1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
|
# encoding: utf-8
from __future__ import unicode_literals
from .common import InfoExtractor
from ..compat import compat_str
class DctpTvIE(InfoExtractor):
_VALID_URL = r'http://www.dctp.tv/(#/)?filme/(?P<id>.+?)/$'
_TEST = {
'url': 'http://www.dctp.tv/filme/videoinstallation-fuer-eine-kaufhausfassade/',
'info_dict': {
'id': '1324',
'display_id': 'videoinstallation-fuer-eine-kaufhausfassade',
'ext': 'flv',
'title': 'Videoinstallation für eine Kaufhausfassade'}
}
def _real_extract(self, url):
video_id = self._match_id(url)
base_url = 'http://dctp-ivms2-restapi.s3.amazonaws.com/'
version_json = self._download_json(
base_url + 'version.json',
video_id, note='Determining file version')
version = version_json['version_name']
info_json = self._download_json(
'{0}{1}/restapi/slugs/{2}.json'.format(base_url, version, video_id),
video_id, note='Fetching object ID')
object_id = compat_str(info_json['object_id'])
meta_json = self._download_json(
'{0}{1}/restapi/media/{2}.json'.format(base_url, version, object_id),
video_id, note='Downloading metadata')
uuid = meta_json['uuid']
title = meta_json['title']
wide = meta_json['is_wide']
if wide:
ratio = '16x9'
else:
ratio = '4x3'
play_path = 'mp4:{0}_dctp_0500_{1}.m4v'.format(uuid, ratio)
servers_json = self._download_json(
'http://www.dctp.tv/streaming_servers/',
video_id, note='Downloading server list')
url = servers_json[0]['endpoint']
return {
'id': object_id,
'title': title,
'format': 'rtmp',
'url': url,
'play_path': play_path,
'rtmp_real_time': True,
'ext': 'flv',
'display_id': video_id
}
|