aboutsummaryrefslogtreecommitdiff
path: root/youtube_dl/extractor
diff options
context:
space:
mode:
Diffstat (limited to 'youtube_dl/extractor')
-rw-r--r--youtube_dl/extractor/pornhub.py39
1 files changed, 23 insertions, 16 deletions
diff --git a/youtube_dl/extractor/pornhub.py b/youtube_dl/extractor/pornhub.py
index 72c351d56..7de585604 100644
--- a/youtube_dl/extractor/pornhub.py
+++ b/youtube_dl/extractor/pornhub.py
@@ -409,14 +409,14 @@ class PornHubUserIE(PornHubPlaylistBaseIE):
@classmethod
def suitable(cls, url):
return (False
- if PornHubUserVideosIE.suitable(url) or PornHubUserVideosUploadIE.suitable(url)
+ if PornHubPagedVideosIE.suitable(url) or PornHubUserVideosUploadIE.suitable(url)
else super(PornHubUserIE, cls).suitable(url))
def _real_extract(self, url):
mobj = re.match(self._VALID_URL, url)
user_id = mobj.group('id')
return self.url_result(
- '%s/videos' % mobj.group('url'), ie=PornHubUserVideosIE.ie_key(),
+ '%s/videos' % mobj.group('url'), ie=PornHubPagedVideosIE.ie_key(),
video_id=user_id)
@@ -426,10 +426,13 @@ class PornHubPagedPlaylistBaseIE(PornHubPlaylistBaseIE):
host = mobj.group('host')
user_id = mobj.group('id')
+ page = int_or_none(self._search_regex(
+ r'\bpage=(\d+)', url, 'page', default=None))
+
page_url = self._make_page_url(url)
entries = []
- for page_num in itertools.count(1):
+ for page_num in (page, ) if page is not None else itertools.count(1):
try:
webpage = self._download_webpage(
page_url, user_id, 'Downloading page %d' % page_num,
@@ -448,10 +451,17 @@ class PornHubPagedPlaylistBaseIE(PornHubPlaylistBaseIE):
return self.playlist_result(orderedSet(entries), user_id)
-class PornHubUserVideosIE(PornHubPagedPlaylistBaseIE):
- _VALID_URL = r'https?://(?:[^/]+\.)?(?P<host>pornhub\.(?:com|net))/(?:(?:user|channel)s|model|pornstar)/(?P<id>[^/]+)/videos'
+class PornHubPagedVideosIE(PornHubPagedPlaylistBaseIE):
+ _VALID_URL = r'''(?x)
+ https?://
+ (?:[^/]+\.)?(?P<host>pornhub\.(?:com|net))/
+ (?:
+ (?:(?:user|channel)s|model|pornstar)/(?P<id>[^/]+)/videos|
+ video/search
+ )
+ '''
_TESTS = [{
- 'url': 'https://www.pornhub.com/model/zoe_ph/videos/upload',
+ 'url': 'https://www.pornhub.com/model/zoe_ph/videos',
'only_matching': True,
}, {
'url': 'http://www.pornhub.com/users/rushandlia/videos',
@@ -463,6 +473,12 @@ class PornHubUserVideosIE(PornHubPagedPlaylistBaseIE):
},
'playlist_mincount': 149,
}, {
+ 'url': 'https://www.pornhub.com/pornstar/jenny-blighe/videos?page=3',
+ 'info_dict': {
+ 'id': 'jenny-blighe',
+ },
+ 'playlist_mincount': 40,
+ }, {
# default sorting as Top Rated Videos
'url': 'https://www.pornhub.com/channels/povd/videos',
'info_dict': {
@@ -485,12 +501,6 @@ class PornHubUserVideosIE(PornHubPagedPlaylistBaseIE):
'url': 'http://www.pornhub.com/users/zoe_ph/videos/public',
'only_matching': True,
}, {
- 'url': 'https://www.pornhub.com/model/jayndrea/videos/upload',
- 'only_matching': True,
- }, {
- 'url': 'https://www.pornhub.com/pornstar/jenny-blighe/videos/upload',
- 'only_matching': True,
- }, {
# Most Viewed Videos
'url': 'https://www.pornhub.com/pornstar/liz-vicious/videos?o=mv',
'only_matching': True,
@@ -507,9 +517,6 @@ class PornHubUserVideosIE(PornHubPagedPlaylistBaseIE):
'url': 'https://www.pornhub.com/pornstar/liz-vicious/videos?o=cm',
'only_matching': True,
}, {
- 'url': 'https://www.pornhub.com/pornstar/liz-vicious/videos/upload',
- 'only_matching': True,
- }, {
'url': 'https://www.pornhub.com/pornstar/liz-vicious/videos/paid',
'only_matching': True,
}, {
@@ -521,7 +528,7 @@ class PornHubUserVideosIE(PornHubPagedPlaylistBaseIE):
def suitable(cls, url):
return (False
if PornHubUserVideosUploadIE.suitable(url)
- else super(PornHubUserVideosIE, cls).suitable(url))
+ else super(PornHubPagedVideosIE, cls).suitable(url))
def _make_page_url(self, url):
return url