0
0
Fork 0
mirror of https://github.com/yt-dlp/yt-dlp.git synced 2024-11-14 01:05:06 +00:00
yt-dlp/yt_dlp/extractor/tnaflix.py
oxamun 989f47b631
[extractor/tnaflix] Fix extractor (#6086)
Closes #6085
Authored by: oxamun, bashonly
2023-02-12 09:51:29 +05:30

337 lines
13 KiB
Python

import re
from .common import InfoExtractor
from ..compat import compat_str
from ..utils import (
fix_xml_ampersands,
float_or_none,
int_or_none,
parse_duration,
str_to_int,
unescapeHTML,
url_basename,
xpath_text,
)
class TNAFlixNetworkBaseIE(InfoExtractor):
# May be overridden in descendants if necessary
_CONFIG_REGEX = [
r'flashvars\.config\s*=\s*escape\("(?P<url>[^"]+)"',
r'<input[^>]+name="config\d?" value="(?P<url>[^"]+)"',
r'config\s*=\s*(["\'])(?P<url>(?:https?:)?//(?:(?!\1).)+)\1',
]
_TITLE_REGEX = r'<input[^>]+name="title" value="([^"]+)"'
_DESCRIPTION_REGEX = r'<input[^>]+name="description" value="([^"]+)"'
_UPLOADER_REGEX = r'<input[^>]+name="username" value="([^"]+)"'
_VIEW_COUNT_REGEX = None
_COMMENT_COUNT_REGEX = None
_AVERAGE_RATING_REGEX = None
_CATEGORIES_REGEX = r'<li[^>]*>\s*<span[^>]+class="infoTitle"[^>]*>Categories:</span>\s*<span[^>]+class="listView"[^>]*>(.+?)</span>\s*</li>'
def _extract_thumbnails(self, flix_xml):
def get_child(elem, names):
for name in names:
child = elem.find(name)
if child is not None:
return child
timeline = get_child(flix_xml, ['timeline', 'rolloverBarImage'])
if timeline is None:
return
pattern_el = get_child(timeline, ['imagePattern', 'pattern'])
if pattern_el is None or not pattern_el.text:
return
first_el = get_child(timeline, ['imageFirst', 'first'])
last_el = get_child(timeline, ['imageLast', 'last'])
if first_el is None or last_el is None:
return
first_text = first_el.text
last_text = last_el.text
if not first_text.isdigit() or not last_text.isdigit():
return
first = int(first_text)
last = int(last_text)
if first > last:
return
width = int_or_none(xpath_text(timeline, './imageWidth', 'thumbnail width'))
height = int_or_none(xpath_text(timeline, './imageHeight', 'thumbnail height'))
return [{
'url': self._proto_relative_url(pattern_el.text.replace('#', compat_str(i)), 'http:'),
'width': width,
'height': height,
} for i in range(first, last + 1)]
def _real_extract(self, url):
mobj = self._match_valid_url(url)
video_id, host = mobj.group('id', 'host')
for display_id_key in ('display_id', 'display_id_2'):
if display_id_key in mobj.groupdict():
display_id = mobj.group(display_id_key)
if display_id:
break
else:
display_id = video_id
webpage = self._download_webpage(url, display_id)
inputs = self._hidden_inputs(webpage)
query = {}
# check for MovieFap-style config
cfg_url = self._proto_relative_url(self._html_search_regex(
self._CONFIG_REGEX, webpage, 'flashvars.config', default=None,
group='url'), 'http:')
if not cfg_url:
cfg_url = inputs.get('config')
# check for TNAFlix-style config
if not cfg_url and inputs.get('vkey') and inputs.get('nkey'):
cfg_url = f'http://cdn-fck.{host}.com/{host}/{inputs["vkey"]}.fid'
query.update({
'key': inputs['nkey'],
'VID': video_id,
'premium': '1',
'vip': '1',
'alpha': '',
})
formats, json_ld = [], {}
# TNAFlix and MovieFap extraction
if cfg_url:
cfg_xml = self._download_xml(
cfg_url, display_id, 'Downloading metadata',
transform_source=fix_xml_ampersands, headers={'Referer': url}, query=query)
def extract_video_url(vl):
# Any URL modification now results in HTTP Error 403: Forbidden
return unescapeHTML(vl.text)
video_link = cfg_xml.find('./videoLink')
if video_link is not None:
formats.append({
'url': extract_video_url(video_link),
'ext': xpath_text(cfg_xml, './videoConfig/type', 'type', default='flv'),
})
