1
0
mirror of https://codeberg.org/polarisfm/youtube-dl synced 2024-11-27 19:04:32 +01:00
youtube-dl/youtube_dl/extractor/biqle.py
2019-04-03 11:08:42 +01:00

98 lines
3.4 KiB
Python

# coding: utf-8
from __future__ import unicode_literals
from .common import InfoExtractor
from .vk import VKIE
from ..utils import (
HEADRequest,
int_or_none,
)
class BIQLEIE(InfoExtractor):
_VALID_URL = r'https?://(?:www\.)?biqle\.(?:com|org|ru)/watch/(?P<id>-?\d+_\d+)'
_TESTS = [{
# Youtube embed
'url': 'https://biqle.ru/watch/-115995369_456239081',
'md5': '97af5a06ee4c29bbf9c001bdb1cf5c06',
'info_dict': {
'id': '8v4f-avW-VI',
'ext': 'mp4',
'title': "PASSE-PARTOUT - L'ete c'est fait pour jouer",
'description': 'Passe-Partout',
'uploader_id': 'mrsimpsonstef3',
'uploader': 'Phanolito',
'upload_date': '20120822',
},
}, {
'url': 'http://biqle.org/watch/-44781847_168547604',
'md5': '7f24e72af1db0edf7c1aaba513174f97',
'info_dict': {
'id': '-44781847_168547604',
'ext': 'mp4',
'title': 'Ребенок в шоке от автоматической мойки',
'timestamp': 1396633454,
'uploader': 'Dmitry Kotov',
'upload_date': '20140404',
'uploader_id': '47850140',
},
}]
def _real_extract(self, url):
video_id = self._match_id(url)
webpage = self._download_webpage(url, video_id)
embed_url = self._proto_relative_url(self._search_regex(
r'<iframe.+?src="((?:https?:)?//daxab\.com/[^"]+)".*?></iframe>',
webpage, 'embed url'))
if VKIE.suitable(embed_url):
return self.url_result(embed_url, VKIE.ie_key(), video_id)
self._request_webpage(
HEADRequest(embed_url), video_id, headers={'Referer': url})
video_id, sig, _, access_token = self._get_cookies(embed_url)['video_ext'].value.split('%3A')
item = self._download_json(
'https://api.vk.com/method/video.get', video_id,
headers={'User-Agent': 'okhttp/3.4.1'}, query={
'access_token': access_token,
'sig': sig,
'v': 5.44,
'videos': video_id,
})['response']['items'][0]
title = item['title']
formats = []
for f_id, f_url in item.get('files', {}).items():
if f_id == 'external':
return self.url_result(f_url)
ext, height = f_id.split('_')
formats.append({
'format_id': height + 'p',
'url': f_url,
'height': int_or_none(height),
'ext': ext,
})
self._sort_formats(formats)
thumbnails = []
for k, v in item.items():
if k.startswith('photo_') and v:
width = k.replace('photo_', '')
thumbnails.append({
'id': width,
'url': v,
'width': int_or_none(width),
})
return {
'id': video_id,
'title': title,
'formats': formats,
'comment_count': int_or_none(item.get('comments')),
'description': item.get('description'),
'duration': int_or_none(item.get('duration')),
'thumbnails': thumbnails,
'timestamp': int_or_none(item.get('date')),
'uploader': item.get('owner_id'),
'view_count': int_or_none(item.get('views')),
}