mirror of
https://github.com/yt-dlp/yt-dlp.git
synced 2024-11-09 10:50:38 +00:00
add96eb9f8
Authored by: seproDev Reviewed-by: bashonly <88596187+bashonly@users.noreply.github.com> Reviewed-by: Simon Sawicki <contact@grub4k.xyz>
298 lines
11 KiB
Python
298 lines
11 KiB
Python
import datetime as dt
|
|
import re
|
|
import urllib.parse
|
|
|
|
from .common import InfoExtractor
|
|
from ..utils import (
|
|
ExtractorError,
|
|
OnDemandPagedList,
|
|
remove_end,
|
|
str_to_int,
|
|
unified_strdate,
|
|
)
|
|
|
|
|
|
class MotherlessIE(InfoExtractor):
|
|
_VALID_URL = r'https?://(?:www\.)?motherless\.com/(?:g/[a-z0-9_]+/|G[VIG]?[A-F0-9]+/)?(?P<id>[A-F0-9]+)'
|
|
_TESTS = [{
|
|
'url': 'http://motherless.com/EE97006',
|
|
'md5': 'cb5e7438f7a3c4e886b7bccc1292a3bc',
|
|
'info_dict': {
|
|
'id': 'EE97006',
|
|
'ext': 'mp4',
|
|
'title': 'Dogging blond Brit getting glazed (comp)',
|
|
'categories': ['UK', 'slag', 'whore', 'dogging', 'cunt', 'cumhound', 'big tits', 'Pearl Necklace'],
|
|
'upload_date': '20230519',
|
|
'uploader_id': 'deathbird',
|
|
'thumbnail': r're:https?://.*\.jpg',
|
|
'age_limit': 18,
|
|
'comment_count': int,
|
|
'view_count': int,
|
|
'like_count': int,
|
|
},
|
|
'params': {
|
|
# Incomplete cert chains
|
|
'nocheckcertificate': True,
|
|
},
|
|
}, {
|
|
'url': 'http://motherless.com/532291B',
|
|
'md5': 'bc59a6b47d1f958e61fbd38a4d31b131',
|
|
'info_dict': {
|
|
'id': '532291B',
|
|
'ext': 'mp4',
|
|
'title': 'Amazing girl playing the omegle game, PERFECT!',
|
|
'categories': ['Amateur', 'webcam', 'omegle', 'pink', 'young', 'masturbate', 'teen',
|
|
'game', 'hairy'],
|
|
'upload_date': '20140622',
|
|
'uploader_id': 'Sulivana7x',
|
|
'thumbnail': r're:https?://.*\.jpg',
|
|
'age_limit': 18,
|
|
},
|
|
'skip': '404',
|
|
}, {
|
|
'url': 'http://motherless.com/g/cosplay/633979F',
|
|
'md5': '0b2a43f447a49c3e649c93ad1fafa4a0',
|
|
'info_dict': {
|
|
'id': '633979F',
|
|
'ext': 'mp4',
|
|
'title': 'Turtlette',
|
|
'categories': ['superheroine heroine superher'],
|
|
'upload_date': '20140827',
|
|
'uploader_id': 'shade0230',
|
|
'thumbnail': r're:https?://.*\.jpg',
|
|
'age_limit': 18,
|
|
'like_count': int,
|
|
'comment_count': int,
|
|
'view_count': int,
|
|
},
|
|
'params': {
|
|
'nocheckcertificate': True,
|
|
},
|
|
}, {
|
|
'url': 'http://motherless.com/8B4BBC1',
|
|
'info_dict': {
|
|
'id': '8B4BBC1',
|
|
'ext': 'mp4',
|
|
'title': 'VIDEO00441.mp4',
|
|
'categories': [],
|
|
'upload_date': '20160214',
|
|
'uploader_id': 'NMWildGirl',
|
|
'thumbnail': r're:https?://.*\.jpg',
|
|
'age_limit': 18,
|
|
'like_count': int,
|
|
'comment_count': int,
|
|
'view_count': int,
|
|
},
|
|
'params': {
|
|
'nocheckcertificate': True,
|
|
},
|
|
}, {
|
|
# see https://motherless.com/videos/recent for recent videos with
|
|
# uploaded date in "ago" format
|
|
'url': 'https://motherless.com/3C3E2CF',
|
|
'info_dict': {
|
|
'id': '3C3E2CF',
|
|
'ext': 'mp4',
|
|
'title': 'a/ Hot Teens',
|
|
'categories': list,
|
|
'upload_date': '20210104',
|
|
'uploader_id': 'anonymous',
|
|
'thumbnail': r're:https?://.*\.jpg',
|
|
'age_limit': 18,
|
|
'like_count': int,
|
|
'comment_count': int,
|
|
'view_count': int,
|
|
},
|
|
'params': {
|
|
'nocheckcertificate': True,
|
|
},
|
|
}]
|
|
|
|
def _real_extract(self, url):
|
|
video_id = self._match_id(url)
|
|
webpage = self._download_webpage(url, video_id)
|
|
|
|
if any(p in webpage for p in (
|
|
'<title>404 - MOTHERLESS.COM<',
|
|
">The page you're looking for cannot be found.<")):
|
|
raise ExtractorError(f'Video {video_id} does not exist', expected=True)
|
|
|
|
if '>The content you are trying to view is for friends only.' in webpage:
|
|
raise ExtractorError(f'Video {video_id} is for friends only', expected=True)
