X-Git-Url: http://git.bitcoin.ninja/index.cgi?a=blobdiff_plain;ds=sidebyside;f=youtube_dl%2Fextractor%2Fpornhub.py;h=e032817f2c18634561dda71f786782373c0b38f8;hb=ddbb4c5c3ea7a64e09834d9628e7e34e03675289;hp=77182bf07f33a58216a07e84769779149e60833a;hpb=eaaaaec042f8e6afa8f8ec6a2a8b137943f802df;p=youtube-dl diff --git a/youtube_dl/extractor/pornhub.py b/youtube_dl/extractor/pornhub.py index 77182bf07..e032817f2 100644 --- a/youtube_dl/extractor/pornhub.py +++ b/youtube_dl/extractor/pornhub.py @@ -1,31 +1,43 @@ # coding: utf-8 from __future__ import unicode_literals +import functools import itertools -import os +import operator +# import os import re from .common import InfoExtractor from ..compat import ( compat_HTTPError, - compat_urllib_parse_unquote, - compat_urllib_parse_unquote_plus, - compat_urllib_parse_urlparse, + # compat_urllib_parse_unquote, + # compat_urllib_parse_unquote_plus, + # compat_urllib_parse_urlparse, ) from ..utils import ( ExtractorError, int_or_none, + js_to_json, orderedSet, - sanitized_Request, + # sanitized_Request, + remove_quotes, str_to_int, ) -from ..aes import ( - aes_decrypt_text -) +# from ..aes import ( +# aes_decrypt_text +# ) class PornHubIE(InfoExtractor): - _VALID_URL = r'https?://(?:[a-z]+\.)?pornhub\.com/(?:view_video\.php\?viewkey=|embed/)(?P[0-9a-z]+)' + IE_DESC = 'PornHub and Thumbzilla' + _VALID_URL = r'''(?x) + https?:// + (?: + (?:[a-z]+\.)?pornhub\.com/(?:(?:view_video\.php|video/show)\?viewkey=|embed/)| + (?:www\.)?thumbzilla\.com/video/ + ) + (?P[\da-z]+) + ''' _TESTS = [{ 'url': 'http://www.pornhub.com/view_video.php?viewkey=648719015', 'md5': '1e19b41231a02eba417839222ac9d58e', @@ -40,6 +52,8 @@ class PornHubIE(InfoExtractor): 'dislike_count': int, 'comment_count': int, 'age_limit': 18, + 'tags': list, + 'categories': list, }, }, { # non-ASCII title @@ -55,6 +69,8 @@ class PornHubIE(InfoExtractor): 'dislike_count': int, 'comment_count': int, 'age_limit': 18, + 'tags': list, + 'categories': list, }, 'params': { 'skip_download': True, @@ -74,14 +90,23 @@ class PornHubIE(InfoExtractor): # removed by uploader 'url': 'http://www.pornhub.com/view_video.php?viewkey=ph572716d15a111', 'only_matching': True, + }, { + # private video + 'url': 'http://www.pornhub.com/view_video.php?viewkey=ph56fd731fce6b7', + 'only_matching': True, + }, { + 'url': 'https://www.thumbzilla.com/video/ph56c6114abd99a/horny-girlfriend-sex', + 'only_matching': True, + }, { + 'url': 'http://www.pornhub.com/video/show?viewkey=648719015', + 'only_matching': True, }] - @classmethod - def _extract_url(cls, webpage): - mobj = re.search( - r']+?src=(["\'])(?P(?:https?:)?//(?:www\.)?pornhub\.com/embed/\d+)\1', webpage) - if mobj: - return mobj.group('url') + @staticmethod + def _extract_urls(webpage): + return re.findall( + r']+?src=["\'](?P(?:https?:)?//(?:www\.)?pornhub\.com/embed/[\da-z]+)', + webpage) def _extract_count(self, pattern, webpage, name): return str_to_int(self._search_regex( @@ -90,13 +115,17 @@ class PornHubIE(InfoExtractor): def _real_extract(self, url): video_id = self._match_id(url) - req = sanitized_Request( - 'http://www.pornhub.com/view_video.php?viewkey=%s' % video_id) - req.add_header('Cookie', 'age_verified=1') - webpage = self._download_webpage(req, video_id) + def dl_webpage(platform): + return self._download_webpage( + 'http://www.pornhub.com/view_video.php?viewkey=%s' % video_id, + video_id, headers={ + 'Cookie': 'age_verified=1; platform=%s' % platform, + }) + + webpage = dl_webpage('pc') error_msg = self._html_search_regex( - r'(?s)]+class=(["\']).*?\bremoved\b.*?\1[^>]*>(?P.+?)', + r'(?s)]+class=(["\'])(?:(?!\1).)*\b(?:removed|userMessageSection)\b(?:(?!\1).)*\1[^>]*>(?P.+?)', webpage, 'error message', default=None, group='error') if error_msg: error_msg = re.sub(r'\s+', ' ', error_msg) @@ -104,10 +133,42 @@ class PornHubIE(InfoExtractor): 'PornHub said: %s' % error_msg, expected=True, video_id=video_id) + tv_webpage = dl_webpage('tv') + + assignments = self._search_regex( + r'(var.+?mediastring.+?)', tv_webpage, + 'encoded url').split(';') + + js_vars = {} + + def parse_js_value(inp): + inp = re.sub(r'/\*(?:(?!\*/).)*?\*/', '', inp) + if '+' in inp: + inps = inp.split('+') + return functools.reduce( + operator.concat, map(parse_js_value, inps)) + inp = inp.strip() + if inp in js_vars: + return js_vars[inp] + return remove_quotes(inp) + + for assn in assignments: + assn = assn.strip() + if not assn: + continue + assn = re.sub(r'var\s+', '', assn) + vname, value = assn.split('=', 1) + js_vars[vname] = parse_js_value(value) + + video_url = js_vars['mediastring'] + + title = self._search_regex( + r'

