from .rtp import RTPIE
from .rts import RTSIE
from .rtve import RTVEALaCartaIE, RTVELiveIE, RTVEInfantilIE
-from .rtvnhnl import RtvnhNlIE
+from .rtvnh import RTVNHIE
from .ruhd import RUHDIE
from .rutube import (
RutubeIE,
--- /dev/null
+# coding: utf-8
+from __future__ import unicode_literals
+
+from .common import InfoExtractor
+
+
+class RTVNHIE(InfoExtractor):
+ _VALID_URL = r'https?://(?:www\.)?rtvnh\.nl/video/(?P<id>[0-9]+)'
+ _TEST = {
+ 'params': {
+ 'hls_prefer_native': True
+ },
+
+ 'url': 'http://www.rtvnh.nl/video/131946',
+ 'md5': '6e1d0ab079e2a00b6161442d3ceacfc1',
+ 'info_dict': {
+ 'id': '131946',
+ 'ext': 'mp4',
+ 'title': 'Grote zoektocht in zee bij Zandvoort naar vermiste vrouw',
+ 'thumbnail': 're:^https?:.*\.jpg$'
+ }
+ }
+
+ def _real_extract(self, url):
+ video_id = self._match_id(url)
+ meta = self._parse_json(self._download_webpage('http://www.rtvnh.nl/video/json?m=' + video_id, video_id), video_id)
+ formats = self._extract_smil_formats('http://www.rtvnh.nl/video/smil?m=' + video_id, video_id)
+
+ for item in meta['source']['fb']:
+ if item.get('type') == 'hls':
+ formats.extend(self._extract_m3u8_formats(item['file'], video_id, ext='mp4'))
+ elif item.get('type') == '':
+ formats.append({'url': item['file']})
+
+ return {
+ 'id': video_id,
+ 'title': meta['title'].strip(),
+ 'thumbnail': meta['image'],
+ 'formats': formats
+ }
+++ /dev/null
-# coding: utf-8
-from __future__ import unicode_literals
-
-from .common import InfoExtractor
-
-
-class RtvnhNlIE(InfoExtractor):
- _VALID_URL = r'https?://(?:www\.)?rtvnh\.nl/video/(?P<id>[0-9]+)'
- _TEST = {
- 'params': {
- 'hls_prefer_native': True
- },
-
- 'url': 'http://www.rtvnh.nl/video/131946',
- 'md5': '6e1d0ab079e2a00b6161442d3ceacfc1',
- 'info_dict': {
- 'id': '131946',
- 'ext': 'mp4',
- 'title': 'Grote zoektocht in zee bij Zandvoort naar vermiste vrouw',
- 'thumbnail': 're:^http:.*\.jpg$'
- }
- }
-
- def _real_extract(self, url):
- video_id = self._match_id(url)
- meta = self._parse_json(self._download_webpage('http://www.rtvnh.nl/video/json?m=' + video_id, video_id), video_id)
- formats = self._extract_smil_formats('http://www.rtvnh.nl/video/smil?m=' + video_id, video_id)
-
- for item in meta['source']['fb']:
- if item.get('type') == 'hls':
- formats.extend(self._extract_m3u8_formats(item['file'], video_id, ext='mp4'))
- elif item.get('type') == '':
- formats.append({'url': item['file']})
-
- return {
- 'id': video_id,
- 'title': meta['title'].strip(),
- 'thumbnail': meta['image'],
- 'formats': formats
- }