|
|
|
|
title = self._html_search_regex(
|
|
(r'(?s)<div[^>]+\bclass=["\']media-meta-title[^>]+>(.+?)</div>',
|
|
r'id="view-upload-title">\s+([^<]+)<'), webpage, 'title')
|
|
video_url = (self._html_search_regex(
|
|
(r'setup\(\{\s*["\']file["\']\s*:\s*(["\'])(?P<url>(?:(?!\1).)+)\1',
|
|
r'fileurl\s*=\s*(["\'])(?P<url>(?:(?!\1).)+)\1'),
|
|
webpage, 'video URL', default=None, group='url')
|
|
or f'http://cdn4.videos.motherlessmedia.com/videos/{video_id}.mp4?fs=opencloud')
|
|
age_limit = self._rta_search(webpage)
|
|
view_count = str_to_int(self._html_search_regex(
|
|
(r'>([\d,.]+)\s+Views<', r'<strong>Views</strong>\s+([^<]+)<'),
|
|
webpage, 'view count', fatal=False))
|
|
like_count = str_to_int(self._html_search_regex(
|
|
(r'>([\d,.]+)\s+Favorites<',
|
|
r'<strong>Favorited</strong>\s+([^<]+)<'),
|
|
webpage, 'like count', fatal=False))
|
|
|
|
upload_date = unified_strdate(self._search_regex(
|
|
r'class=["\']count[^>]+>(\d+\s+[a-zA-Z]{3}\s+\d{4})<', webpage,
|
|
'upload date', default=None))
|
|
if not upload_date:
|
|
uploaded_ago = self._search_regex(
|
|
r'>\s*(\d+[hd])\s+[aA]go\b', webpage, 'uploaded ago',
|
|
default=None)
|
|
if uploaded_ago:
|
|
delta = int(uploaded_ago[:-1])
|
|
_AGO_UNITS = {
|
|
'h': 'hours',
|
|
'd': 'days',
|
|
}
|
|
kwargs = {_AGO_UNITS.get(uploaded_ago[-1]): delta}
|
|
upload_date = (dt.datetime.now(dt.timezone.utc) - dt.timedelta(**kwargs)).strftime('%Y%m%d')
|
|
|
|
comment_count = len(re.findall(r'''class\s*=\s*['"]media-comment-contents\b''', webpage))
|
|
uploader_id = self._html_search_regex(
|
|
(r'''<span\b[^>]+\bclass\s*=\s*["']username\b[^>]*>([^<]+)</span>''',
|
|
r'''(?s)['"](?:media-meta-member|thumb-member-username)\b[^>]+>\s*<a\b[^>]+\bhref\s*=\s*['"]/m/([^"']+)'''),
|
|
webpage, 'uploader_id', fatal=False)
|
|
categories = self._html_search_meta('keywords', webpage, default='')
|
|
categories = [cat.strip() for cat in categories.split(',') if cat.strip()]
|
|
|
|
return {
|
|
'id': video_id,
|
|
'title': title,
|
|
'upload_date': upload_date,
|
|
'uploader_id': uploader_id,
|
|
'thumbnail': self._og_search_thumbnail(webpage),
|
|
'categories': categories,
|
|
'view_count': view_count,
|
|
'like_count': like_count,
|
|
'comment_count': comment_count,
|
|
'age_limit': age_limit,
|
|
'url': video_url,
|
|
}
|
|
|
|
|
|
class MotherlessPaginatedIE(InfoExtractor):
|
|
_EXTRA_QUERY = {}
|
|
_PAGE_SIZE = 60
|
|
|
|
def _correct_path(self, url, item_id):
|
|
raise NotImplementedError('This method must be implemented by subclasses')
|
|
|
|
def _extract_entries(self, webpage, base):
|
|
for mobj in re.finditer(r'href="[^"]*(?P<href>/[A-F0-9]+)"\s+title="(?P<title>[^"]+)',
|
|
webpage):
|
|
video_url = urllib.parse.urljoin(base, mobj.group('href'))
|
|
video_id = MotherlessIE.get_temp_id(video_url)
|
|
|
|
if video_id:
|
|
yield self.url_result(video_url, MotherlessIE, video_id, mobj.group('title'))
|
|
|
|
def _real_extract(self, url):
|
|
item_id = self._match_id(url)
|
|
real_url = self._correct_path(url, item_id)
|
|
webpage = self._download_webpage(real_url, item_id, 'Downloading page 1')
|
|
|
|
def get_page(idx):
|
|
page = idx + 1
|
|
current_page = webpage if not idx else self._download_webpage(
|
|
real_url, item_id, note=f'Downloading page {page}', query={'page': page, **self._EXTRA_QUERY})
|
|
yield from self._extract_entries(current_page, real_url)
|
|
|
|
return self.playlist_result(
|
|
OnDemandPagedList(get_page, self._PAGE_SIZE), item_id,
|
|
remove_end(self._html_extract_title(webpage), ' | MOTHERLESS.COM ™'))
|
|
|
|
|
|
class MotherlessGroupIE(MotherlessPaginatedIE):
|
|
_VALID_URL = r'https?://(?:www\.)?motherless\.com/g[vifm]?/(?P<id>[a-z0-9_]+)/?(?:$|[#?])'
