mirror of
https://github.com/ytdl-org/youtube-dl.git
synced 2024-01-07 17:16:08 +00:00
[vk.com] Updated the extract_videos_from_page function with a much simpler 1-liner.
This commit is contained in:
parent
53d1cd1f77
commit
e1e8b6897b
|
@ -11,7 +11,8 @@ from ..utils import (
|
||||||
compat_urllib_parse,
|
compat_urllib_parse,
|
||||||
compat_str,
|
compat_str,
|
||||||
unescapeHTML,
|
unescapeHTML,
|
||||||
unified_strdate)
|
unified_strdate,
|
||||||
|
orderedSet)
|
||||||
|
|
||||||
|
|
||||||
class VKIE(InfoExtractor):
|
class VKIE(InfoExtractor):
|
||||||
|
@ -216,13 +217,6 @@ class VKUserVideosIE(InfoExtractor):
|
||||||
'playlist_mincount': 4,
|
'playlist_mincount': 4,
|
||||||
}
|
}
|
||||||
|
|
||||||
def extract_videos_from_page(self, page):
|
|
||||||
ids_in_page = []
|
|
||||||
for mobj in re.finditer(r'href="/video([0-9_]+)"', page):
|
|
||||||
if mobj.group(1) not in ids_in_page:
|
|
||||||
ids_in_page.append(mobj.group(1))
|
|
||||||
return ids_in_page
|
|
||||||
|
|
||||||
def _real_extract(self, url):
|
def _real_extract(self, url):
|
||||||
# Extract page id
|
# Extract page id
|
||||||
mobj = re.match(self._VALID_URL, url)
|
mobj = re.match(self._VALID_URL, url)
|
||||||
|
@ -232,7 +226,7 @@ class VKUserVideosIE(InfoExtractor):
|
||||||
# Download page and get video ids
|
# Download page and get video ids
|
||||||
page_id = mobj.group(1)
|
page_id = mobj.group(1)
|
||||||
page = self._download_webpage(url, page_id)
|
page = self._download_webpage(url, page_id)
|
||||||
video_ids = self.extract_videos_from_page(page)
|
video_ids = orderedSet(m.group(1) for m in re.finditer(r'href="/video([0-9_]+)"', page))
|
||||||
|
|
||||||
self._downloader.to_screen('[vk] User videos %s: Found %i videos' % (page_id, len(video_ids)))
|
self._downloader.to_screen('[vk] User videos %s: Found %i videos' % (page_id, len(video_ids)))
|
||||||
|
|
||||||
|
|
Loading…
Reference in a new issue