summaryrefslogtreecommitdiffstats
path: root/yt_dlp/extractor/nate.py
diff options
context:
space:
mode:
authorDaniel Baumann <daniel.baumann@progress-linux.org>2024-08-05 09:06:10 +0000
committerDaniel Baumann <daniel.baumann@progress-linux.org>2024-08-05 09:06:10 +0000
commit7e21328585afda6d66f98ca476301680eeffac32 (patch)
tree266a4e7b20443f94572748781d71fc0375a15037 /yt_dlp/extractor/nate.py
parentAdding upstream version 2024.05.27. (diff)
downloadyt-dlp-7e21328585afda6d66f98ca476301680eeffac32.tar.xz
yt-dlp-7e21328585afda6d66f98ca476301680eeffac32.zip
Adding upstream version 2024.07.01.upstream/2024.07.01
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to 'yt_dlp/extractor/nate.py')
-rw-r--r--yt_dlp/extractor/nate.py24
1 files changed, 12 insertions, 12 deletions
diff --git a/yt_dlp/extractor/nate.py b/yt_dlp/extractor/nate.py
index 5e74caa..bbc641f 100644
--- a/yt_dlp/extractor/nate.py
+++ b/yt_dlp/extractor/nate.py
@@ -29,7 +29,7 @@ class NateIE(InfoExtractor):
'uploader_id': '3606',
'tags': 'count:59',
},
- 'params': {'skip_download': True}
+ 'params': {'skip_download': True},
}, {
'url': 'https://tv.nate.com/clip/4300566',
'info_dict': {
@@ -47,7 +47,7 @@ class NateIE(InfoExtractor):
'uploader_id': '27987',
'tags': 'count:20',
},
- 'params': {'skip_download': True}
+ 'params': {'skip_download': True},
}]
_QUALITY = {
@@ -60,8 +60,8 @@ class NateIE(InfoExtractor):
}
def _real_extract(self, url):
- id = self._match_id(url)
- video_data = self._download_json(f'https://tv.nate.com/api/v1/clip/{id}', id)
+ video_id = self._match_id(url)
+ video_data = self._download_json(f'https://tv.nate.com/api/v1/clip/{video_id}', video_id)
formats = [{
'format_id': f_url[-2:],
'url': f_url,
@@ -69,7 +69,7 @@ class NateIE(InfoExtractor):
'quality': int_or_none(f_url[-2:]),
} for f_url in video_data.get('smcUriList') or []]
return {
- 'id': id,
+ 'id': video_id,
'title': video_data.get('clipTitle'),
'description': video_data.get('synopsis'),
'thumbnail': video_data.get('contentImg'),
@@ -102,19 +102,19 @@ class NateProgramIE(InfoExtractor):
},
}]
- def _entries(self, id):
+ def _entries(self, playlist_id):
for page_num in itertools.count(1):
- program_data = self._download_json(f'https://tv.nate.com/api/v1/program/{id}/clip/ranking?size=20&page={page_num}',
- id, note=f'Downloading page {page_num}')
+ program_data = self._download_json(
+ f'https://tv.nate.com/api/v1/program/{playlist_id}/clip/ranking?size=20&page={page_num}',
+ playlist_id, note=f'Downloading page {page_num}')
for clip in program_data.get('content') or []:
clip_id = clip.get('clipSeq')
if clip_id:
yield self.url_result(
- 'https://tv.nate.com/clip/%s' % clip_id,
- ie=NateIE.ie_key(), video_id=clip_id)
+ f'https://tv.nate.com/clip/{clip_id}', NateIE, playlist_id)
if program_data.get('last'):
break
def _real_extract(self, url):
- id = self._match_id(url)
- return self.playlist_result(self._entries(id), playlist_id=id)
+ playlist_id = self._match_id(url)
+ return self.playlist_result(self._entries(playlist_id), playlist_id=playlist_id)