1
0
Fork 0
mirror of https://github.com/blackjack4494/yt-dlc.git synced 2024-12-22 15:57:23 +00:00

[channel9] Cleanup

This commit is contained in:
dst 2013-12-12 15:19:23 +07:00
parent df53747436
commit 4d2ebb6bd7

View file

@ -3,10 +3,7 @@
import re
from .common import InfoExtractor
from ..utils import (
format_bytes,
ExtractorError,
)
from ..utils import ExtractorError
class Channel9IE(InfoExtractor):
'''
@ -51,7 +48,6 @@ class Channel9IE(InfoExtractor):
]
_RSS_URL = 'http://channel9.msdn.com/%s/RSS'
_EXTRACT_ENTRY_ITEMS_FROM_RSS = False
# Sorted by quality
_known_formats = ['MP3', 'MP4', 'Mid Quality WMV', 'Mid Quality MP4', 'High Quality WMV', 'High Quality MP4']
@ -90,37 +86,6 @@ class Channel9IE(InfoExtractor):
formats.sort(key=lambda fmt: self._known_formats.index(fmt['format_id']))
return formats
def _formats_from_rss_item(self, item):
def process_formats(elem):
formats = []
for media_content in elem.findall('./{http://search.yahoo.com/mrss/}content'):
url = media_content.attrib['url']
# Ignore unrelated media
if url.endswith('.ism/manifest'):
continue
format_note = media_content.attrib['type']
filesize = int(media_content.attrib['fileSize'])
formats.append({'url': url,
'format_note': format_note,
'format': '%s %s' % (format_note, format_bytes(filesize)),
'filesize': filesize,
})
return formats
formats = []
for media_group in item.findall('./{http://search.yahoo.com/mrss/}group'):
formats.extend(process_formats(media_group))
# Sometimes there are no media:groups in item, but there is media:content
# right in item (usually when there is the only media source)
formats.extend(process_formats(item))
# Sort by file size
formats.sort(key=lambda fmt: fmt['filesize'])
return formats
def _extract_title(self, html):
title = self._html_search_meta(u'title', html, u'title')
if title is None:
@ -274,61 +239,12 @@ class Channel9IE(InfoExtractor):
return contents
def _extract_content_rss(self, rss):
'''
Extracts links to entry items right out of RSS feed.
This approach is faster than extracting from web pages
one by one, but suffers from some problems.
Pros:
- no need to download additional pages
- provides more media links
- accurate file size
Cons:
- fewer meta data provided
- links to media files have no appropriate data that may be used as format_id
- RSS does not contain links to presentation materials (slides, zip)
'''
entries = []
for item in rss.findall('./channel/item'):
url = item.find('./link').text
video_id = url.split('/')[-1]
formats = self._formats_from_rss_item(item)
if len(formats) == 0:
self._downloader.report_warning(u'The recording for session %s is not yet available' % video_id)
continue
title = item.find('./title').text
description = item.find('./description').text
thumbnail = item.find('./{http://search.yahoo.com/mrss/}thumbnail').text
duration_e = item.find('./{http://www.itunes.com/dtds/podcast-1.0.dtd}duration')
duration = duration_e.text if duration_e is not None else 0
speakers_e = item.find('./{http://purl.org/dc/elements/1.1/}creator')
speakers = speakers_e.text.split(', ') if speakers_e is not None and speakers_e.text else []
entries.append({'_type': 'video',
'id': video_id,
'formats': formats,
'title': title,
'description': description,
'thumbnail': thumbnail,
'duration': duration,
'session_speakers': speakers,
})
return entries
def _extract_list(self, content_path):
rss = self._download_xml(self._RSS_URL % content_path, content_path, u'Downloading RSS')
if self._EXTRACT_ENTRY_ITEMS_FROM_RSS:
return self._extract_content_rss(rss)
else:
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, content_path, title_text)
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, content_path, title_text)
def _real_extract(self, url):
mobj = re.match(self._VALID_URL, url)