mirror of
https://github.com/ytdl-org/youtube-dl.git
synced 2024-12-15 13:47:48 +01:00
parent
24eb13b1c6
commit
1a57b8c18c
@ -1109,4 +1109,3 @@ from .zingmp3 import (
|
|||||||
ZingMp3SongIE,
|
ZingMp3SongIE,
|
||||||
ZingMp3AlbumIE,
|
ZingMp3AlbumIE,
|
||||||
)
|
)
|
||||||
from .zippcast import ZippCastIE
|
|
||||||
|
@ -1,94 +0,0 @@
|
|||||||
from __future__ import unicode_literals
|
|
||||||
|
|
||||||
import re
|
|
||||||
|
|
||||||
from .common import InfoExtractor
|
|
||||||
from ..utils import (
|
|
||||||
determine_ext,
|
|
||||||
str_to_int,
|
|
||||||
)
|
|
||||||
|
|
||||||
|
|
||||||
class ZippCastIE(InfoExtractor):
|
|
||||||
_VALID_URL = r'https?://(?:www\.)?zippcast\.com/(?:video/|videoview\.php\?.*\bvplay=)(?P<id>[0-9a-zA-Z]+)'
|
|
||||||
_TESTS = [{
|
|
||||||
# m3u8, hq direct link
|
|
||||||
'url': 'http://www.zippcast.com/video/c9cfd5c7e44dbc29c81',
|
|
||||||
'md5': '5ea0263b5606866c4d6cda0fc5e8c6b6',
|
|
||||||
'info_dict': {
|
|
||||||
'id': 'c9cfd5c7e44dbc29c81',
|
|
||||||
'ext': 'mp4',
|
|
||||||
'title': '[Vinesauce] Vinny - Digital Space Traveler',
|
|
||||||
'description': 'Muted on youtube, but now uploaded in it\'s original form.',
|
|
||||||
'thumbnail': 're:^https?://.*\.jpg$',
|
|
||||||
'uploader': 'vinesauce',
|
|
||||||
'view_count': int,
|
|
||||||
'categories': ['Entertainment'],
|
|
||||||
'tags': list,
|
|
||||||
},
|
|
||||||
}, {
|
|
||||||
# f4m, lq ipod direct link
|
|
||||||
'url': 'http://www.zippcast.com/video/b79c0a233e9c6581775',
|
|
||||||
'only_matching': True,
|
|
||||||
}, {
|
|
||||||
'url': 'http://www.zippcast.com/videoview.php?vplay=c9cfd5c7e44dbc29c81&auto=no',
|
|
||||||
'only_matching': True,
|
|
||||||
}]
|
|
||||||
|
|
||||||
def _real_extract(self, url):
|
|
||||||
video_id = self._match_id(url)
|
|
||||||
|
|
||||||
webpage = self._download_webpage(
|
|
||||||
'http://www.zippcast.com/video/%s' % video_id, video_id)
|
|
||||||
|
|
||||||
formats = []
|
|
||||||
video_url = self._search_regex(
|
|
||||||
r'<source[^>]+src=(["\'])(?P<url>.+?)\1', webpage,
|
|
||||||
'video url', default=None, group='url')
|
|
||||||
if video_url:
|
|
||||||
formats.append({
|
|
||||||
'url': video_url,
|
|
||||||
'format_id': 'http',
|
|
||||||
'preference': 0, # direct link is almost always of worse quality
|
|
||||||
})
|
|
||||||
src_url = self._search_regex(
|
|
||||||
r'src\s*:\s*(?:escape\()?(["\'])(?P<url>http://.+?)\1',
|
|
||||||
webpage, 'src', default=None, group='url')
|
|
||||||
ext = determine_ext(src_url)
|
|
||||||
if ext == 'm3u8':
|
|
||||||
formats.extend(self._extract_m3u8_formats(
|
|
||||||
src_url, video_id, 'mp4', entry_protocol='m3u8_native',
|
|
||||||
m3u8_id='hls', fatal=False))
|
|
||||||
elif ext == 'f4m':
|
|
||||||
formats.extend(self._extract_f4m_formats(
|
|
||||||
src_url, video_id, f4m_id='hds', fatal=False))
|
|
||||||
self._sort_formats(formats)
|
|
||||||
|
|
||||||
title = self._og_search_title(webpage)
|
|
||||||
description = self._og_search_description(webpage) or self._html_search_meta(
|
|
||||||
'description', webpage)
|
|
||||||
uploader = self._search_regex(
|
|
||||||
r'<a[^>]+href="https?://[^/]+/profile/[^>]+>([^<]+)</a>',
|
|
||||||
webpage, 'uploader', fatal=False)
|
|
||||||
thumbnail = self._og_search_thumbnail(webpage)
|
|
||||||
view_count = str_to_int(self._search_regex(
|
|
||||||
r'>([\d,.]+) views!', webpage, 'view count', fatal=False))
|
|
||||||
|
|
||||||
categories = re.findall(
|
|
||||||
r'<a[^>]+href="https?://[^/]+/categories/[^"]+">([^<]+),?<',
|
|
||||||
webpage)
|
|
||||||
tags = re.findall(
|
|
||||||
r'<a[^>]+href="https?://[^/]+/search/tags/[^"]+">([^<]+),?<',
|
|
||||||
webpage)
|
|
||||||
|
|
||||||
return {
|
|
||||||
'id': video_id,
|
|
||||||
'title': title,
|
|
||||||
'description': description,
|
|
||||||
'thumbnail': thumbnail,
|
|
||||||
'uploader': uploader,
|
|
||||||
'view_count': view_count,
|
|
||||||
'categories': categories,
|
|
||||||
'tags': tags,
|
|
||||||
'formats': formats,
|
|
||||||
}
|
|
Loading…
Reference in New Issue
Block a user