0
0
Fork 0
mirror of https://github.com/yt-dlp/yt-dlp.git synced 2024-11-27 03:03:01 +00:00
yt-dlp/yt_dlp/extractor/rumble.py

333 lines
12 KiB
Python
Raw Normal View History

import itertools
import re
from .common import InfoExtractor
from ..compat import compat_HTTPError
from ..utils import (
ExtractorError,
UnsupportedError,
clean_html,
get_element_by_class,
int_or_none,
parse_count,
parse_iso8601,
traverse_obj,
2022-03-16 20:07:04 +00:00
unescapeHTML,
)
class RumbleEmbedIE(InfoExtractor):
_VALID_URL = r'https?://(?:www\.)?rumble\.com/embed/(?:[0-9a-z]+\.)?(?P<id>[0-9a-z]+)'
_EMBED_REGEX = [fr'(?:<(?:script|iframe)[^>]+\bsrc=|["\']embedUrl["\']\s*:\s*)["\'](?P<url>{_VALID_URL})']
_TESTS = [{
'url': 'https://rumble.com/embed/v5pv5f',
'md5': '36a18a049856720189f30977ccbb2c34',
'info_dict': {
'id': 'v5pv5f',
'ext': 'mp4',
'title': 'WMAR 2 News Latest Headlines | October 20, 6pm',
'timestamp': 1571611968,
'upload_date': '20191020',
'channel_url': 'https://rumble.com/c/WMAR',
'channel': 'WMAR',
'thumbnail': 'https://sp.rmbl.ws/s8/1/5/M/z/1/5Mz1a.OvCc-small-WMAR-2-News-Latest-Headline.jpg',
'duration': 234,
'uploader': 'WMAR',
'live_status': 'not_live',
}
2022-03-16 20:07:04 +00:00
}, {
'url': 'https://rumble.com/embed/vslb7v',
'md5': '7418035de1a30a178b8af34dc2b6a52b',
'info_dict': {
'id': 'vslb7v',
'ext': 'mp4',
'title': 'Defense Sec. says US Commitment to NATO Defense \'Ironclad\'',
'timestamp': 1645142135,
'upload_date': '20220217',
'channel_url': 'https://rumble.com/c/CyberTechNews',
'channel': 'CTNews',
'thumbnail': 'https://sp.rmbl.ws/s8/6/7/i/9/h/7i9hd.OvCc.jpg',
'duration': 901,
'uploader': 'CTNews',
'live_status': 'not_live',
2022-03-16 20:07:04 +00:00
}
}, {
'url': 'https://rumble.com/embed/vunh1h',
'info_dict': {
'id': 'vunh1h',
'ext': 'mp4',
'title': 'Gideon, op zoek naar de waarheid including ENG SUBS',
'timestamp': 1647197663,
'upload_date': '20220313',
'channel_url': 'https://rumble.com/user/BLCKBX',
'channel': 'BLCKBX',
'thumbnail': r're:https://.+\.jpg',
'duration': 5069,
'uploader': 'BLCKBX',
'live_status': 'not_live',
'subtitles': {
'en': [
{
'url': r're:https://.+\.vtt',
'name': 'English',
'ext': 'vtt'
}
]
},
},
'params': {'skip_download': True}
}, {
'url': 'https://rumble.com/embed/v1essrt',
'info_dict': {
'id': 'v1essrt',
'ext': 'mp4',
'title': 'startswith:lofi hip hop radio - beats to relax/study',
'timestamp': 1661519399,
'upload_date': '20220826',
'channel_url': 'https://rumble.com/c/LofiGirl',
'channel': 'Lofi Girl',
'thumbnail': r're:https://.+\.jpg',
'duration': None,
'uploader': 'Lofi Girl',
'live_status': 'is_live',
},
'params': {'skip_download': True}
}, {
'url': 'https://rumble.com/embed/v1amumr',
'info_dict': {
'id': 'v1amumr',
'ext': 'webm',
'fps': 60,
'title': 'Turning Point USA 2022 Student Action Summit DAY 1 - Rumble Exclusive Live',
'timestamp': 1658518457,
'upload_date': '20220722',
'channel_url': 'https://rumble.com/c/RumbleEvents',
'channel': 'Rumble Events',
'thumbnail': r're:https://.+\.jpg',
'duration': 16427,
'uploader': 'Rumble Events',
'live_status': 'was_live',
},
'params': {'skip_download': True}
}, {
'url': 'https://rumble.com/embed/ufe9n.v5pv5f',
'only_matching': True,
}]
_WEBPAGE_TESTS = [
{
'note': 'Rumble JS embed',
