mirror of
https://github.com/yt-dlp/yt-dlp.git
synced 2024-11-09 20:52:30 +01:00
214 lines
7.5 KiB
Python
214 lines
7.5 KiB
Python
# coding: utf-8
|
|
from __future__ import unicode_literals
|
|
|
|
import re
|
|
|
|
from .common import InfoExtractor
|
|
from ..utils import (
|
|
decode_packed_codes,
|
|
determine_ext,
|
|
ExtractorError,
|
|
int_or_none,
|
|
NO_DEFAULT,
|
|
urlencode_postdata,
|
|
)
|
|
|
|
|
|
class XFileShareIE(InfoExtractor):
|
|
_SITES = (
|
|
(r'daclips\.(?:in|com)', 'DaClips'),
|
|
(r'filehoot\.com', 'FileHoot'),
|
|
(r'gorillavid\.(?:in|com)', 'GorillaVid'),
|
|
(r'movpod\.in', 'MovPod'),
|
|
(r'powerwatch\.pw', 'PowerWatch'),
|
|
(r'rapidvideo\.ws', 'Rapidvideo.ws'),
|
|
(r'thevideobee\.to', 'TheVideoBee'),
|
|
(r'vidto\.(?:me|se)', 'Vidto'),
|
|
(r'streamin\.to', 'Streamin.To'),
|
|
(r'xvidstage\.com', 'XVIDSTAGE'),
|
|
(r'vidabc\.com', 'Vid ABC'),
|
|
(r'vidbom\.com', 'VidBom'),
|
|
(r'vidlo\.us', 'vidlo'),
|
|
(r'rapidvideo\.(?:cool|org)', 'RapidVideo.TV'),
|
|
(r'fastvideo\.me', 'FastVideo.me'),
|
|
)
|
|
|
|
IE_DESC = 'XFileShare based sites: %s' % ', '.join(list(zip(*_SITES))[1])
|
|
_VALID_URL = (r'https?://(?P<host>(?:www\.)?(?:%s))/(?:embed-)?(?P<id>[0-9a-zA-Z]+)'
|
|
% '|'.join(site for site in list(zip(*_SITES))[0]))
|
|
|
|
_FILE_NOT_FOUND_REGEXES = (
|
|
r'>(?:404 - )?File Not Found<',
|
|
r'>The file was removed by administrator<',
|
|
)
|
|
|
|
_TESTS = [{
|
|
'url': 'http://gorillavid.in/06y9juieqpmi',
|
|
'md5': '5ae4a3580620380619678ee4875893ba',
|
|
'info_dict': {
|
|
'id': '06y9juieqpmi',
|
|
'ext': 'mp4',
|
|
'title': 'Rebecca Black My Moment Official Music Video Reaction-6GK87Rc8bzQ',
|
|
'thumbnail': r're:http://.*\.jpg',
|
|
},
|
|
}, {
|
|
'url': 'http://gorillavid.in/embed-z08zf8le23c6-960x480.html',
|
|
'only_matching': True,
|
|
}, {
|
|
'url': 'http://daclips.in/3rso4kdn6f9m',
|
|
'md5': '1ad8fd39bb976eeb66004d3a4895f106',
|
|
'info_dict': {
|
|
'id': '3rso4kdn6f9m',
|
|
'ext': 'mp4',
|
|
'title': 'Micro Pig piglets ready on 16th July 2009-bG0PdrCdxUc',
|
|
'thumbnail': r're:http://.*\.jpg',
|
|
}
|
|
}, {
|
|
'url': 'http://movpod.in/0wguyyxi1yca',
|
|
'only_matching': True,
|
|
}, {
|
|
'url': 'http://filehoot.com/3ivfabn7573c.html',
|
|
'info_dict': {
|
|
'id': '3ivfabn7573c',
|
|
'ext': 'mp4',
|
|
'title': 'youtube-dl test video \'äBaW_jenozKc.mp4.mp4',
|
|
'thumbnail': r're:http://.*\.jpg',
|
|
},
|
|
'skip': 'Video removed',
|
|
}, {
|
|
'url': 'http://vidto.me/ku5glz52nqe1.html',
|
|
'info_dict': {
|
|
'id': 'ku5glz52nqe1',
|
|
'ext': 'mp4',
|
|
'title': 'test'
|
|
}
|
|
}, {
|
|
'url': 'http://powerwatch.pw/duecjibvicbu',
|
|
'info_dict': {
|
|
'id': 'duecjibvicbu',
|
|
'ext': 'mp4',
|
|
'title': 'Big Buck Bunny trailer',
|
|
},
|
|
}, {
|
|
'url': 'http://xvidstage.com/e0qcnl03co6z',
|
|
'info_dict': {
|
|
'id': 'e0qcnl03co6z',
|
|
'ext': 'mp4',
|
|
'title': 'Chucky Prank 2015.mp4',
|
|
},
|
|
}, {
|
|
# removed by administrator
|
|
'url': 'http://xvidstage.com/amfy7atlkx25',
|
|
'only_matching': True,
|
|
}, {
|
|
'url': 'http://vidabc.com/i8ybqscrphfv',
|
|
'info_dict': {
|
|
'id': 'i8ybqscrphfv',
|
|
'ext': 'mp4',
|
|
'title': 're:Beauty and the Beast 2017',
|
|
},
|
|
'params': {
|
|
'skip_download': True,
|
|
},
|
|
}, {
|
|
'url': 'http://www.rapidvideo.cool/b667kprndr8w',
|
|
'only_matching': True,