for item in cfg_xml.findall('./quality/item'):
video_link = item.find('./videoLink')
if video_link is None:
continue
res = item.find('res')
format_id = None if res is None else res.text
height = int_or_none(self._search_regex(
r'^(\d+)[pP]', format_id, 'height', default=None))
formats.append({
'url': self._proto_relative_url(extract_video_url(video_link), 'http:'),
'format_id': format_id,
'height': height,
})
thumbnails = self._extract_thumbnails(cfg_xml) or []
thumbnails.append({
'url': self._proto_relative_url(xpath_text(cfg_xml, './startThumb', 'thumbnail'), 'http:')
})
# check for EMPFlix-style JSON and extract
else:
player = self._download_json(
f'http://www.{host}.com/ajax/video-player/{video_id}', video_id,
headers={'Referer': url}).get('html', '')
for mobj in re.finditer(r'<source src="(?P<src>[^"]+)"', player):
video_url = mobj.group('src')
height = self._search_regex(r'-(\d+)p\.', url_basename(video_url), 'height', default=None)
formats.append({
'url': self._proto_relative_url(video_url, 'http:'),
'ext': url_basename(video_url).split('.')[-1],
'height': int_or_none(height),
'format_id': f'{height}p' if height else url_basename(video_url).split('.')[0],
})
thumbnail = self._proto_relative_url(self._search_regex(
r'data-poster="([^"]+)"', player, 'thumbnail', default=None), 'http:')
thumbnails = [{'url': thumbnail}] if thumbnail else None
json_ld = self._search_json_ld(webpage, display_id, default={})
def extract_field(pattern, name):
return self._html_search_regex(pattern, webpage, name, default=None) if pattern else None
return {
'id': video_id,
'display_id': display_id,
'title': (extract_field(self._TITLE_REGEX, 'title')
or self._og_search_title(webpage, default=None)
or json_ld.get('title')),
'description': extract_field(self._DESCRIPTION_REGEX, 'description') or json_ld.get('description'),
'thumbnails': thumbnails,
'duration': parse_duration(
self._html_search_meta('duration', webpage, 'duration', default=None)) or json_ld.get('duration'),
'age_limit': self._rta_search(webpage) or 18,
'uploader': extract_field(self._UPLOADER_REGEX, 'uploader') or json_ld.get('uploader'),
'view_count': str_to_int(extract_field(self._VIEW_COUNT_REGEX, 'view count')),
'comment_count': str_to_int(extract_field(self._COMMENT_COUNT_REGEX, 'comment count')),
'average_rating': float_or_none(extract_field(self._AVERAGE_RATING_REGEX, 'average rating')),
'categories': list(map(str.strip, (extract_field(self._CATEGORIES_REGEX, 'categories') or '').split(','))),
'formats': formats,
}
class TNAFlixNetworkEmbedIE(TNAFlixNetworkBaseIE):
_VALID_URL = r'https?://player\.(?P<host>tnaflix|empflix)\.com/video/(?P<id>\d+)'
_EMBED_REGEX = [r'<iframe[^>]+?src=(["\'])(?P<url>(?:https?:)?//player\.(?:tna|emp)flix\.com/video/\d+)\1']
_TESTS = [{
'url': 'https://player.tnaflix.com/video/6538',
'info_dict': {
'id': '6538',
'display_id': '6538',
'ext': 'mp4',
'title': 'Educational xxx video (G Spot)',
'description': 'md5:b4fab8f88a8621c8fabd361a173fe5b8',
'thumbnail': r're:https?://.*\.jpg$',
'age_limit': 18,
'duration': 164,
'uploader': 'bobwhite39',
'categories': list,
},
'params': {
'skip_download': True,
},
}, {
'url': 'http://player.empflix.com/video/33051',
'only_matching': True,
}]
def _real_extract(self, url):
mobj = self._match_valid_url(url)
video_id, host = mobj.group('id', 'host')
return self.url_result(f'http://www.{host}.com/category/{video_id}/video{video_id}')
class TNAEMPFlixBaseIE(TNAFlixNetworkBaseIE):
_DESCRIPTION_REGEX = r'(?s)>Description:</[^>]+>(.+?)<'
_UPLOADER_REGEX = r'<span>by\s*<a[^>]+\bhref=["\']/profile/[^>]+>([^<]+)<'
_CATEGORIES_REGEX = r'(?s)<span[^>]*>Categories:</span>(.+?)</div>'
