1
0
mirror of https://github.com/ytdl-org/youtube-dl.git synced 2024-12-23 00:16:48 +00:00

[DoodStream] Add extractor from yt-dlp back-port and improve

This commit is contained in:
dirkf 2022-06-04 08:07:54 +01:00
parent 04fd3289d3
commit 3687a4f9bd
2 changed files with 120 additions and 0 deletions

View File

@ -0,0 +1,119 @@
# coding: utf-8
from __future__ import unicode_literals
import random
import string
import time
from ..compat import compat_filter as filter
from ..utils import (
clean_html,
ExtractorError,
get_element_by_class,
parse_duration,
parse_filesize,
update_url_query,
unified_strdate,
url_or_none,
)
from .common import InfoExtractor
class DoodStreamIE(InfoExtractor):
_VALID_URL = r'https?://(?:www\.)?dood\.(?:to|watch)/[ed]/(?P<id>[a-z0-9]+)'
_TESTS = [{
'url': 'http://dood.to/e/5s1wmbdacezb',
'md5': '4568b83b31e13242b3f1ff96c55f0595',
'info_dict': {
'id': '5s1wmbdacezb',
'ext': 'mp4',
'title': 'Kat Wonders - Monthly May 2020',
'description': 'Kat Wonders - Monthly May 2020 | DoodStream.com',
'thumbnail': 'https://img.doodcdn.com/snaps/flyus84qgl2fsk4g.jpg',
},
'skip': 'Video not found',
}, {
'url': 'http://dood.watch/d/5s1wmbdacezb',
'md5': '4568b83b31e13242b3f1ff96c55f0595',
'info_dict': {
'id': '5s1wmbdacezb',
'ext': 'mp4',
'title': 'Kat Wonders - Monthly May 2020',
'description': 'Kat Wonders - Monthly May 2020 | DoodStream.com',
'thumbnail': 'https://img.doodcdn.com/snaps/flyus84qgl2fsk4g.jpg',
},
'skip': 'Video not found',
}, {
'url': 'https://dood.to/d/jzrxn12t2s7n',
'md5': '3207e199426eca7c2aa23c2872e6728a',
'info_dict': {
'id': 'jzrxn12t2s7n',
'ext': 'mp4',
'title': 'Stacy Cruz Cute ALLWAYSWELL',
'description': 'Stacy Cruz Cute ALLWAYSWELL | DoodStream.com',
'thumbnail': 'https://img.doodcdn.com/snaps/8edqd5nppkac3x8u.jpg',
},
'skip': 'Video not found',
}, {
'url': 'https://dood.to/d/is34uy8wvaet',
'md5': '04740d3ba93bcd638aa7a097d9226710',
'info_dict': {
'id': 'is34uy8wvaet',
'ext': 'mp4',
'title': 'Akhanda (2021) Telugu DVDScr MP3 700MB - DoodStream',
'upload_date': '20211202',
'filesize_approx': int,
'duration': 9886,
}
}]
def _real_extract(self, url):
video_id = self._match_id(url)
url = 'https://dood.to/e/' + video_id
headers = {
'User-Agent': 'Mozilla/5.0', # (Windows NT 6.1; WOW64; rv:53.0) Gecko/20100101 Firefox/66.0',
}
webpage = self._download_webpage(url, video_id, headers=headers)
title = self._html_search_meta(('og:title', 'twitter:title'), webpage, default=None)
if not title:
title = self._html_search_regex(r'<title\b[^>]*>([^<]+?)(?:\|\s+DoodStream\s*)?</title', webpage, 'title')
if title == 'Video not found':
raise ExtractorError(title, expected=True)
token = self._html_search_regex(r'''[?&]token=([a-z0-9]+)[&']''', webpage, 'token')
headers.update({
# 'User-Agent': 'Mozilla/5.0', # (Windows NT 6.1; WOW64; rv:53.0) Gecko/20100101 Firefox/66.0',
'referer': url
})
pass_md5 = self._html_search_regex(r'(/pass_md5.*?)\'', webpage, 'pass_md5')
final_url = (
self._download_webpage('https://dood.to' + pass_md5, video_id, headers=headers, note='Downloading final URL')
+ ''.join((random.choice(string.ascii_letters + string.digits) for _ in range(10)))
)
final_url = update_url_query(final_url, {'token': token, 'expiry': int(time.time() * 1000), })
thumb = next(filter(None, (url_or_none(self._html_search_meta(x, webpage, default=None))
for x in ('og:image', 'twitter:image'))), None)
description = self._html_search_meta(
('og:description', 'description', 'twitter:description'), webpage, default=None)
webpage = self._download_webpage('https://dood.to/d/' + video_id, video_id, headers=headers, fatal=False)
def get_class_text(x):
return clean_html(get_element_by_class(x, webpage))
return {
'id': video_id,
'title': title,
'url': final_url,
'http_headers': headers,
'ext': 'mp4',
'upload_date': unified_strdate(get_class_text('uploadate')),
'duration': parse_duration(get_class_text('length')),
'filesize_approx': parse_filesize(get_class_text('size')),
'description': description,
'thumbnail': thumb,
}

View File

@ -299,6 +299,7 @@ from .democracynow import DemocracynowIE
from .dfb import DFBIE from .dfb import DFBIE
from .dhm import DHMIE from .dhm import DHMIE
from .digg import DiggIE from .digg import DiggIE
from .doodstream import DoodStreamIE
from .dotsub import DotsubIE from .dotsub import DotsubIE
from .douyutv import ( from .douyutv import (
DouyuShowIE, DouyuShowIE,