0
0
Fork 0
mirror of https://github.com/yt-dlp/yt-dlp.git synced 2024-11-18 01:35:12 +00:00
yt-dlp/yt_dlp/extractor/channel9.py

261 lines
10 KiB
Python
Raw Normal View History

2014-01-28 19:00:56 +00:00
from __future__ import unicode_literals
import re
from .common import InfoExtractor
from ..utils import (
clean_html,
int_or_none,
parse_iso8601,
2017-03-23 16:47:43 +00:00
qualities,
unescapeHTML,
)
2014-11-23 19:41:03 +00:00
class Channel9IE(InfoExtractor):
2014-01-28 19:00:56 +00:00
IE_DESC = 'Channel 9'
IE_NAME = 'channel9'
_VALID_URL = r'https?://(?:www\.)?(?:channel9\.msdn\.com|s\.ch9\.ms)/(?P<contentpath>.+?)(?P<rss>/RSS)?/?(?:[?#&]|$)'
_TESTS = [{
'url': 'http://channel9.msdn.com/Events/TechEd/Australia/2013/KOS002',
'md5': '32083d4eaf1946db6d454313f44510ca',
'info_dict': {
'id': '6c413323-383a-49dc-88f9-a22800cab024',
'ext': 'wmv',
'title': 'Developer Kick-Off Session: Stuff We Love',
'description': 'md5:b80bf9355a503c193aff7ec6cd5a7731',
'duration': 4576,
'thumbnail': r're:https?://.*\.jpg',
'timestamp': 1377717420,
'upload_date': '20130828',
'session_code': 'KOS002',
'session_room': 'Arena 1A',
'session_speakers': 'count:5',
},
}, {
'url': 'http://channel9.msdn.com/posts/Self-service-BI-with-Power-BI-nuclear-testing',
'md5': 'dcf983ee6acd2088e7188c3cf79b46bc',
'info_dict': {
'id': 'fe8e435f-bb93-4e01-8e97-a28c01887024',
'ext': 'wmv',
'title': 'Self-service BI with Power BI - nuclear testing',
'description': 'md5:2d17fec927fc91e9e17783b3ecc88f54',
'duration': 1540,
'thumbnail': r're:https?://.*\.jpg',
'timestamp': 1386381991,
'upload_date': '20131207',
'authors': ['Mike Wilmot'],
},
}, {
# low quality mp4 is best
'url': 'https://channel9.msdn.com/Events/CPP/CppCon-2015/Ranges-for-the-Standard-Library',
'info_dict': {
'id': '33ad69d2-6a4e-4172-83a1-a523013dec76',
'ext': 'mp4',
'title': 'Ranges for the Standard Library',
'description': 'md5:9895e0a9fd80822d2f01c454b8f4a372',
'duration': 5646,
'thumbnail': r're:https?://.*\.jpg',
'upload_date': '20150930',
'timestamp': 1443640735,
},
'params': {
'skip_download': True,
},
}, {
'url': 'https://channel9.msdn.com/Events/DEVintersection/DEVintersection-2016/RSS',
'info_dict': {
'id': 'Events/DEVintersection/DEVintersection-2016',
'title': 'DEVintersection 2016 Orlando Sessions',
},
'playlist_mincount': 14,
}, {
'url': 'https://channel9.msdn.com/Niners/Splendid22/Queue/76acff796e8f411184b008028e0d492b/RSS',
'only_matching': True,
}, {
'url': 'https://channel9.msdn.com/Events/Speakers/scott-hanselman/RSS?UrlSafeName=scott-hanselman',
'only_matching': True,
}]
_RSS_URL = 'http://channel9.msdn.com/%s/RSS'
@staticmethod
def _extract_urls(webpage):
return re.findall(
r'<iframe[^>]+src=["\'](https?://channel9\.msdn\.com/(?:[^/]+/)+)player\b',
webpage)
def _extract_list(self, video_id, rss_url=None):
if not rss_url:
rss_url = self._RSS_URL % video_id
rss = self._download_xml(rss_url, video_id, 'Downloading RSS')
2013-12-12 08:19:23 +00:00
entries = [self.url_result(session_url.text, 'Channel9')
for session_url in rss.findall('./channel/item/link')]
title_text = rss.find('./channel/title').text
return self.playlist_result(entries, video_id, title_text)
def _real_extract(self, url):
content_path, rss = re.match(self._VALID_URL, url).groups()
if rss:
return self._extract_list(content_path, url)
webpage = self._download_webpage(
url, content_path, 'Downloading web page')
episode_data = self._search_regex(
r"data-episode='([^']+)'", webpage, 'episode data', default=None)
if episode_data:
episode_data = self._parse_json(unescapeHTML(
episode_data), content_path)
content_id = episode_data['contentId']
is_session = '/Sessions(' in episode_data['api']
content_url = 'https://channel9.msdn.com/odata' + episode_data['api'] + '?$select=Captions,CommentCount,MediaLengthInSeconds,PublishedDate,Rating,RatingCount,Title,VideoMP4High,VideoMP4Low,VideoMP4Medium,VideoPlayerPreviewImage,VideoWMV,VideoWMVHQ,Views,'
