1
0
mirror of https://codeberg.org/polarisfm/youtube-dl synced 2024-11-24 01:14:32 +01:00
youtube-dl/youtube_dl/extractor/tvigle.py

139 lines
4.9 KiB
Python
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

# coding: utf-8
from __future__ import unicode_literals
import re
from .common import InfoExtractor
from ..utils import (
ExtractorError,
float_or_none,
int_or_none,
parse_age_limit,
try_get,
url_or_none,
)
class TvigleIE(InfoExtractor):
IE_NAME = 'tvigle'
IE_DESC = 'Интернет-телевидение Tvigle.ru'
_VALID_URL = r'https?://(?:www\.)?(?:tvigle\.ru/(?:[^/]+/)+(?P<display_id>[^/]+)/$|cloud\.tvigle\.ru/video/(?P<id>\d+))'
_GEO_BYPASS = False
_GEO_COUNTRIES = ['RU']
_TESTS = [
{
'url': 'http://www.tvigle.ru/video/sokrat/',
'info_dict': {
'id': '1848932',
'display_id': 'sokrat',
'ext': 'mp4',
'title': 'Сократ',
'description': 'md5:d6b92ffb7217b4b8ebad2e7665253c17',
'duration': 6586,
'age_limit': 12,
},
'skip': 'georestricted',
},
{
'url': 'http://www.tvigle.ru/video/vladimir-vysotskii/vedushchii-teleprogrammy-60-minut-ssha-o-vladimire-vysotskom/',
'info_dict': {
'id': '5142516',
'ext': 'flv',
'title': 'Ведущий телепрограммы «60 минут» (США) о Владимире Высоцком',
'description': 'md5:027f7dc872948f14c96d19b4178428a4',
'duration': 186.080,
'age_limit': 0,
},
'skip': 'georestricted',
}, {
'url': 'https://cloud.tvigle.ru/video/5267604/',
'only_matching': True,
}
]
def _real_extract(self, url):
mobj = re.match(self._VALID_URL, url)
video_id = mobj.group('id')
display_id = mobj.group('display_id')
if not video_id:
webpage = self._download_webpage(url, display_id)
video_id = self._html_search_regex(
(r'<div[^>]+class=["\']player["\'][^>]+id=["\'](\d+)',
r'cloudId\s*=\s*["\'](\d+)',
r'class="video-preview current_playing" id="(\d+)"'),
webpage, 'video id')
video_data = self._download_json(
'http://cloud.tvigle.ru/api/play/video/%s/' % video_id, display_id)
item = video_data['playlist']['items'][0]
videos = item.get('videos')
error_message = item.get('errorMessage')
if not videos and error_message:
if item.get('isGeoBlocked') is True:
self.raise_geo_restricted(
msg=error_message, countries=self._GEO_COUNTRIES)
else:
raise ExtractorError(
'%s returned error: %s' % (self.IE_NAME, error_message),
expected=True)
title = item['title']
description = item.get('description')
thumbnail = item.get('thumbnail')
duration = float_or_none(item.get('durationMilliseconds'), 1000)
age_limit = parse_age_limit(item.get('ageRestrictions'))
formats = []
for vcodec, url_or_fmts in item['videos'].items():
if vcodec == 'hls':
m3u8_url = url_or_none(url_or_fmts)
if not m3u8_url:
continue
formats.extend(self._extract_m3u8_formats(
m3u8_url, video_id, ext='mp4', entry_protocol='m3u8_native',
m3u8_id='hls', fatal=False))
elif vcodec == 'dash':
mpd_url = url_or_none(url_or_fmts)
if not mpd_url:
continue
formats.extend(self._extract_mpd_formats(
mpd_url, video_id, mpd_id='dash', fatal=False))
else:
if not isinstance(url_or_fmts, dict):
continue
for format_id, video_url in url_or_fmts.items():
if format_id == 'm3u8':
continue
video_url = url_or_none(video_url)
if not video_url:
continue
height = self._search_regex(
r'^(\d+)[pP]$', format_id, 'height', default=None)
filesize = int_or_none(try_get(
item, lambda x: x['video_files_size'][vcodec][format_id]))
formats.append({
'url': video_url,
'format_id': '%s-%s' % (vcodec, format_id),
'vcodec': vcodec,
'height': int_or_none(height),
'filesize': filesize,
})
self._sort_formats(formats)
return {
'id': video_id,
'display_id': display_id,
'title': title,
'description': description,
'thumbnail': thumbnail,
'duration': duration,
'age_limit': age_limit,
'formats': formats,
}