aboutsummaryrefslogtreecommitdiff
path: root/youtube_dl/extractor/sockshare.py
diff options
context:
space:
mode:
Diffstat (limited to 'youtube_dl/extractor/sockshare.py')
-rw-r--r--youtube_dl/extractor/sockshare.py17
1 files changed, 9 insertions, 8 deletions
diff --git a/youtube_dl/extractor/sockshare.py b/youtube_dl/extractor/sockshare.py
index cbf2d7abe..75b634bc6 100644
--- a/youtube_dl/extractor/sockshare.py
+++ b/youtube_dl/extractor/sockshare.py
@@ -5,7 +5,6 @@ from ..utils import (
ExtractorError,
compat_urllib_parse,
compat_urllib_request,
- determine_ext,
)
import re
@@ -34,7 +33,7 @@ class SockshareIE(InfoExtractor):
webpage = self._download_webpage(url, video_id)
if re.search(self._FILE_DELETED_REGEX, webpage) is not None:
- raise ExtractorError(u'Video %s does not exist' % video_id,
+ raise ExtractorError('Video %s does not exist' % video_id,
expected=True)
confirm_hash = self._html_search_regex(r'''(?x)<input\s+
@@ -54,19 +53,21 @@ class SockshareIE(InfoExtractor):
req.add_header('Host', 'www.sockshare.com')
req.add_header('Content-type', 'application/x-www-form-urlencoded')
- webpage = self._download_webpage(req, video_id, 'Downloading video page')
+ webpage = self._download_webpage(
+ req, video_id, 'Downloading video page')
- video_url = self._html_search_regex(r'<a href="([^"]*)".+class="download_file_link"', webpage, 'file url')
+ video_url = self._html_search_regex(
+ r'<a href="([^"]*)".+class="download_file_link"',
+ webpage, 'file url')
video_url = "http://www.sockshare.com" + video_url
title = self._html_search_regex(r'<h1>(.+)<strong>', webpage, 'title')
- thumbnail = self._html_search_regex(r'<img\ssrc="([^"]*)".+name="bg"',
- webpage, 'thumbnail')
- ext = determine_ext(title)
+ thumbnail = self._html_search_regex(
+ r'<img\s+src="([^"]*)".+?name="bg"',
+ webpage, 'thumbnail')
formats = [{
'format_id': 'sd',
'url': video_url,
- 'ext': ext,
}]
return {