mirror of
https://github.com/yt-dlp/yt-dlp.git
synced 2024-12-13 17:16:16 +01:00
85 lines
2.6 KiB
Python
85 lines
2.6 KiB
Python
from __future__ import unicode_literals
|
|
|
|
from .common import InfoExtractor
|
|
from ..utils import int_or_none
|
|
|
|
|
|
class NewgroundsIE(InfoExtractor):
|
|
_VALID_URL = r'https?://(?:www\.)?newgrounds\.com/(?:audio/listen|portal/view)/(?P<id>[0-9]+)'
|
|
_TESTS = [{
|
|
'url': 'https://www.newgrounds.com/audio/listen/549479',
|
|
'md5': 'fe6033d297591288fa1c1f780386f07a',
|
|
'info_dict': {
|
|
'id': '549479',
|
|
'ext': 'mp3',
|
|
'title': 'B7 - BusMode',
|
|
'uploader': 'Burn7',
|
|
}
|
|
}, {
|
|
'url': 'https://www.newgrounds.com/portal/view/673111',
|
|
'md5': '3394735822aab2478c31b1004fe5e5bc',
|
|
'info_dict': {
|
|
'id': '673111',
|
|
'ext': 'mp4',
|
|
'title': 'Dancin',
|
|
'uploader': 'Squirrelman82',
|
|
},
|
|
}, {
|
|
# source format unavailable, additional mp4 formats
|
|
'url': 'http://www.newgrounds.com/portal/view/689400',
|
|
'info_dict': {
|
|
'id': '689400',
|
|
'ext': 'mp4',
|
|
'title': 'ZTV News Episode 8',
|
|
'uploader': 'BennettTheSage',
|
|
},
|
|
'params': {
|
|
'skip_download': True,
|
|
},
|
|
}]
|
|
|
|
def _real_extract(self, url):
|
|
media_id = self._match_id(url)
|
|
|
|
webpage = self._download_webpage(url, media_id)
|
|
|
|
title = self._html_search_regex(
|
|
r'<title>([^>]+)</title>', webpage, 'title')
|
|
|
|
video_url = self._parse_json(self._search_regex(
|
|
r'"url"\s*:\s*("[^"]+"),', webpage, ''), media_id)
|
|
|
|
formats = [{
|
|
'url': video_url,
|
|
'format_id': 'source',
|
|
'quality': 1,
|
|
}]
|
|
|
|
max_resolution = int_or_none(self._search_regex(
|
|
r'max_resolution["\']\s*:\s*(\d+)', webpage, 'max resolution',
|
|
default=None))
|
|
if max_resolution:
|
|
url_base = video_url.rpartition('.')[0]
|
|
for resolution in (360, 720, 1080):
|
|
if resolution > max_resolution:
|
|
break
|
|
formats.append({
|
|
'url': '%s.%dp.mp4' % (url_base, resolution),
|
|
'format_id': '%dp' % resolution,
|
|
'height': resolution,
|
|
})
|
|
|
|
self._check_formats(formats, media_id)
|
|
self._sort_formats(formats)
|
|
|
|
uploader = self._html_search_regex(
|
|
r'(?:Author|Writer)\s*<a[^>]+>([^<]+)', webpage, 'uploader',
|
|
fatal=False)
|
|
|
|
return {
|
|
'id': media_id,
|
|
'title': title,
|
|
'uploader': uploader,
|
|
'formats': formats,
|
|
}
|