aboutsummaryrefslogtreecommitdiff
path: root/youtube_dl
diff options
context:
space:
mode:
authorRemita Amine <remitamine@gmail.com>2017-03-19 00:45:04 +0100
committerRemita Amine <remitamine@gmail.com>2017-03-19 00:45:38 +0100
commit772b5ff57f702dd76986d1db17068da2116a2800 (patch)
tree9b560b120c409222cce5f960fc09058884c3f179 /youtube_dl
parentf68ef1e2abd876ffca65544fd3e42756f9c33be3 (diff)
[toongoggles] Add new extractor(closes #12171)
Diffstat (limited to 'youtube_dl')
-rw-r--r--youtube_dl/extractor/extractors.py1
-rw-r--r--youtube_dl/extractor/toongoggles.py81
2 files changed, 82 insertions, 0 deletions
diff --git a/youtube_dl/extractor/extractors.py b/youtube_dl/extractor/extractors.py
index 6b4742ed8..97d68d9ca 100644
--- a/youtube_dl/extractor/extractors.py
+++ b/youtube_dl/extractor/extractors.py
@@ -986,6 +986,7 @@ from .tnaflix import (
)
from .toggle import ToggleIE
from .tonline import TOnlineIE
+from .toongoggles import ToonGogglesIE
from .toutv import TouTvIE
from .toypics import ToypicsUserIE, ToypicsIE
from .traileraddict import TrailerAddictIE
diff --git a/youtube_dl/extractor/toongoggles.py b/youtube_dl/extractor/toongoggles.py
new file mode 100644
index 000000000..b5ba1c01d
--- /dev/null
+++ b/youtube_dl/extractor/toongoggles.py
@@ -0,0 +1,81 @@
+# coding: utf-8
+from __future__ import unicode_literals
+
+import re
+
+from .common import InfoExtractor
+from ..utils import (
+ int_or_none,
+ parse_duration,
+)
+
+
+class ToonGogglesIE(InfoExtractor):
+ _VALID_URL = r'https?://(?:www\.)?toongoggles\.com/shows/(?P<show_id>\d+)(?:/[^/]+/episodes/(?P<episode_id>\d+))?'
+ _TESTS = [{
+ 'url': 'http://www.toongoggles.com/shows/217143/bernard-season-2/episodes/217147/football',
+ 'md5': '18289fc2b951eff6b953a9d8f01e6831',
+ 'info_dict': {
+ 'id': '217147',
+ 'ext': 'mp4',
+ 'title': 'Football',
+ 'uploader_id': '1',
+ 'description': 'Bernard decides to play football in order to be better than Lloyd and tries to beat him no matter how, he even cheats.',
+ 'upload_date': '20160718',
+ 'timestamp': 1468879330,
+ }
+ }, {
+ 'url': 'http://www.toongoggles.com/shows/227759/om-nom-stories-around-the-world',
+ 'info_dict': {
+ 'id': '227759',
+ 'title': 'Om Nom Stories Around The World',
+ },
+ 'playlist_mincount': 11,
+ }]
+
+ def _call_api(self, action, page_id, query):
+ query.update({
+ 'for_ng': 1,
+ 'for_web': 1,
+ 'show_meta': 1,
+ 'version': 7.0,
+ })
+ return self._download_json('http://api.toongoggles.com/' + action, page_id, query=query)
+
+ def _parse_episode_data(self, episode_data):
+ title = episode_data['episode_name']
+
+ return {
+ '_type': 'url_transparent',
+ 'id': episode_data['episode_id'],
+ 'title': title,
+ 'url': 'kaltura:513551:' + episode_data['entry_id'],
+ 'thumbnail': episode_data.get('thumbnail_url'),
+ 'description': episode_data.get('description'),
+ 'duration': parse_duration(episode_data.get('hms')),
+ 'series': episode_data.get('show_name'),
+ 'season_number': int_or_none(episode_data.get('season_num')),
+ 'episode_id': episode_data.get('episode_id'),
+ 'episode': title,
+ 'episode_number': int_or_none(episode_data.get('episode_num')),
+ 'categories': episode_data.get('categories'),
+ 'ie_key': 'Kaltura',
+ }
+
+ def _real_extract(self, url):
+ show_id, episode_id = re.match(self._VALID_URL, url).groups()
+ if episode_id:
+ episode_data = self._call_api('search', episode_id, {
+ 'filter': 'episode',
+ 'id': episode_id,
+ })['objects'][0]
+ return self._parse_episode_data(episode_data)
+ else:
+ show_data = self._call_api('getepisodesbyshow', show_id, {
+ 'max': 1000000000,
+ 'showid': show_id,
+ })
+ entries = []
+ for episode_data in show_data.get('objects', []):
+ entries.append(self._parse_episode_data(episode_data))
+ return self.playlist_result(entries, show_id, show_data.get('show_name'))