From f0c3a223cd446525b5321df7a6e0800ee1f88721 Mon Sep 17 00:00:00 2001 From: Grabien Date: Sun, 8 Aug 2021 23:20:54 +0300 Subject: [PATCH] - rumble plugin direct links support --- youtube_dl/extractor/rumble.py | 75 +++++++++++++++++++++++++++++++++- 1 file changed, 73 insertions(+), 2 deletions(-) diff --git a/youtube_dl/extractor/rumble.py b/youtube_dl/extractor/rumble.py index 4a0225109..1aac82abb 100644 --- a/youtube_dl/extractor/rumble.py +++ b/youtube_dl/extractor/rumble.py @@ -10,9 +10,11 @@ from ..utils import ( try_get, ) +import re + class RumbleEmbedIE(InfoExtractor): - _VALID_URL = r'https?://(?:www\.)?rumble\.com/embed/(?:[0-9a-z]+\.)?(?P[0-9a-z]+)' + _VALID_URL = r'https?://(?:www\.)?rumble\.com/(?:embed/(?:[0-9a-z]+\.)?(?P[0-9a-z]+)|(.*?)\.html)' _TESTS = [{ 'url': 'https://rumble.com/embed/v5pv5f', 'md5': '36a18a049856720189f30977ccbb2c34', @@ -26,10 +28,79 @@ class RumbleEmbedIE(InfoExtractor): }, { 'url': 'https://rumble.com/embed/ufe9n.v5pv5f', 'only_matching': True, + }, { + 'url': 'https://rumble.com/vhlrar-mike-lindell-to-confront-brian-kemp-and-doug-ducey-over-election-fraud.html', + 'only_matching': True, }] def _real_extract(self, url): - video_id = self._match_id(url) + if re.match(r'https?://(?:www\.)?rumble\.com/(.*?)\.html', url): + video_id = "" + content, urlh = self._download_webpage_handle(url, video_id) + video_id = re.findall(r'"embedUrl":"https://rumble\.com/embed/(.*?)/"', content)[0] + else: + video_id = self._match_id(url) + video = self._download_json( + 'https://rumble.com/embedJS/', video_id, + query={'request': 'video', 'v': video_id}) + title = video['title'] + + formats = [] + for height, ua in (video.get('ua') or {}).items(): + for i in range(2): + f_url = try_get(ua, lambda x: x[i], compat_str) + if f_url: + ext = determine_ext(f_url) + f = { + 'ext': ext, + 'format_id': '%s-%sp' % (ext, height), + 'height': int_or_none(height), + 'url': f_url, + } + bitrate = try_get(ua, lambda x: x[i + 2]['bitrate']) + if bitrate: + f['tbr'] = int_or_none(bitrate) + formats.append(f) + self._sort_formats(formats) + + author = video.get('author') or {} + + return { + 'id': video_id, + 'title': title, + 'formats': formats, + 'thumbnail': video.get('i'), + 'timestamp': parse_iso8601(video.get('pubDate')), + 'channel': author.get('name'), + 'channel_url': author.get('url'), + 'duration': int_or_none(video.get('duration')), + } + + +class RumbleRegularIE(RumbleEmbedIE): + _VALID_URL = r'https?://(?:www\.)?rumble\.com/(.*?)\.html' + _TESTS = [{ + 'url': 'https://rumble.com/vhlrar-mike-lindell-to-confront-brian-kemp-and-doug-ducey-over-election-fraud.html', + 'md5': '36a18a049856720189f30977ccbb2c34', + 'info_dict': { + 'id': 'v5pv5f', + 'ext': 'mp4', + 'title': 'WMAR 2 News Latest Headlines | October 20, 6pm', + 'timestamp': 1571611968, + 'upload_date': '20191020', + } + }, { + 'url': 'https://rumble.com/vhlrar-mike-lindell-to-confront-brian-kemp-and-doug-ducey-over-election-fraud.html', + 'only_matching': True, + }] + + def _real_extract(self, url): + video_id = "" + content, urlh = self._download_webpage_handle(url, video_id) + video_id = re.findall(r'"embedUrl":"https://rumble\.com/embed/(.*?)/"', content)[0] + + # this is + #video_id = self._match_id(url) video = self._download_json( 'https://rumble.com/embedJS/', video_id, query={'request': 'video', 'v': video_id})