mirror of
https://github.com/ytdl-org/youtube-dl
synced 2024-12-30 03:25:51 +01:00
89 lines
2.7 KiB
Python
89 lines
2.7 KiB
Python
# coding: utf-8
|
|
from __future__ import unicode_literals
|
|
|
|
from .common import InfoExtractor
|
|
from ..compat import compat_str
|
|
from ..utils import (
|
|
float_or_none,
|
|
int_or_none,
|
|
try_get,
|
|
)
|
|
|
|
|
|
class CarambaTVIE(InfoExtractor):
|
|
_VALID_URL = r'(?:carambatv:|https?://video1\.carambatv\.ru/v/)(?P<id>\d+)'
|
|
_TESTS = [{
|
|
'url': 'http://video1.carambatv.ru/v/191910501',
|
|
'md5': '2f4a81b7cfd5ab866ee2d7270cb34a2a',
|
|
'info_dict': {
|
|
'id': '191910501',
|
|
'ext': 'mp4',
|
|
'title': '[BadComedian] - Разборка в Маниле (Абсолютный обзор)',
|
|
'thumbnail': 're:^https?://.*\.jpg',
|
|
'duration': 2678.31,
|
|
},
|
|
}, {
|
|
'url': 'carambatv:191910501',
|
|
'only_matching': True,
|
|
}]
|
|
|
|
def _real_extract(self, url):
|
|
video_id = self._match_id(url)
|
|
|
|
video = self._download_json(
|
|
'http://video1.carambatv.ru/v/%s/videoinfo.js' % video_id,
|
|
video_id)
|
|
|
|
title = video['title']
|
|
|
|
base_url = video.get('video') or 'http://video1.carambatv.ru/v/%s/' % video_id
|
|
|
|
formats = [{
|
|
'url': base_url + f['fn'],
|
|
'height': int_or_none(f.get('height')),
|
|
'format_id': '%sp' % f['height'] if f.get('height') else None,
|
|
} for f in video['qualities'] if f.get('fn')]
|
|
self._sort_formats(formats)
|
|
|
|
thumbnail = video.get('splash')
|
|
duration = float_or_none(try_get(
|
|
video, lambda x: x['annotations'][0]['end_time'], compat_str))
|
|
|
|
return {
|
|
'id': video_id,
|
|
'title': title,
|
|
'thumbnail': thumbnail,
|
|
'duration': duration,
|
|
'formats': formats,
|
|
}
|
|
|
|
|
|
class CarambaTVPageIE(InfoExtractor):
|
|
_VALID_URL = r'https?://carambatv\.ru/(?:[^/]+/)+(?P<id>[^/?#&]+)'
|
|
_TEST = {
|
|
'url': 'http://carambatv.ru/movie/bad-comedian/razborka-v-manile/',
|
|
'md5': '',
|
|
'info_dict': {
|
|
'id': '191910501',
|
|
'ext': 'mp4',
|
|
'title': '[BadComedian] - Разборка в Маниле (Абсолютный обзор)',
|
|
'thumbnail': 're:^https?://.*\.jpg$',
|
|
'duration': 2678.31,
|
|
},
|
|
}
|
|
|
|
def _real_extract(self, url):
|
|
video_id = self._match_id(url)
|
|
|
|
webpage = self._download_webpage(url, video_id)
|
|
|
|
video_url = self._og_search_property('video:iframe', webpage, default=None)
|
|
|
|
if not video_url:
|
|
video_id = self._search_regex(
|
|
r'(?:video_id|crmb_vuid)\s*[:=]\s*["\']?(\d+)',
|
|
webpage, 'video id')
|
|
video_url = 'carambatv:%s' % video_id
|
|
|
|
return self.url_result(video_url, CarambaTVIE.ie_key())
|