'url': 'https://therightscoop.com/what-does-9-plus-1-plus-1-equal-listen-to-this-audio-of-attempted-kavanaugh-assassins-call-and-youll-get-it',
'md5': '4701209ac99095592e73dbba21889690',
'info_dict': {
'id': 'v15eqxl',
'ext': 'mp4',
'channel': 'Mr Producer Media',
'duration': 92,
'title': '911 Audio From The Man Who Wanted To Kill Supreme Court Justice Kavanaugh',
'channel_url': 'https://rumble.com/c/RichSementa',
'thumbnail': 'https://sp.rmbl.ws/s8/1/P/j/f/A/PjfAe.OvCc-small-911-Audio-From-The-Man-Who-.jpg',
'timestamp': 1654892716,
'uploader': 'Mr Producer Media',
'upload_date': '20220610',
'live_status': 'not_live',
}
},
]
@classmethod
def _extract_embed_urls(cls, url, webpage):
embeds = tuple(super()._extract_embed_urls(url, webpage))
if embeds:
return embeds
return [f'https://rumble.com/embed/{mobj.group("id")}' for mobj in re.finditer(
r'<script>\s*Rumble\(\s*"play"\s*,\s*{\s*[\'"]video[\'"]\s*:\s*[\'"](?P<id>[0-9a-z]+)[\'"]', webpage)]
def _real_extract(self, url):
video_id = self._match_id(url)
video = self._download_json(
'https://rumble.com/embedJS/u3/', video_id,
query={'request': 'video', 'ver': 2, 'v': video_id})
sys_msg = traverse_obj(video, ('sys', 'msg'))
if sys_msg:
self.report_warning(sys_msg, video_id=video_id)
if video.get('live') == 0:
live_status = 'not_live' if video.get('livestream_has_dvr') is None else 'was_live'
elif video.get('live') == 1:
live_status = 'is_upcoming' if video.get('livestream_has_dvr') else 'was_live'
elif video.get('live') == 2:
live_status = 'is_live'
else:
live_status = None
formats = []
for ext, ext_info in (video.get('ua') or {}).items():
for height, video_info in (ext_info or {}).items():
meta = video_info.get('meta') or {}
if not video_info.get('url'):
continue
if ext == 'hls':
if meta.get('live') is True and video.get('live') == 1:
live_status = 'post_live'
formats.extend(self._extract_m3u8_formats(
video_info['url'], video_id,
ext='mp4', m3u8_id='hls', fatal=False, live=live_status == 'is_live'))
continue
formats.append({
'ext': ext,
'url': video_info['url'],
'format_id': '%s-%sp' % (ext, height),
'height': int_or_none(height),
'fps': video.get('fps'),
**traverse_obj(meta, {
'tbr': 'bitrate',
'filesize': 'size',
'width': 'w',
'height': 'h',
}, default={})
})
subtitles = {
lang: [{
'url': sub_info['path'],
'name': sub_info.get('language') or '',
}] for lang, sub_info in (video.get('cc') or {}).items() if sub_info.get('path')
}
author = video.get('author') or {}
thumbnails = traverse_obj(video, ('t', ..., {'url': 'i', 'width': 'w', 'height': 'h'}))
if not thumbnails and video.get('i'):
thumbnails = [{'url': video['i']}]
if live_status in {'is_live', 'post_live'}:
duration = None
else:
duration = int_or_none(video.get('duration'))
return {
'id': video_id,
'title': unescapeHTML(video.get('title')),
'formats': formats,
'subtitles': subtitles,
'thumbnails': thumbnails,
'timestamp': parse_iso8601(video.get('pubDate')),
'channel': author.get('name'),
'channel_url': author.get('url'),
'duration': duration,
'uploader': author.get('name'),
'live_status': live_status,
}
class RumbleIE(InfoExtractor):
_VALID_URL = r'https?://(?:www\.)?rumble\.com/(?P<id>v(?!ideos)[\w.-]+)[^/]*$'
_EMBED_REGEX = [r'<a class=video-item--a href=(?P<url>/v[\w.-]+\.html)>']
