aboutsummaryrefslogtreecommitdiff
path: root/youtube_dl/extractor/wdr.py
diff options
context:
space:
mode:
authorYen Chi Hsuan <yan12125@gmail.com>2016-06-09 13:04:30 +0800
committerYen Chi Hsuan <yan12125@gmail.com>2016-06-09 13:04:30 +0800
commit50918c4ee01be6c1218a72bef35838216b2bf8d1 (patch)
tree883f9fbf10ce4fb757ab12d12ed89dde2caebc8f /youtube_dl/extractor/wdr.py
parent6c33d24b46ecfb1f2ce790e21f2410149fdfb095 (diff)
[wdr] Support radio players (closes #6147)
Diffstat (limited to 'youtube_dl/extractor/wdr.py')
-rw-r--r--youtube_dl/extractor/wdr.py29
1 files changed, 24 insertions, 5 deletions
diff --git a/youtube_dl/extractor/wdr.py b/youtube_dl/extractor/wdr.py
index 6174eb19f..059e2aa08 100644
--- a/youtube_dl/extractor/wdr.py
+++ b/youtube_dl/extractor/wdr.py
@@ -10,12 +10,13 @@ from ..utils import (
strip_jsonp,
unified_strdate,
ExtractorError,
+ urlhandle_detect_ext,
)
class WDRIE(InfoExtractor):
_CURRENT_MAUS_URL = r'https?://(?:www\.)wdrmaus.de/(?:[^/]+/){1,2}[^/?#]+\.php5'
- _PAGE_REGEX = r'/mediathek/(?P<media_type>[^/]+)/(?P<type>[^/]+)/(?P<display_id>.+)\.html'
+ _PAGE_REGEX = r'/(?:mediathek/)?(?P<media_type>[^/]+)/(?P<type>[^/]+)/(?P<display_id>.+)\.html'
_VALID_URL = r'(?P<page_url>https?://(?:www\d\.)?wdr\d?\.de)' + _PAGE_REGEX + '|' + _CURRENT_MAUS_URL
_TESTS = [
@@ -97,6 +98,16 @@ class WDRIE(InfoExtractor):
'description': '- Die Sendung mit der Maus -',
},
},
+ {
+ 'url': 'http://www1.wdr.de/radio/player/radioplayer116~_layout-popupVersion.html',
+ 'info_dict': {
+ 'id': 'mdb-869971',
+ 'ext': 'mp3',
+ 'title': 'Funkhaus Europa Livestream',
+ 'description': 'md5:2309992a6716c347891c045be50992e4',
+ 'upload_date': '20160101',
+ },
+ }
]
def _real_extract(self, url):
@@ -107,9 +118,10 @@ class WDRIE(InfoExtractor):
webpage = self._download_webpage(url, display_id)
# for wdr.de the data-extension is in a tag with the class "mediaLink"
+ # for wdr.de radio players, in a tag with the class "wdrrPlayerPlayBtn"
# for wdrmaus its in a link to the page in a multiline "videoLink"-tag
json_metadata = self._html_search_regex(
- r'class=(?:"mediaLink\b[^"]*"[^>]+|"videoLink\b[^"]*"[\s]*>\n[^\n]*)data-extension="([^"]+)"',
+ r'class=(?:"(?:mediaLink|wdrrPlayerPlayBtn)\b[^"]*"[^>]+|"videoLink\b[^"]*"[\s]*>\n[^\n]*)data-extension="([^"]+)"',
webpage, 'media link', default=None, flags=re.MULTILINE)
if not json_metadata:
@@ -143,15 +155,22 @@ class WDRIE(InfoExtractor):
for tag_name in ['videoURL', 'audioURL']:
if tag_name in metadata_media_alt:
alt_url = metadata_media_alt[tag_name]
- if determine_ext(alt_url) == 'm3u8':
+ ext = determine_ext(alt_url)
+ if ext == 'm3u8':
m3u_fmt = self._extract_m3u8_formats(
alt_url, display_id, 'mp4', 'm3u8_native',
m3u8_id='hls')
formats.extend(m3u_fmt)
else:
- formats.append({
+ a_format = {
'url': alt_url
- })
+ }
+ if ext == 'unknown_video':
+ urlh = self._request_webpage(
+ alt_url, display_id, note='Determining extension')
+ ext = urlhandle_detect_ext(urlh)
+ a_format['ext'] = ext
+ formats.append(a_format)
# check if there are flash-streams for this video
if 'dflt' in metadata_media_resource and 'videoURL' in metadata_media_resource['dflt']: