aboutsummaryrefslogtreecommitdiff
path: root/youtube_dl
diff options
context:
space:
mode:
Diffstat (limited to 'youtube_dl')
-rw-r--r--youtube_dl/extractor/__init__.py1
-rw-r--r--youtube_dl/extractor/thechive.py60
2 files changed, 61 insertions, 0 deletions
diff --git a/youtube_dl/extractor/__init__.py b/youtube_dl/extractor/__init__.py
index ddb9d6670..b064a3a7c 100644
--- a/youtube_dl/extractor/__init__.py
+++ b/youtube_dl/extractor/__init__.py
@@ -481,6 +481,7 @@ from .tenplay import TenPlayIE
from .testurl import TestURLIE
from .testtube import TestTubeIE
from .tf1 import TF1IE
+from .thechive import TheChiveIE
from .theonion import TheOnionIE
from .theplatform import ThePlatformIE
from .thesixtyone import TheSixtyOneIE
diff --git a/youtube_dl/extractor/thechive.py b/youtube_dl/extractor/thechive.py
new file mode 100644
index 000000000..df1a7998d
--- /dev/null
+++ b/youtube_dl/extractor/thechive.py
@@ -0,0 +1,60 @@
+from __future__ import unicode_literals
+
+import re
+
+from .common import InfoExtractor
+from ..utils import unified_strdate
+
+
+
+class TheChiveIE(InfoExtractor):
+ _VALID_URL = r'http://(www\.)?thechive\.com/[^/]+/[^/]+/[^/]+/(?P<video_id>[A-Za-z\-]+)'
+ _TEST = {
+ 'url': "http://thechive.com/2015/02/20/so-thats-what-a-set-of-redneck-bagpipes-sound-like-video/",
+ 'md5': "366710dda77cfa727bdef3523ba8466f",
+ 'info_dict': {
+ 'id': "so-thats-what-a-set-of-redneck-bagpipes-sound-like-video",
+ 'title': "So that's what a set of redneck bagpipes sound like... (Video)",
+ 'description': "Okay that was pretty good. Now play Freebird!...",
+ 'thumbnail': "https://thechive.files.wordpress.com/2015/02/0_07dghz0w-thumbnail2.jpg",
+ 'author': "Ben",
+ 'upload_date': "20150220",
+ 'ext': "mp4"
+ }
+ }
+
+ def _real_extract(self, url):
+ mobj = re.match(self._VALID_URL, url)
+ video_id = mobj.group('video_id')
+ webpage = self._download_webpage(url, video_id)
+
+ title = self._og_search_title(webpage)
+ description = self._html_search_regex(r'(?s)<meta name="description" content="(.*?)" />', webpage, 'description')
+ thumbnail = self._og_search_thumbnail(webpage)
+ author = self._html_search_regex(
+ r'(?s)itemprop="author">(.+?)</span>', webpage, 'author', fatal=False).capitalize()
+ upload_date = unified_strdate(self._html_search_regex(
+ r'(?s)itemprop="datePublished" datetime="(.+?)">', webpage, 'upload_date', fatal=False))
+
+ # Adapted from extractor/musicvault.py
+ VIDEO_URL_TEMPLATE = 'http://cdnapi.kaltura.com/p/%(uid)s/sp/%(wid)s/playManifest/entryId/%(entry_id)s/format/url/protocol/http'
+
+ kaltura_id = self._search_regex(
+ r'entry_id=([^"]+)',
+ webpage, 'kaltura ID')
+ video_url = VIDEO_URL_TEMPLATE % {
+ 'entry_id': kaltura_id,
+ 'wid': self._search_regex(r'partner_id/([0-9]+)\?', webpage, 'wid'),
+ 'uid': self._search_regex(r'uiconf_id/([0-9]+)/', webpage, 'uid'),
+ }
+
+ return {
+ 'url': video_url,
+ 'id': video_id,
+ 'title': title,
+ 'description': description,
+ 'thumbnail': thumbnail,
+ 'author': author,
+ 'upload_date': upload_date,
+ 'ext': 'mp4'
+ } \ No newline at end of file