([^>]+)

', tv_webpage, 'title', default=None) + # video_title from flashvars contains whitespace instead of non-ASCII (see # http://www.pornhub.com/view_video.php?viewkey=1331683002), not relying # on that anymore. - title = self._html_search_meta( + title = title or self._html_search_meta( 'twitter:title', webpage, default=None) or self._search_regex( (r']+class=["\']title["\'][^>]*>(?P[^<]+)', r'<div[^>]+data-video-title=(["\'])(?P<title>.+?)\1', @@ -137,39 +198,18 @@ class PornHubIE(InfoExtractor): comment_count = self._extract_count( r'All Comments\s*<span>\(([\d,.]+)\)', webpage, 'comment') - video_urls = list(map(compat_urllib_parse_unquote, re.findall(r"player_quality_[0-9]{3}p\s*=\s*'([^']+)'", webpage))) - if webpage.find('"encrypted":true') != -1: - password = compat_urllib_parse_unquote_plus( - self._search_regex(r'"video_title":"([^"]+)', webpage, 'password')) - video_urls = list(map(lambda s: aes_decrypt_text(s, password, 32).decode('utf-8'), video_urls)) - - formats = [] - for video_url in video_urls: - path = compat_urllib_parse_urlparse(video_url).path - extension = os.path.splitext(path)[1][1:] - format = path.split('/')[5].split('_')[:2] - format = '-'.join(format) - - m = re.match(r'^(?P<height>[0-9]+)[pP]-(?P<tbr>[0-9]+)[kK]$', format) - if m is None: - height = None - tbr = None - else: - height = int(m.group('height')) - tbr = int(m.group('tbr')) - - formats.append({ - 'url': video_url, - 'ext': extension, - 'format': format, - 'format_id': format, - 'tbr': tbr, - 'height': height, - }) - self._sort_formats(formats) + page_params = self._parse_json(self._search_regex( + r'page_params\.zoneDetails\[([\'"])[^\'"]+\1\]\s*=\s*(?P<data>{[^}]+})', + webpage, 'page parameters', group='data', default='{}'), + video_id, transform_source=js_to_json, fatal=False) + tags = categories = None + if page_params: + tags = page_params.get('tags', '').split(',') + categories = page_params.get('categories', '').split(',') return { 'id': video_id, + 'url': video_url, 'uploader': video_uploader, 'title': title, 'thumbnail': thumbnail, @@ -178,8 +218,10 @@ class PornHubIE(InfoExtractor): 'like_count': like_count, 'dislike_count': dislike_count, 'comment_count': comment_count, - 'formats': formats, + # 'formats': formats, 'age_limit': 18, + 'tags': tags, + 'categories': categories, } @@ -199,26 +241,36 @@ class PornHubPlaylistBaseIE(InfoExtractor): webpage = self._download_webpage(url, playlist_id) - entries = self._extract_entries(webpage) + # Only process container div with main playlist content skipping + # drop-down menu that uses similar pattern for videos (see + # https://github.com/rg3/youtube-dl/issues/11594). + container = self._search_regex( + r'(?s)(<div[^>]+class=["\']container.+)', webpage, + 'container', default=webpage) + + entries = self._extract_entries(container) playlist = self._parse_json( self._search_regex( - r'playlistObject\s*=\s*({.+?});', webpage, 'playlist'), - playlist_id) + r'(?:playlistObject|PLAYLIST_VIEW)\s*=\s*({.+?});', webpage, + 'playlist', default='{}'), + playlist_id, fatal=False) + title = playlist.get('title') or self._search_regex( + r'>Videos\s+in\s+(.+?)\s+[Pp]laylist<', webpage, 'title', fatal=False) return self.playlist_result( - entries, playlist_id, playlist.get('title'), playlist.get('description')) + entries, playlist_id, title, playlist.get('description')) class PornHubPlaylistIE(PornHubPlaylistBaseIE): _VALID_URL = r'https?://(?:www\.)?pornhub\.com/playlist/(?P<id>\d+)' _TESTS = [{ - 'url': 'http://www.pornhub.com/playlist/6201671', + 'url': 'http://www.pornhub.com/playlist/4667351', 'info_dict': { - 'id': '6201671', - 'title': 'P0p4', + 'id': '4667351', + 'title': 'Nataly Hot', }, - 'playlist_mincount': 35, + 'playlist_mincount': 2, }] @@ -247,6 +299,7 @@ class PornHubUserVideosIE(PornHubPlaylistBaseIE): except ExtractorError as e: if isinstance(e.cause, compat_HTTPError) and e.cause.code == 404: break + raise page_entries = self._extract_entries(webpage) if not page_entries: break