mirror of
https://github.com/ytdl-org/youtube-dl
synced 2024-11-25 06:56:50 +01:00
156 lines
6.3 KiB
Python
156 lines
6.3 KiB
Python
import json
|
|
import os
|
|
import re
|
|
import xml.etree.ElementTree
|
|
|
|
from .common import InfoExtractor
|
|
from ..utils import (
|
|
ExtractorError,
|
|
formatSeconds,
|
|
)
|
|
|
|
|
|
class JustinTVIE(InfoExtractor):
|
|
"""Information extractor for justin.tv and twitch.tv"""
|
|
# TODO: One broadcast may be split into multiple videos. The key
|
|
# 'broadcast_id' is the same for all parts, and 'broadcast_part'
|
|
# starts at 1 and increases. Can we treat all parts as one video?
|
|
|
|
_VALID_URL = r"""(?x)^(?:http://)?(?:www\.)?(?:twitch|justin)\.tv/
|
|
(?:
|
|
(?P<channelid>[^/]+)|
|
|
(?:(?:[^/]+)/b/(?P<videoid>[^/]+))|
|
|
(?:(?:[^/]+)/c/(?P<chapterid>[^/]+))
|
|
)
|
|
/?(?:\#.*)?$
|
|
"""
|
|
_JUSTIN_PAGE_LIMIT = 100
|
|
IE_NAME = u'justin.tv'
|
|
_TEST = {
|
|
u'url': u'http://www.twitch.tv/thegamedevhub/b/296128360',
|
|
u'file': u'296128360.flv',
|
|
u'md5': u'ecaa8a790c22a40770901460af191c9a',
|
|
u'info_dict': {
|
|
u"upload_date": u"20110927",
|
|
u"uploader_id": 25114803,
|
|
u"uploader": u"thegamedevhub",
|
|
u"title": u"Beginner Series - Scripting With Python Pt.1"
|
|
}
|
|
}
|
|
|
|
def report_download_page(self, channel, offset):
|
|
"""Report attempt to download a single page of videos."""
|
|
self.to_screen(u'%s: Downloading video information from %d to %d' %
|
|
(channel, offset, offset + self._JUSTIN_PAGE_LIMIT))
|
|
|
|
# Return count of items, list of *valid* items
|
|
def _parse_page(self, url, video_id):
|
|
info_json = self._download_webpage(url, video_id,
|
|
u'Downloading video info JSON',
|
|
u'unable to download video info JSON')
|
|
|
|
response = json.loads(info_json)
|
|
if type(response) != list:
|
|
error_text = response.get('error', 'unknown error')
|
|
raise ExtractorError(u'Justin.tv API: %s' % error_text)
|
|
info = []
|
|
for clip in response:
|
|
video_url = clip['video_file_url']
|
|
if video_url:
|
|
video_extension = os.path.splitext(video_url)[1][1:]
|
|
video_date = re.sub('-', '', clip['start_time'][:10])
|
|
video_uploader_id = clip.get('user_id', clip.get('channel_id'))
|
|
video_id = clip['id']
|
|
video_title = clip.get('title', video_id)
|
|
info.append({
|
|
'id': video_id,
|
|
'url': video_url,
|
|
'title': video_title,
|
|
'uploader': clip.get('channel_name', video_uploader_id),
|
|
'uploader_id': video_uploader_id,
|
|
'upload_date': video_date,
|
|
'ext': video_extension,
|
|
})
|
|
return (len(response), info)
|
|
|
|
def _real_extract(self, url):
|
|
mobj = re.match(self._VALID_URL, url)
|
|
if mobj is None:
|
|
raise ExtractorError(u'invalid URL: %s' % url)
|
|
|
|
api_base = 'http://api.justin.tv'
|
|
paged = False
|
|
if mobj.group('channelid'):
|
|
paged = True
|
|
video_id = mobj.group('channelid')
|
|
api = api_base + '/channel/archives/%s.json' % video_id
|
|
elif mobj.group('chapterid'):
|
|
chapter_id = mobj.group('chapterid')
|
|
|
|
webpage = self._download_webpage(url, chapter_id)
|
|
m = re.search(r'PP\.archive_id = "([0-9]+)";', webpage)
|
|
if not m:
|
|
raise ExtractorError(u'Cannot find archive of a chapter')
|
|
archive_id = m.group(1)
|
|
|
|
api = api_base + '/broadcast/by_chapter/%s.xml' % chapter_id
|
|
chapter_info_xml = self._download_webpage(api, chapter_id,
|
|
note=u'Downloading chapter information',
|
|
errnote=u'Chapter information download failed')
|
|
doc = xml.etree.ElementTree.fromstring(chapter_info_xml)
|
|
for a in doc.findall('.//archive'):
|
|
if archive_id == a.find('./id').text:
|
|
break
|
|
else:
|
|
raise ExtractorError(u'Could not find chapter in chapter information')
|
|
|
|
video_url = a.find('./video_file_url').text
|
|
video_ext = video_url.rpartition('.')[2] or u'flv'
|
|
|
|
chapter_api_url = u'https://api.twitch.tv/kraken/videos/c' + chapter_id
|
|
chapter_info_json = self._download_webpage(chapter_api_url, u'c' + chapter_id,
|
|
note='Downloading chapter metadata',
|
|
errnote='Download of chapter metadata failed')
|
|
chapter_info = json.loads(chapter_info_json)
|
|
|
|
bracket_start = int(doc.find('.//bracket_start').text)
|
|
bracket_end = int(doc.find('.//bracket_end').text)
|
|
|
|
# TODO determine start (and probably fix up file)
|
|
# youtube-dl -v http://www.twitch.tv/firmbelief/c/1757457
|
|
#video_url += u'?start=' + TODO:start_timestamp
|
|
# bracket_start is 13290, but we want 51670615
|
|
self._downloader.report_warning(u'Chapter detected, but we can just download the whole file. '
|
|
u'Chapter starts at %s and ends at %s' % (formatSeconds(bracket_start), formatSeconds(bracket_end)))
|
|
|
|
info = {
|
|
'id': u'c' + chapter_id,
|
|
'url': video_url,
|
|
'ext': video_ext,
|
|
'title': chapter_info['title'],
|
|
'thumbnail': chapter_info['preview'],
|
|
'description': chapter_info['description'],
|
|
'uploader': chapter_info['channel']['display_name'],
|
|
'uploader_id': chapter_info['channel']['name'],
|
|
}
|
|
return [info]
|
|
else:
|
|
video_id = mobj.group('videoid')
|
|
api = api_base + '/broadcast/by_archive/%s.json' % video_id
|
|
|
|
self.report_extraction(video_id)
|
|
|
|
info = []
|
|
offset = 0
|
|
limit = self._JUSTIN_PAGE_LIMIT
|
|
while True:
|
|
if paged:
|
|
self.report_download_page(video_id, offset)
|
|
page_url = api + ('?offset=%d&limit=%d' % (offset, limit))
|
|
page_count, page_info = self._parse_page(page_url, video_id)
|
|
info.extend(page_info)
|
|
if not paged or page_count != limit:
|
|
break
|
|
offset += limit
|
|
return info
|