aboutsummaryrefslogtreecommitdiff
path: root/youtube_dl/extractor
diff options
context:
space:
mode:
Diffstat (limited to 'youtube_dl/extractor')
-rw-r--r--youtube_dl/extractor/__init__.py1
-rw-r--r--youtube_dl/extractor/pornoxo.py66
2 files changed, 67 insertions, 0 deletions
diff --git a/youtube_dl/extractor/__init__.py b/youtube_dl/extractor/__init__.py
index f1a5b2608..1bf5c51b4 100644
--- a/youtube_dl/extractor/__init__.py
+++ b/youtube_dl/extractor/__init__.py
@@ -267,6 +267,7 @@ from .podomatic import PodomaticIE
from .pornhd import PornHdIE
from .pornhub import PornHubIE
from .pornotube import PornotubeIE
+from .pornoxo import PornoXOIE
from .promptfile import PromptFileIE
from .prosiebensat1 import ProSiebenSat1IE
from .pyvideo import PyvideoIE
diff --git a/youtube_dl/extractor/pornoxo.py b/youtube_dl/extractor/pornoxo.py
new file mode 100644
index 000000000..0d9656d39
--- /dev/null
+++ b/youtube_dl/extractor/pornoxo.py
@@ -0,0 +1,66 @@
+from __future__ import unicode_literals
+
+import re
+
+from .common import InfoExtractor
+from ..utils import (
+ parse_duration,
+ str_to_int,
+)
+
+class PornoXOIE(InfoExtractor):
+ _VALID_URL = r'https?://(?:www\.)?pornoxo\.com/videos/(?P<id>\d+)/(?P<display_id>[^/]+)\.html'
+ _TEST = {
+ 'url': 'http://www.pornoxo.com/videos/7564/striptease-from-sexy-secretary.html',
+ 'md5': '582f28ecbaa9e6e24cb90f50f524ce87',
+ 'info_dict': {
+ 'id': '7564',
+ 'ext': 'flv',
+ 'title': 'Striptease From Sexy Secretary!',
+ 'description': 'Striptease From Sexy Secretary!',
+ 'categories': list, # NSFW
+ 'thumbnail': 're:https?://.*\.jpg$',
+ 'age_limit': 18,
+ }
+ }
+
+ def _real_extract(self, url):
+ mobj = re.match(self._VALID_URL, url)
+ video_id = mobj.group('id')
+
+ webpage = self._download_webpage(url, video_id)
+
+ video_url = self._html_search_regex(
+ r'\'file\'\s*:\s*"([^"]+)"', webpage, 'video_url')
+
+ title = self._html_search_regex(
+ r'<title>([^<]+)\s*-\s*PornoXO', webpage, 'title')
+
+ description = self._html_search_regex(
+ r'<meta name="description" content="([^"]+)\s*featuring',
+ webpage, 'description', fatal=False)
+
+ thumbnail = self._html_search_regex(
+ r'\'image\'\s*:\s*"([^"]+)"', webpage, 'thumbnail', fatal=False)
+
+ view_count = str_to_int(self._html_search_regex(
+ r'Views:\s*(\d+)', webpage, 'view count', fatal=False))
+
+ categories_str = self._html_search_regex(
+ r'<meta name="description" content=".*featuring\s*([^"]+)"',
+ webpage, 'categories', fatal=False)
+ categories = (
+ None if categories_str is None
+ else categories_str.split(','))
+
+
+ return {
+ 'id': video_id,
+ 'url': video_url,
+ 'title': title,
+ 'description': description,
+ 'thumbnail': thumbnail,
+ 'categories': categories,
+ 'view_count': view_count,
+ 'age_limit': 18,
+ }