mirror of
https://github.com/ytdl-org/youtube-dl.git
synced 2024-11-25 12:15:07 +01:00
69 lines
2.5 KiB
Python
69 lines
2.5 KiB
Python
# coding: utf-8
|
|
from __future__ import unicode_literals
|
|
|
|
import re
|
|
|
|
from .common import InfoExtractor
|
|
from ..utils import (
|
|
float_or_none,
|
|
ExtractorError,
|
|
)
|
|
|
|
|
|
class UplynkIE(InfoExtractor):
|
|
IE_NAME = 'uplynk'
|
|
_VALID_URL = r'https?://.*?\.uplynk\.com/(?P<path>ext/[0-9a-f]{32}/(?P<external_id>[^/?&]+)|(?P<id>[0-9a-f]{32}))\.(?:m3u8|json)(?:.*?\bpbs=(?P<session_id>[^&]+))?'
|
|
_TEST = {
|
|
'url': 'http://content.uplynk.com/e89eaf2ce9054aa89d92ddb2d817a52e.m3u8',
|
|
'info_dict': {
|
|
'id': 'e89eaf2ce9054aa89d92ddb2d817a52e',
|
|
'ext': 'mp4',
|
|
'title': '030816-kgo-530pm-solar-eclipse-vid_web.mp4',
|
|
'uploader_id': '4413701bf5a1488db55b767f8ae9d4fa',
|
|
},
|
|
'params': {
|
|
# m3u8 download
|
|
'skip_download': True,
|
|
},
|
|
}
|
|
|
|
def _extract_uplynk_info(self, uplynk_content_url):
|
|
path, external_id, video_id, session_id = re.match(UplynkIE._VALID_URL, uplynk_content_url).groups()
|
|
display_id = video_id or external_id
|
|
formats = self._extract_m3u8_formats('http://content.uplynk.com/%s.m3u8' % path, display_id, 'mp4')
|
|
if session_id:
|
|
for f in formats:
|
|
f['extra_param_to_segment_url'] = 'pbs=' + session_id
|
|
self._sort_formats(formats)
|
|
asset = self._download_json('http://content.uplynk.com/player/assetinfo/%s.json' % path, display_id)
|
|
if asset.get('error') == 1:
|
|
raise ExtractorError('% said: %s' % (self.IE_NAME, asset['msg']), expected=True)
|
|
|
|
return {
|
|
'id': asset['asset'],
|
|
'title': asset['desc'],
|
|
'thumbnail': asset.get('default_poster_url'),
|
|
'duration': float_or_none(asset.get('duration')),
|
|
'uploader_id': asset.get('owner'),
|
|
'formats': formats,
|
|
}
|
|
|
|
def _real_extract(self, url):
|
|
return self._extract_uplynk_info(url)
|
|
|
|
|
|
class UplynkPreplayIE(UplynkIE):
|
|
IE_NAME = 'uplynk:preplay'
|
|
_VALID_URL = r'https?://.*?\.uplynk\.com/preplay2?/(?P<path>ext/[0-9a-f]{32}/(?P<external_id>[^/?&]+)|(?P<id>[0-9a-f]{32}))\.json'
|
|
_TEST = None
|
|
|
|
def _real_extract(self, url):
|
|
path, external_id, video_id = re.match(self._VALID_URL, url).groups()
|
|
display_id = video_id or external_id
|
|
preplay = self._download_json(url, display_id)
|
|
content_url = 'http://content.uplynk.com/%s.m3u8' % path
|
|
session_id = preplay.get('sid')
|
|
if session_id:
|
|
content_url += '?pbs=' + session_id
|
|
return self._extract_uplynk_info(content_url)
|