mirror of
https://github.com/ytdl-org/youtube-dl.git
synced 2024-11-25 00:21:24 +01:00
Compare commits
9 Commits
c5dffddb55
...
00bfe6430e
Author | SHA1 | Date | |
---|---|---|---|
|
00bfe6430e | ||
|
e29f8e8707 | ||
|
9179a0294b | ||
|
0b817e51c2 | ||
|
ff968566a5 | ||
|
3525025a6f | ||
|
d4d99b1b51 | ||
|
001d5b8395 | ||
|
8d2881a442 |
|
@ -1534,6 +1534,7 @@ from .voxmedia import (
|
||||||
VoxMediaVolumeIE,
|
VoxMediaVolumeIE,
|
||||||
VoxMediaIE,
|
VoxMediaIE,
|
||||||
)
|
)
|
||||||
|
from .voe import VOEIE
|
||||||
from .vrt import VRTIE
|
from .vrt import VRTIE
|
||||||
from .vrak import VrakIE
|
from .vrak import VrakIE
|
||||||
from .vrv import (
|
from .vrv import (
|
||||||
|
|
75
youtube_dl/extractor/voe.py
Normal file
75
youtube_dl/extractor/voe.py
Normal file
|
@ -0,0 +1,75 @@
|
||||||
|
# coding: utf-8
|
||||||
|
from __future__ import unicode_literals
|
||||||
|
|
||||||
|
from .common import InfoExtractor
|
||||||
|
from ..compat import compat_base64_b64decode
|
||||||
|
from ..utils import (
|
||||||
|
int_or_none,
|
||||||
|
js_to_json,
|
||||||
|
url_or_none,
|
||||||
|
)
|
||||||
|
|
||||||
|
|
||||||
|
class VOEIE(InfoExtractor):
|
||||||
|
IE_NAME = 'voe'
|
||||||
|
IE_DESC = 'VOE.SX'
|
||||||
|
_VALID_URL = r'https?://voe\.sx(?P<path>/(e/)?(?P<id>[a-z0-9]+))'
|
||||||
|
_TESTS = [{
|
||||||
|
'url': 'https://voe.sx/e/ng7ja5n5n2y8',
|
||||||
|
'info_dict': {
|
||||||
|
'id': 'ng7ja5n5n2y8',
|
||||||
|
'title': 'md5:8dd774de9b73851151d80ef6baaea7f1',
|
||||||
|
'thumbnail': r're:^https?://.*\.jpg$',
|
||||||
|
'ext': 'm3u8',
|
||||||
|
},
|
||||||
|
}]
|
||||||
|
|
||||||
|
def _real_extract(self, url):
|
||||||
|
video_id, video_path = self._match_valid_url(url).group('id', 'path')
|
||||||
|
|
||||||
|
webpage = self._download_webpage(
|
||||||
|
'https://voe.sx/e/%s' % video_id, video_id)
|
||||||
|
|
||||||
|
player_url = self._search_regex(
|
||||||
|
r'''("|')(?P<url>https://(?!voe\.sx/)[^/]+%s)\1\s*;''' % (video_path,),
|
||||||
|
webpage, 'redirect', group='url', default=None)
|
||||||
|
if player_url:
|
||||||
|
webpage = self._download_webpage(
|
||||||
|
player_url, video_id, note='Redirecting to player page')
|
||||||
|
|
||||||
|
sources = self._search_json(
|
||||||
|
r'\bsources\s*=', webpage, 'sources', video_id, transform_source=js_to_json)
|
||||||
|
|
||||||
|
title = self._search_regex(
|
||||||
|
r'<title>(?:Watch\s+)?(?P<title>.+?)(?:-\s+VOE\s+\|.+)?</title>',
|
||||||
|
webpage, 'title', group='title')
|
||||||
|
|
||||||
|
formats = []
|
||||||
|
for fmt in ('mp4', 'hls'):
|
||||||
|
if fmt not in sources:
|
||||||
|
continue
|
||||||
|
sources[fmt] = url_or_none(sources[fmt]) or url_or_none(compat_base64_b64decode(sources[fmt]).decode('utf-8'))
|
||||||
|
f_url = sources.get('hls')
|
||||||
|
if f_url:
|
||||||
|
formats.extend(self._extract_m3u8_formats(
|
||||||
|
f_url, video_id, entry_protocol='m3u8_native', fatal=False))
|
||||||
|
f_url = sources.get('mp4')
|
||||||
|
if f_url:
|
||||||
|
formats.append({
|
||||||
|
'url': f_url,
|
||||||
|
'ext': 'mp4',
|
||||||
|
'height': int_or_none(sources.get('video_height')),
|
||||||
|
})
|
||||||
|
|
||||||
|
self._sort_formats(formats)
|
||||||
|
|
||||||
|
thumbnail = url_or_none(self._search_regex(
|
||||||
|
r'''(?:VOEPlayer\s*\.\s*|data-)poster\s*=\s*("|')(?P<thumbnail>(?:(?!\1)\S)+)\1''',
|
||||||
|
webpage, 'thumbnail', group='thumbnail', default=None))
|
||||||
|
|
||||||
|
return {
|
||||||
|
'id': video_id,
|
||||||
|
'title': title,
|
||||||
|
'formats': formats,
|
||||||
|
'thumbnail': thumbnail,
|
||||||
|
}
|
Loading…
Reference in New Issue
Block a user