ 'thumbnail': 'http://video.ch9.ms/ch9/87e1/0300391f-a455-4c72-bec3-4422f19287e1/selfservicenuk_512.jpg',
+ 'authors': ['Mike Wilmot'],
},
}
]
@@ -83,7 +84,7 @@ class Channel9IE(InfoExtractor):
'format_id': x.group('quality'),
'format_note': x.group('note'),
'format': '%s (%s)' % (x.group('quality'), x.group('note')),
- 'filesize': self._restore_bytes(x.group('filesize')), # File size is approximate
+ 'filesize': self._restore_bytes(x.group('filesize')), # File size is approximate
'preference': self._known_formats.index(x.group('quality')),
'vcodec': 'none' if x.group('note') == 'Audio only' else None,
} for x in list(re.finditer(FORMAT_REGEX, html)) if x.group('quality') in self._known_formats]
@@ -94,7 +95,7 @@ class Channel9IE(InfoExtractor):
def _extract_title(self, html):
title = self._html_search_meta('title', html, 'title')
- if title is None:
+ if title is None:
title = self._og_search_title(html)
TITLE_SUFFIX = ' (Channel 9)'
if title is not None and title.endswith(TITLE_SUFFIX):
@@ -115,7 +116,7 @@ class Channel9IE(InfoExtractor):
return self._html_search_meta('description', html, 'description')
def _extract_duration(self, html):
- m = re.search(r'data-video_duration="(?P<hours>\d{2}):(?P<minutes>\d{2}):(?P<seconds>\d{2})"', html)
+ m = re.search(r'"length": *"(?P<hours>\d{2}):(?P<minutes>\d{2}):(?P<seconds>\d{2})"', html)
return ((int(m.group('hours')) * 60 * 60) + (int(m.group('minutes')) * 60) + int(m.group('seconds'))) if m else None
def _extract_slides(self, html):
@@ -167,7 +168,7 @@ class Channel9IE(InfoExtractor):
return re.findall(r'<a href="/Events/Speakers/[^"]+">([^<]+)</a>', html)
def _extract_content(self, html, content_path):
- # Look for downloadable content
+ # Look for downloadable content
formats = self._formats_from_html(html)
slides = self._extract_slides(html)
zip_ = self._extract_zip(html)
@@ -187,32 +188,33 @@ class Channel9IE(InfoExtractor):
view_count = self._extract_view_count(html)
comment_count = self._extract_comment_count(html)
- common = {'_type': 'video',
- 'id': content_path,
- 'description': description,
- 'thumbnail': thumbnail,
- 'duration': duration,
- 'avg_rating': avg_rating,
- 'rating_count': rating_count,
- 'view_count': view_count,
- 'comment_count': comment_count,
- }
+ common = {
+ '_type': 'video',
+ 'id': content_path,
+ 'description': description,
+ 'thumbnail': thumbnail,
+ 'duration': duration,
+ 'avg_rating': avg_rating,
+ 'rating_count': rating_count,
+ 'view_count': view_count,
+ 'comment_count': comment_count,
+ }
result = []
if slides is not None:
d = common.copy()
- d.update({ 'title': title + '-Slides', 'url': slides })
+ d.update({'title': title + '-Slides', 'url': slides})
result.append(d)
if zip_ is not None:
d = common.copy()
- d.update({ 'title': title + '-Zip', 'url': zip_ })
+ d.update({'title': title + '-Zip', 'url': zip_})
result.append(d)
if len(formats) > 0:
d = common.copy()
- d.update({ 'title': title, 'formats': formats })
+ d.update({'title': title, 'formats': formats})
result.append(d)
return result
@@ -234,16 +236,17 @@ class Channel9IE(InfoExtractor):
if contents is None:
return contents
- session_meta = {'session_code': self._extract_session_code(html),
- 'session_day': self._extract_session_day(html),
- 'session_room': self._extract_session_room(html),
- 'session_speakers': self._extract_session_speakers(html),
- }
+ session_meta = {
+ 'session_code': self._extract_session_code(html),
+ 'session_day': self._extract_session_day(html),
+ 'session_room': self._extract_session_room(html),
+ 'session_speakers': self._extract_session_speakers(html),
+ }
for content in contents:
content.update(session_meta)
- return contents
+ return self.playlist_result(contents)
def _extract_list(self, content_path):
rss = self._download_xml(self._RSS_URL % content_path, content_path, 'Downloading RSS')
@@ -258,16 +261,17 @@ class Channel9IE(InfoExtractor):
webpage = self._download_webpage(url, content_path, 'Downloading web page')
- page_type_m = re.search(r'<meta name="Search.PageType" content="(?P<pagetype>[^"]+)"/>', webpage)
- if page_type_m is None:
- raise ExtractorError('Search.PageType not found, don\'t know how to process this page', expected=True)
-
- page_type = page_type_m.group('pagetype')
- if page_type == 'List': # List page, may contain list of 'item'-like objects
+ page_type_m = re.search(r'<meta name="WT.entryid" content="(?P<pagetype>[^:]+)[^"]+"/>', webpage)
+ if page_type_m is not None:
+ page_type = page_type_m.group('pagetype')
+ if page_type == 'Entry': # Any 'item'-like page, may contain downloadable content
+ return self._extract_entry_item(webpage, content_path)
+ elif page_type == 'Session': # Event session page, may contain downloadable content
+ return self._extract_session(webpage, content_path)
+ elif page_type == 'Event':
+ return self._extract_list(content_path)
+ else:
+ raise ExtractorError('Unexpected WT.entryid %s' % page_type, expected=True)
+
+ else: # Assuming list
return self._extract_list(content_path)
- elif page_type == 'Entry.Item': # Any 'item'-like page, may contain downloadable content
- return self._extract_entry_item(webpage, content_path)
- elif page_type == 'Session': # Event session page, may contain downloadable content
- return self._extract_session(webpage, content_path)
- else:
- raise ExtractorError('Unexpected Search.PageType %s' % page_type, expected=True) \ No newline at end of file