|
|
}, {
|
|
'url': 'http://www.fastvideo.me/k8604r8nk8sn/FAST_FURIOUS_8_-_Trailer_italiano_ufficiale.mp4.html',
|
|
'only_matching': True,
|
|
}, {
|
|
'url': 'http://vidto.se/1tx1pf6t12cg.html',
|
|
'only_matching': True,
|
|
}]
|
|
|
|
@staticmethod
|
|
def _extract_urls(webpage):
|
|
return [
|
|
mobj.group('url')
|
|
for mobj in re.finditer(
|
|
r'<iframe\b[^>]+\bsrc=(["\'])(?P<url>(?:https?:)?//(?:%s)/embed-[0-9a-zA-Z]+.*?)\1'
|
|
% '|'.join(site for site in list(zip(*XFileShareIE._SITES))[0]),
|
|
webpage)]
|
|
|
|
def _real_extract(self, url):
|
|
mobj = re.match(self._VALID_URL, url)
|
|
video_id = mobj.group('id')
|
|
|
|
url = 'http://%s/%s' % (mobj.group('host'), video_id)
|
|
webpage = self._download_webpage(url, video_id)
|
|
|
|
if any(re.search(p, webpage) for p in self._FILE_NOT_FOUND_REGEXES):
|
|
raise ExtractorError('Video %s does not exist' % video_id, expected=True)
|
|
|
|
fields = self._hidden_inputs(webpage)
|
|
|
|
if fields['op'] == 'download1':
|
|
countdown = int_or_none(self._search_regex(
|
|
r'<span id="countdown_str">(?:[Ww]ait)?\s*<span id="cxc">(\d+)</span>\s*(?:seconds?)?</span>',
|
|
webpage, 'countdown', default=None))
|
|
if countdown:
|
|
self._sleep(countdown, video_id)
|
|
|
|
webpage = self._download_webpage(
|
|
url, video_id, 'Downloading video page',
|
|
data=urlencode_postdata(fields), headers={
|
|
'Referer': url,
|
|
'Content-type': 'application/x-www-form-urlencoded',
|
|
})
|
|
|
|
title = (self._search_regex(
|
|
(r'style="z-index: [0-9]+;">([^<]+)</span>',
|
|
r'<td nowrap>([^<]+)</td>',
|
|
r'h4-fine[^>]*>([^<]+)<',
|
|
r'>Watch (.+) ',
|
|
r'<h2 class="video-page-head">([^<]+)</h2>',
|
|
r'<h2 style="[^"]*color:#403f3d[^"]*"[^>]*>([^<]+)<'), # streamin.to
|
|
webpage, 'title', default=None) or self._og_search_title(
|
|
webpage, default=None) or video_id).strip()
|
|
|
|
def extract_formats(default=NO_DEFAULT):
|
|
urls = []
|
|
for regex in (
|
|
r'(?:file|src)\s*:\s*(["\'])(?P<url>http(?:(?!\1).)+\.(?:m3u8|mp4|flv)(?:(?!\1).)*)\1',
|
|
r'file_link\s*=\s*(["\'])(?P<url>http(?:(?!\1).)+)\1',
|
|
r'addVariable\((\\?["\'])file\1\s*,\s*(\\?["\'])(?P<url>http(?:(?!\2).)+)\2\)',
|
|
r'<embed[^>]+src=(["\'])(?P<url>http(?:(?!\1).)+\.(?:m3u8|mp4|flv)(?:(?!\1).)*)\1'):
|
|
for mobj in re.finditer(regex, webpage):
|
|
video_url = mobj.group('url')
|
|
if video_url not in urls:
|
|
urls.append(video_url)
|
|
formats = []
|
|
for video_url in urls:
|
|
if determine_ext(video_url) == 'm3u8':
|
|
formats.extend(self._extract_m3u8_formats(
|
|
video_url, video_id, 'mp4',
|
|
entry_protocol='m3u8_native', m3u8_id='hls',
|
|
fatal=False))
|
|
else:
|
|
formats.append({
|
|
'url': video_url,
|
|
'format_id': 'sd',
|
|
})
|
|
if not formats and default is not NO_DEFAULT:
|
|
return default
|
|
self._sort_formats(formats)
|
|
return formats
|
|
|
|
formats = extract_formats(default=None)
|
|
|
|
if not formats:
|
|
webpage = decode_packed_codes(self._search_regex(
|
|
r"(}\('(.+)',(\d+),(\d+),'[^']*\b(?:file|embed)\b[^']*'\.split\('\|'\))",
|
|
webpage, 'packed code'))
|
|
formats = extract_formats()
|
|
|
|
thumbnail = self._search_regex(
|
|
r'image\s*:\s*["\'](http[^"\']+)["\'],', webpage, 'thumbnail', default=None)
|
|
|
|
return {
|
|
'id': video_id,
|
|
'title': title,
|
|
'thumbnail': thumbnail,
|
|
'formats': formats,
|
|
}
|