1
0
mirror of https://github.com/yt-dlp/yt-dlp.git synced 2024-11-09 12:42:28 +01:00
yt-dlp/yt_dlp/extractor/mediaklikk.py

105 lines
4.3 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
from ..utils import (
unified_strdate
)
from .common import InfoExtractor
from ..compat import (
compat_urllib_parse_unquote,
compat_str
)
class MediaKlikkIE(InfoExtractor):
_VALID_URL = r'''(?x)^https?:\/\/(?:www\.)?
(?:mediaklikk|m4sport|hirado|petofilive)\.hu\/.*?videok?\/
(?:(?P<year>[0-9]{4})/(?P<month>[0-9]{1,2})/(?P<day>[0-9]{1,2})/)?
(?P<id>[^/#?_]+)'''
_TESTS = [{
# mediaklikk. date in html.
'url': 'https://mediaklikk.hu/video/hazajaro-delnyugat-bacska-a-duna-menten-palankatol-doroszloig/',
'info_dict': {
'id': '4754129',
'title': 'Hazajáró, DÉLNYUGAT-BÁCSKA A Duna mentén Palánkától Doroszlóig',
'ext': 'mp4',
'upload_date': '20210901',
'thumbnail': 'http://mediaklikk.hu/wp-content/uploads/sites/4/2014/02/hazajarouj_JO.jpg'
}
}, {
# m4sport
'url': 'https://m4sport.hu/video/2021/08/30/gyemant-liga-parizs/',
'info_dict': {
'id': '4754999',
'title': 'Gyémánt Liga, Párizs',
'ext': 'mp4',
'upload_date': '20210830',
'thumbnail': 'http://m4sport.hu/wp-content/uploads/sites/4/2021/08/vlcsnap-2021-08-30-18h21m20s10-1024x576.jpg'
}
}, {
# m4sport with *video/ url and no date
'url': 'https://m4sport.hu/bl-video/real-madrid-chelsea-1-1/',
'info_dict': {
'id': '4492099',
'title': 'Real Madrid - Chelsea 1-1',
'ext': 'mp4',
'thumbnail': 'http://m4sport.hu/wp-content/uploads/sites/4/2021/04/Sequence-01.Still001-1024x576.png'
}
}, {
# hirado
'url': 'https://hirado.hu/videok/felteteleket-szabott-a-fovaros/',
'info_dict': {
'id': '4760120',
'title': 'Feltételeket szabott a főváros',
'ext': 'mp4',
'thumbnail': 'http://hirado.hu/wp-content/uploads/sites/4/2021/09/vlcsnap-2021-09-01-20h20m37s165.jpg'
}
}, {
# petofilive
'url': 'https://petofilive.hu/video/2021/06/07/tha-shudras-az-akusztikban/',
'info_dict': {
'id': '4571948',
'title': 'Tha Shudras az Akusztikban',
'ext': 'mp4',
'upload_date': '20210607',
'thumbnail': 'http://petofilive.hu/wp-content/uploads/sites/4/2021/06/vlcsnap-2021-06-07-22h14m23s915-1024x576.jpg'
}
}]
def _real_extract(self, url):
mobj = self._match_valid_url(url)
display_id = mobj.group('id')
webpage = self._download_webpage(url, display_id)
player_data_str = self._html_search_regex(
r'mtva_player_manager\.player\(document.getElementById\(.*\),\s?(\{.*\}).*\);', webpage, 'player data')
player_data = self._parse_json(player_data_str, display_id, compat_urllib_parse_unquote)
video_id = compat_str(player_data['contentId'])
title = player_data.get('title') or self._og_search_title(webpage, fatal=False) or \
self._html_search_regex(r'<h\d+\b[^>]+\bclass="article_title">([^<]+)<', webpage, 'title')
upload_date = unified_strdate(
'%s-%s-%s' % (mobj.group('year'), mobj.group('month'), mobj.group('day')))
if not upload_date:
upload_date = unified_strdate(self._html_search_regex(
r'<p+\b[^>]+\bclass="article_date">([^<]+)<', webpage, 'upload date', default=None))
player_data['video'] = player_data.pop('token')
player_page = self._download_webpage('https://player.mediaklikk.hu/playernew/player.php', video_id, query=player_data)
playlist_url = self._proto_relative_url(compat_urllib_parse_unquote(
self._html_search_regex(r'\"file\":\s*\"(\\?/\\?/.*playlist\.m3u8)\"', player_page, 'playlist_url')).replace('\\/', '/'))
formats = self._extract_wowza_formats(
playlist_url, video_id, skip_protocols=['f4m', 'smil', 'dash'])
self._sort_formats(formats)
return {
'id': video_id,
'title': title,
'display_id': display_id,
'formats': formats,
'upload_date': upload_date,
'thumbnail': player_data.get('bgImage') or self._og_search_thumbnail(webpage)
}