summaryrefslogtreecommitdiffstats
path: root/yt_dlp/extractor/cloudy.py
diff options
context:
space:
mode:
Diffstat (limited to 'yt_dlp/extractor/cloudy.py')
-rw-r--r--yt_dlp/extractor/cloudy.py57
1 files changed, 57 insertions, 0 deletions
diff --git a/yt_dlp/extractor/cloudy.py b/yt_dlp/extractor/cloudy.py
new file mode 100644
index 0000000..848643e
--- /dev/null
+++ b/yt_dlp/extractor/cloudy.py
@@ -0,0 +1,57 @@
+from .common import InfoExtractor
+from ..utils import (
+ str_to_int,
+ unified_strdate,
+)
+
+
+class CloudyIE(InfoExtractor):
+ _IE_DESC = 'cloudy.ec'
+ _VALID_URL = r'https?://(?:www\.)?cloudy\.ec/(?:v/|embed\.php\?.*?\bid=)(?P<id>[A-Za-z0-9]+)'
+ _TESTS = [{
+ 'url': 'https://www.cloudy.ec/v/af511e2527aac',
+ 'md5': '29832b05028ead1b58be86bf319397ca',
+ 'info_dict': {
+ 'id': 'af511e2527aac',
+ 'ext': 'mp4',
+ 'title': 'Funny Cats and Animals Compilation june 2013',
+ 'upload_date': '20130913',
+ 'view_count': int,
+ }
+ }, {
+ 'url': 'http://www.cloudy.ec/embed.php?autoplay=1&id=af511e2527aac',
+ 'only_matching': True,
+ }]
+
+ def _real_extract(self, url):
+ video_id = self._match_id(url)
+
+ webpage = self._download_webpage(
+ 'https://www.cloudy.ec/embed.php', video_id, query={
+ 'id': video_id,
+ 'playerPage': 1,
+ 'autoplay': 1,
+ })
+
+ info = self._parse_html5_media_entries(url, webpage, video_id)[0]
+
+ webpage = self._download_webpage(
+ 'https://www.cloudy.ec/v/%s' % video_id, video_id, fatal=False)
+
+ if webpage:
+ info.update({
+ 'title': self._search_regex(
+ r'<h\d[^>]*>([^<]+)<', webpage, 'title'),
+ 'upload_date': unified_strdate(self._search_regex(
+ r'>Published at (\d{4}-\d{1,2}-\d{1,2})', webpage,
+ 'upload date', fatal=False)),
+ 'view_count': str_to_int(self._search_regex(
+ r'([\d,.]+) views<', webpage, 'view count', fatal=False)),
+ })
+
+ if not info.get('title'):
+ info['title'] = video_id
+
+ info['id'] = video_id
+
+ return info