|
|
_TESTS = [{
|
|
'url': 'http://motherless.com/gv/movie_scenes',
|
|
'info_dict': {
|
|
'id': 'movie_scenes',
|
|
'title': 'Movie Scenes - Videos - Hot and sexy scenes from "regular" movies... Beautiful actresses fully',
|
|
},
|
|
'playlist_mincount': 540,
|
|
}, {
|
|
'url': 'http://motherless.com/g/sex_must_be_funny',
|
|
'info_dict': {
|
|
'id': 'sex_must_be_funny',
|
|
'title': 'Sex must be funny',
|
|
},
|
|
'playlist_count': 0,
|
|
}, {
|
|
'url': 'https://motherless.com/gv/beautiful_cock',
|
|
'info_dict': {
|
|
'id': 'beautiful_cock',
|
|
'title': 'Beautiful Cock',
|
|
},
|
|
'playlist_mincount': 2040,
|
|
}]
|
|
|
|
def _correct_path(self, url, item_id):
|
|
return urllib.parse.urljoin(url, f'/gv/{item_id}')
|
|
|
|
|
|
class MotherlessGalleryIE(MotherlessPaginatedIE):
|
|
_VALID_URL = r'https?://(?:www\.)?motherless\.com/G[VIG]?(?P<id>[A-F0-9]+)/?(?:$|[#?])'
|
|
_TESTS = [{
|
|
'url': 'https://motherless.com/GV338999F',
|
|
'info_dict': {
|
|
'id': '338999F',
|
|
'title': 'Random',
|
|
},
|
|
'playlist_mincount': 171,
|
|
}, {
|
|
'url': 'https://motherless.com/GVABD6213',
|
|
'info_dict': {
|
|
'id': 'ABD6213',
|
|
'title': 'Cuties',
|
|
},
|
|
'playlist_mincount': 2,
|
|
}, {
|
|
'url': 'https://motherless.com/GVBCF7622',
|
|
'info_dict': {
|
|
'id': 'BCF7622',
|
|
'title': 'Vintage',
|
|
},
|
|
'playlist_count': 0,
|
|
}, {
|
|
'url': 'https://motherless.com/G035DE2F',
|
|
'info_dict': {
|
|
'id': '035DE2F',
|
|
'title': 'General',
|
|
},
|
|
'playlist_mincount': 420,
|
|
}]
|
|
|
|
def _correct_path(self, url, item_id):
|
|
return urllib.parse.urljoin(url, f'/GV{item_id}')
|
|
|
|
|
|
class MotherlessUploaderIE(MotherlessPaginatedIE):
|
|
_VALID_URL = r'https?://(?:www\.)?motherless\.com/u/(?P<id>\w+)/?(?:$|[?#])'
|
|
_TESTS = [{
|
|
'url': 'https://motherless.com/u/Mrgo4hrs2023',
|
|
'info_dict': {
|
|
'id': 'Mrgo4hrs2023',
|
|
'title': "Mrgo4hrs2023's Uploads - Videos",
|
|
},
|
|
'playlist_mincount': 32,
|
|
}, {
|
|
'url': 'https://motherless.com/u/Happy_couple?t=v',
|
|
'info_dict': {
|
|
'id': 'Happy_couple',
|
|
'title': "Happy_couple's Uploads - Videos",
|
|
},
|
|
'playlist_mincount': 8,
|
|
}]
|
|
|
|
_EXTRA_QUERY = {'t': 'v'}
|
|
|
|
def _correct_path(self, url, item_id):
|
|
return urllib.parse.urljoin(url, f'/u/{item_id}?t=v')
|