1
0
Fork 0
mirror of https://github.com/ytdl-org/youtube-dl.git synced 2024-11-30 06:01:50 +00:00
youtube-dl/youtube_dl/extractor/iprima.py
2014-03-18 19:33:28 +07:00

107 lines
3.5 KiB
Python

# -*- coding: utf-8 -*-
from __future__ import unicode_literals
import re
from random import random
from math import floor
from .common import InfoExtractor
from ..utils import (
compat_urllib_request,
ExtractorError,
)
class IPrimaIE(InfoExtractor):
_VALID_URL = r'https?://play\.iprima\.cz/[^?#]+/(?P<id>[^?#]+)'
_TESTS = [{
'url': 'http://play.iprima.cz/particka/particka-92',
'info_dict': {
'id': '39152',
'ext': 'flv',
'title': 'Partička (92)',
'description': 'md5:3740fda51464da35a2d4d0670b8e4fd6',
'thumbnail': 'http://play.iprima.cz/sites/default/files/image_crops/image_620x349/3/491483_particka-92_image_620x349.jpg',
},
'params': {
'skip_download': True, # requires rtmpdump
},
}, {
'url': 'http://play.iprima.cz/particka/tchibo-particka-jarni-moda',
'info_dict': {
'id': '9718337',
'ext': 'flv',
'title': 'Tchibo Partička - Jarní móda',
'description': 'md5:589f8f59f414220621ff8882eb3ce7be',
'thumbnail': 're:^http:.*\.jpg$',
},
'params': {
'skip_download': True, # requires rtmpdump
},
'skip': 'Do not have permission to access this page',
}]
def _real_extract(self, url):
mobj = re.match(self._VALID_URL, url)
video_id = mobj.group('id')
webpage = self._download_webpage(url, video_id)
if re.search(r'Nemáte oprávnění přistupovat na tuto stránku\.\s*</div>', webpage):
raise ExtractorError(
'%s said: You do not have permission to access this page' % self.IE_NAME, expected=True)
player_url = (
'http://embed.livebox.cz/iprimaplay/player-embed-v2.js?__tok%s__=%s' %
(floor(random()*1073741824), floor(random()*1073741824))
)
req = compat_urllib_request.Request(player_url)
req.add_header('Referer', url)
playerpage = self._download_webpage(req, video_id)
base_url = ''.join(re.findall(r"embed\['stream'\] = '(.+?)'.+'(\?auth=)'.+'(.+?)';", playerpage)[1])
zoneGEO = self._html_search_regex(r'"zoneGEO":(.+?),', webpage, 'zoneGEO')
if zoneGEO != '0':
base_url = base_url.replace('token', 'token_' + zoneGEO)
formats = []
for format_id in ['lq', 'hq', 'hd']:
filename = self._html_search_regex(
r'"%s_id":(.+?),' % format_id, webpage, 'filename')
if filename == 'null':
continue
real_id = self._search_regex(
r'Prima-(?:[0-9]{10}|WEB)-([0-9]+)[-_]',
filename, 'real video id')
if format_id == 'lq':
quality = 0
elif format_id == 'hq':
quality = 1
elif format_id == 'hd':
quality = 2
filename = 'hq/' + filename
formats.append({
'format_id': format_id,
'url': base_url,
'quality': quality,
'play_path': 'mp4:' + filename.replace('"', '')[:-4],
'rtmp_live': True,
'ext': 'flv',
})
self._sort_formats(formats)
return {
'id': real_id,
'title': self._og_search_title(webpage),
'thumbnail': self._og_search_thumbnail(webpage),
'formats': formats,
'description': self._og_search_description(webpage),
}