if is_session:
content_url += 'Code,Description,Room,Slides,Speakers,ZipFile&$expand=Speakers'
else:
content_url += 'Authors,Body&$expand=Authors'
content_data = self._download_json(content_url, content_id)
title = content_data['Title']
2017-03-23 16:47:43 +00:00
QUALITIES = (
'mp3',
'wmv', 'mp4',
'wmv-low', 'mp4-low',
'wmv-mid', 'mp4-mid',
'wmv-high', 'mp4-high',
)
quality_key = qualities(QUALITIES)
def quality(quality_id, format_url):
return (len(QUALITIES) if '_Source.' in format_url
else quality_key(quality_id))
formats = []
2017-03-23 16:47:43 +00:00
urls = set()
SITE_QUALITIES = {
'MP3': 'mp3',
'MP4': 'mp4',
'Low Quality WMV': 'wmv-low',
'Low Quality MP4': 'mp4-low',
'Mid Quality WMV': 'wmv-mid',
'Mid Quality MP4': 'mp4-mid',
'High Quality WMV': 'wmv-high',
'High Quality MP4': 'mp4-high',
}
formats_select = self._search_regex(
r'(?s)<select[^>]+name=["\']format[^>]+>(.+?)</select', webpage,
'formats select', default=None)
if formats_select:
for mobj in re.finditer(
r'<option\b[^>]+\bvalue=(["\'])(?P<url>(?:(?!\1).)+)\1[^>]*>\s*(?P<format>[^<]+?)\s*<',
formats_select):
format_url = mobj.group('url')
if format_url in urls:
continue
urls.add(format_url)
format_id = mobj.group('format')
quality_id = SITE_QUALITIES.get(format_id, format_id)
formats.append({
'url': format_url,
'format_id': quality_id,
'quality': quality(quality_id, format_url),
'vcodec': 'none' if quality_id == 'mp3' else None,
})
API_QUALITIES = {
'VideoMP4Low': 'mp4-low',
'VideoWMV': 'wmv-mid',
'VideoMP4Medium': 'mp4-mid',
'VideoMP4High': 'mp4-high',
'VideoWMVHQ': 'wmv-hq',
}
for format_id, q in API_QUALITIES.items():
q_url = content_data.get(format_id)
if not q_url or q_url in urls:
continue
2017-03-23 16:47:43 +00:00
urls.add(q_url)
formats.append({
'url': q_url,
2017-03-23 16:47:43 +00:00
'format_id': q,
'quality': quality(q, q_url),
})
2017-03-23 16:47:43 +00:00
slides = content_data.get('Slides')
zip_file = content_data.get('ZipFile')
if not formats and not slides and not zip_file:
self.raise_no_formats(
'None of recording, slides or zip are available for %s' % content_path)
self._sort_formats(formats)
subtitles = {}
for caption in content_data.get('Captions', []):
caption_url = caption.get('Url')
if not caption_url:
continue
subtitles.setdefault(caption.get('Language', 'en'), []).append({
'url': caption_url,
'ext': 'vtt',
})
common = {
'id': content_id,
'title': title,
'description': clean_html(content_data.get('Description') or content_data.get('Body')),
'thumbnail': content_data.get('VideoPlayerPreviewImage'),
'duration': int_or_none(content_data.get('MediaLengthInSeconds')),
'timestamp': parse_iso8601(content_data.get('PublishedDate')),
'avg_rating': int_or_none(content_data.get('Rating')),
'rating_count': int_or_none(content_data.get('RatingCount')),
'view_count': int_or_none(content_data.get('Views')),
'comment_count': int_or_none(content_data.get('CommentCount')),
'subtitles': subtitles,
}
if is_session:
speakers = []
for s in content_data.get('Speakers', []):
speaker_name = s.get('FullName')
if not speaker_name:
continue
speakers.append(speaker_name)
common.update({
'session_code': content_data.get('Code'),
'session_room': content_data.get('Room'),
'session_speakers': speakers,
})
2014-11-04 22:20:34 +00:00
else:
authors = []
for a in content_data.get('Authors', []):
author_name = a.get('DisplayName')
if not author_name:
continue
authors.append(author_name)
common['authors'] = authors
contents = []
if slides:
d = common.copy()
d.update({'title': title + '-Slides', 'url': slides})
contents.append(d)
if zip_file:
d = common.copy()
d.update({'title': title + '-Zip', 'url': zip_file})
contents.append(d)
if formats:
d = common.copy()
d.update({'title': title, 'formats': formats})
contents.append(d)
return self.playlist_result(contents)
else:
return self._extract_list(content_path)