aboutsummaryrefslogtreecommitdiff
path: root/youtube_dl/extractor
diff options
context:
space:
mode:
Diffstat (limited to 'youtube_dl/extractor')
-rw-r--r--youtube_dl/extractor/xnxx.py46
1 files changed, 20 insertions, 26 deletions
diff --git a/youtube_dl/extractor/xnxx.py b/youtube_dl/extractor/xnxx.py
index 1177a4b14..7a73b2430 100644
--- a/youtube_dl/extractor/xnxx.py
+++ b/youtube_dl/extractor/xnxx.py
@@ -1,55 +1,49 @@
+# encoding: utf-8
+from __future__ import unicode_literals
+
import re
from .common import InfoExtractor
from ..utils import (
compat_urllib_parse,
-
- ExtractorError,
)
class XNXXIE(InfoExtractor):
- _VALID_URL = r'^(?:https?://)?(?:video|www)\.xnxx\.com/video([0-9]+)/(.*)'
- VIDEO_URL_RE = r'flv_url=(.*?)&'
- VIDEO_TITLE_RE = r'<title>(.*?)\s+-\s+XNXX.COM'
- VIDEO_THUMB_RE = r'url_bigthumb=(.*?)&amp;'
+ _VALID_URL = r'^https?://(?:video|www)\.xnxx\.com/video(?P<id>[0-9]+)/(.*)'
_TEST = {
- u'url': u'http://video.xnxx.com/video1135332/lida_naked_funny_actress_5_',
- u'file': u'1135332.flv',
- u'md5': u'0831677e2b4761795f68d417e0b7b445',
- u'info_dict': {
- u"title": u"lida \u00bb Naked Funny Actress (5)",
- u"age_limit": 18,
+ 'url': 'http://video.xnxx.com/video1135332/lida_naked_funny_actress_5_',
+ 'md5': '0831677e2b4761795f68d417e0b7b445',
+ 'info_dict': {
+ 'id': '1135332',
+ 'ext': 'flv',
+ 'title': 'lida ยป Naked Funny Actress (5)',
+ 'age_limit': 18,
}
}
def _real_extract(self, url):
mobj = re.match(self._VALID_URL, url)
- if mobj is None:
- raise ExtractorError(u'Invalid URL: %s' % url)
- video_id = mobj.group(1)
+ video_id = mobj.group('id')
# Get webpage content
webpage = self._download_webpage(url, video_id)
- video_url = self._search_regex(self.VIDEO_URL_RE,
- webpage, u'video URL')
+ video_url = self._search_regex(r'flv_url=(.*?)&amp;',
+ webpage, 'video URL')
video_url = compat_urllib_parse.unquote(video_url)
- video_title = self._html_search_regex(self.VIDEO_TITLE_RE,
- webpage, u'title')
+ video_title = self._html_search_regex(r'<title>(.*?)\s+-\s+XNXX.COM',
+ webpage, 'title')
- video_thumbnail = self._search_regex(self.VIDEO_THUMB_RE,
- webpage, u'thumbnail', fatal=False)
+ video_thumbnail = self._search_regex(r'url_bigthumb=(.*?)&amp;',
+ webpage, 'thumbnail', fatal=False)
- return [{
+ return {
'id': video_id,
'url': video_url,
- 'uploader': None,
- 'upload_date': None,
'title': video_title,
'ext': 'flv',
'thumbnail': video_thumbnail,
- 'description': None,
'age_limit': 18,
- }]
+ }