class TNAFlixIE(TNAEMPFlixBaseIE):
_VALID_URL = r'https?://(?:www\.)?(?P<host>tnaflix)\.com/[^/]+/(?P<display_id>[^/]+)/video(?P<id>\d+)'
_TITLE_REGEX = r'<title>(.+?) - (?:TNAFlix Porn Videos|TNAFlix\.com)</title>'
_TESTS = [{
# anonymous uploader, no categories
'url': 'http://www.tnaflix.com/porn-stars/Carmella-Decesare-striptease/video553878',
'md5': '7e569419fe6d69543d01e6be22f5f7c4',
'info_dict': {
'id': '553878',
'display_id': 'Carmella-Decesare-striptease',
'ext': 'mp4',
'title': 'Carmella Decesare - striptease',
'thumbnail': r're:https?://.*\.jpg$',
'duration': 91,
'age_limit': 18,
'categories': list,
}
}, {
# non-anonymous uploader, categories
'url': 'https://www.tnaflix.com/teen-porn/Educational-xxx-video/video6538',
'md5': 'add5a9fa7f4da53d3e9d0845ac58f20c',
'info_dict': {
'id': '6538',
'display_id': 'Educational-xxx-video',
'ext': 'mp4',
'title': 'Educational xxx video (G Spot)',
'description': 'md5:b4fab8f88a8621c8fabd361a173fe5b8',
'thumbnail': r're:https?://.*\.jpg$',
'duration': 164,
'age_limit': 18,
'uploader': 'bobwhite39',
'categories': list,
}
}, {
'url': 'https://www.tnaflix.com/amateur-porn/bunzHD-Ms.Donk/video358632',
'only_matching': True,
}]
class EMPFlixIE(TNAEMPFlixBaseIE):
_VALID_URL = r'https?://(?:www\.)?(?P<host>empflix)\.com/(?:videos/(?P<display_id>.+?)-|[^/]+/(?P<display_id_2>[^/]+)/video)(?P<id>[0-9]+)'
_TESTS = [{
'url': 'http://www.empflix.com/amateur-porn/Amateur-Finger-Fuck/video33051',
'md5': 'd761c7b26601bd14476cd9512f2654fc',
'info_dict': {
'id': '33051',
'display_id': 'Amateur-Finger-Fuck',
'ext': 'mp4',
'title': 'Amateur Finger Fuck',
'description': 'Amateur solo finger fucking.',
'thumbnail': r're:https?://.*\.jpg$',
'duration': 83,
'age_limit': 18,
'uploader': None,
'categories': list,
}
}, {
'url': 'http://www.empflix.com/videos/[AROMA][ARMD-718]-Aoi-Yoshino-Sawa-25826.html',
'only_matching': True,
}, {
'url': 'http://www.empflix.com/videos/Amateur-Finger-Fuck-33051.html',
'only_matching': True,
}]
class MovieFapIE(TNAFlixNetworkBaseIE):
_VALID_URL = r'https?://(?:www\.)?(?P<host>moviefap)\.com/videos/(?P<id>[0-9a-f]+)/(?P<display_id>[^/]+)\.html'
_VIEW_COUNT_REGEX = r'<br>Views\s*<strong>([\d,.]+)</strong>'
_COMMENT_COUNT_REGEX = r'<span[^>]+id="comCount"[^>]*>([\d,.]+)</span>'
_AVERAGE_RATING_REGEX = r'Current Rating\s*<br>\s*<strong>([\d.]+)</strong>'
_CATEGORIES_REGEX = r'(?s)<div[^>]+id="vid_info"[^>]*>\s*<div[^>]*>.+?</div>(.*?)<br>'
_TESTS = [{
# normal, multi-format video
'url': 'http://www.moviefap.com/videos/be9867c9416c19f54a4a/experienced-milf-amazing-handjob.html',
'md5': '26624b4e2523051b550067d547615906',
'info_dict': {
'id': 'be9867c9416c19f54a4a',
'display_id': 'experienced-milf-amazing-handjob',
'ext': 'mp4',
'title': 'Experienced MILF Amazing Handjob',
'description': 'Experienced MILF giving an Amazing Handjob',
'thumbnail': r're:https?://.*\.jpg$',
'age_limit': 18,
'uploader': 'darvinfred06',
'view_count': int,
'comment_count': int,
'average_rating': float,
'categories': ['Amateur', 'Masturbation', 'Mature', 'Flashing'],
}
}, {
# quirky single-format case where the extension is given as fid, but the video is really an flv
'url': 'http://www.moviefap.com/videos/e5da0d3edce5404418f5/jeune-couple-russe.html',
'md5': 'fa56683e291fc80635907168a743c9ad',
'info_dict': {
'id': 'e5da0d3edce5404418f5',
'display_id': 'jeune-couple-russe',
'ext': 'flv',
'title': 'Jeune Couple Russe',
'description': 'Amateur',
'thumbnail': r're:https?://.*\.jpg$',
'age_limit': 18,
'uploader': 'whiskeyjar',
'view_count': int,
'comment_count': int,
'average_rating': float,
'categories': ['Amateur', 'Teen'],
},
'skip': 'This video does not exist',
}]