_TESTS = [{
'add_ie': ['RumbleEmbed'],
'url': 'https://rumble.com/vdmum1-moose-the-dog-helps-girls-dig-a-snow-fort.html',
'md5': '53af34098a7f92c4e51cf0bd1c33f009',
'info_dict': {
'id': 'vb0ofn',
'ext': 'mp4',
'timestamp': 1612662578,
'uploader': 'LovingMontana',
'channel': 'LovingMontana',
'upload_date': '20210207',
'title': 'Winter-loving dog helps girls dig a snow fort ',
'description': 'Moose the dog is more than happy to help with digging out this epic snow fort. Great job, Moose!',
'channel_url': 'https://rumble.com/c/c-546523',
'thumbnail': r're:https://.+\.jpg',
'duration': 103,
'like_count': int,
'view_count': int,
'live_status': 'not_live',
}
}, {
'url': 'http://www.rumble.com/vDMUM1?key=value',
'only_matching': True,
}]
_WEBPAGE_TESTS = [{
'url': 'https://rumble.com/videos?page=2',
'playlist_count': 25,
'info_dict': {
'id': 'videos?page=2',
'title': 'All videos',
'description': 'Browse videos uploaded to Rumble.com',
'age_limit': 0,
},
}, {
'url': 'https://rumble.com/live-videos',
'playlist_mincount': 19,
'info_dict': {
'id': 'live-videos',
'title': 'Live Videos',
'description': 'Live videos on Rumble.com',
'age_limit': 0,
},
}, {
'url': 'https://rumble.com/search/video?q=rumble&sort=views',
'playlist_count': 24,
'info_dict': {
'id': 'video?q=rumble&sort=views',
'title': 'Search results for: rumble',
'age_limit': 0,
},
}]
def _real_extract(self, url):
page_id = self._match_id(url)
webpage = self._download_webpage(url, page_id)
url_info = next(RumbleEmbedIE.extract_from_webpage(self._downloader, url, webpage), None)
if not url_info:
raise UnsupportedError(url)
release_ts_str = self._search_regex(
r'(?:Livestream begins|Streamed on):\s+<time datetime="([^"]+)',
webpage, 'release date', fatal=False, default=None)
view_count_str = self._search_regex(r'<span class="media-heading-info">([\d,]+) Views',
webpage, 'view count', fatal=False, default=None)
return self.url_result(
url_info['url'], ie_key=url_info['ie_key'], url_transparent=True,
view_count=parse_count(view_count_str),
release_timestamp=parse_iso8601(release_ts_str),
like_count=parse_count(get_element_by_class('rumbles-count', webpage)),
description=clean_html(get_element_by_class('media-description', webpage)),
)
class RumbleChannelIE(InfoExtractor):
_VALID_URL = r'(?P<url>https?://(?:www\.)?rumble\.com/(?:c|user)/(?P<id>[^&?#$/]+))'
_TESTS = [{
'url': 'https://rumble.com/c/Styxhexenhammer666',
'playlist_mincount': 1160,
'info_dict': {
'id': 'Styxhexenhammer666',
},
}, {
'url': 'https://rumble.com/user/goldenpoodleharleyeuna',
'playlist_mincount': 4,
'info_dict': {
'id': 'goldenpoodleharleyeuna',
},
}]
def entries(self, url, playlist_id):
for page in itertools.count(1):
try:
webpage = self._download_webpage(f'{url}?page={page}', playlist_id, note='Downloading page %d' % page)
except ExtractorError as e:
if isinstance(e.cause, compat_HTTPError) and e.cause.code == 404:
break
raise
for video_url in re.findall(r'class=video-item--a\s?href=([^>]+\.html)', webpage):
yield self.url_result('https://rumble.com' + video_url)
def _real_extract(self, url):
url, playlist_id = self._match_valid_url(url).groups()
return self.playlist_result(self.entries(url, playlist_id), playlist_id=playlist_id)