]> Raphaƫl G. Git Repositories - youtubedl/blobdiff - youtube_dl/extractor/xhamster.py
debian/copyright: Update my copyright years.
[youtubedl] / youtube_dl / extractor / xhamster.py
index 0f1feeffd777ccc152b1ea1c4bb15dc00dc6b494..a5b94d2794166d452464b728ec40b2b258459c64 100644 (file)
+from __future__ import unicode_literals
+
+import itertools
 import re
 
 from .common import InfoExtractor
+from ..compat import compat_str
 from ..utils import (
-    compat_urllib_parse,
-
+    clean_html,
+    determine_ext,
+    dict_get,
+    extract_attributes,
     ExtractorError,
+    int_or_none,
+    parse_duration,
+    try_get,
+    unified_strdate,
+    url_or_none,
 )
 
 
 class XHamsterIE(InfoExtractor):
-    """Information Extractor for xHamster"""
-    _VALID_URL = r'(?:http://)?(?:www.)?xhamster\.com/movies/(?P<id>[0-9]+)/.*\.html'
+    _DOMAINS = r'(?:xhamster\.(?:com|one|desi)|xhms\.pro|xhamster[27]\.com)'
+    _VALID_URL = r'''(?x)
+                    https?://
+                        (?:.+?\.)?%s/
+                        (?:
+                            movies/(?P<id>\d+)/(?P<display_id>[^/]*)\.html|
+                            videos/(?P<display_id_2>[^/]*)-(?P<id_2>\d+)
+                        )
+                    ''' % _DOMAINS
+    _TESTS = [{
+        'url': 'https://xhamster.com/videos/femaleagent-shy-beauty-takes-the-bait-1509445',
+        'md5': '98b4687efb1ffd331c4197854dc09e8f',
+        'info_dict': {
+            'id': '1509445',
+            'display_id': 'femaleagent-shy-beauty-takes-the-bait',
+            'ext': 'mp4',
+            'title': 'FemaleAgent Shy beauty takes the bait',
+            'timestamp': 1350194821,
+            'upload_date': '20121014',
+            'uploader': 'Ruseful2011',
+            'duration': 893,
+            'age_limit': 18,
+        },
+    }, {
+        'url': 'https://xhamster.com/videos/britney-spears-sexy-booty-2221348?hd=',
+        'info_dict': {
+            'id': '2221348',
+            'display_id': 'britney-spears-sexy-booty',
+            'ext': 'mp4',
+            'title': 'Britney Spears  Sexy Booty',
+            'timestamp': 1379123460,
+            'upload_date': '20130914',
+            'uploader': 'jojo747400',
+            'duration': 200,
+            'age_limit': 18,
+        },
+        'params': {
+            'skip_download': True,
+        },
+    }, {
+        # empty seo, unavailable via new URL schema
+        'url': 'http://xhamster.com/movies/5667973/.html',
+        'info_dict': {
+            'id': '5667973',
+            'ext': 'mp4',
+            'title': '....',
+            'timestamp': 1454948101,
+            'upload_date': '20160208',
+            'uploader': 'parejafree',
+            'duration': 72,
+            'age_limit': 18,
+        },
+        'params': {
+            'skip_download': True,
+        },
+    }, {
+        # mobile site
+        'url': 'https://m.xhamster.com/videos/cute-teen-jacqueline-solo-masturbation-8559111',
+        'only_matching': True,
+    }, {
+        'url': 'https://xhamster.com/movies/2272726/amber_slayed_by_the_knight.html',
+        'only_matching': True,
+    }, {
+        # This video is visible for marcoalfa123456's friends only
+        'url': 'https://it.xhamster.com/movies/7263980/la_mia_vicina.html',
+        'only_matching': True,
+    }, {
+        # new URL schema
+        'url': 'https://pt.xhamster.com/videos/euro-pedal-pumping-7937821',
+        'only_matching': True,
+    }, {
+        'url': 'https://xhamster.one/videos/femaleagent-shy-beauty-takes-the-bait-1509445',
+        'only_matching': True,
+    }, {
+        'url': 'https://xhamster.desi/videos/femaleagent-shy-beauty-takes-the-bait-1509445',
+        'only_matching': True,
+    }, {
+        'url': 'https://xhamster2.com/videos/femaleagent-shy-beauty-takes-the-bait-1509445',
+        'only_matching': True,
+    }, {
+        'url': 'http://xhamster.com/movies/1509445/femaleagent_shy_beauty_takes_the_bait.html',
+        'only_matching': True,
+    }, {
+        'url': 'http://xhamster.com/movies/2221348/britney_spears_sexy_booty.html?hd',
+        'only_matching': True,
+    }]
+
+    def _real_extract(self, url):
+        mobj = re.match(self._VALID_URL, url)
+        video_id = mobj.group('id') or mobj.group('id_2')
+        display_id = mobj.group('display_id') or mobj.group('display_id_2')
+
+        desktop_url = re.sub(r'^(https?://(?:.+?\.)?)m\.', r'\1', url)
+        webpage = self._download_webpage(desktop_url, video_id)
+
+        error = self._html_search_regex(
+            r'<div[^>]+id=["\']videoClosed["\'][^>]*>(.+?)</div>',
+            webpage, 'error', default=None)
+        if error:
+            raise ExtractorError(error, expected=True)
+
+        age_limit = self._rta_search(webpage)
+
+        def get_height(s):
+            return int_or_none(self._search_regex(
+                r'^(\d+)[pP]', s, 'height', default=None))
+
+        initials = self._parse_json(
+            self._search_regex(
+                r'window\.initials\s*=\s*({.+?})\s*;\s*\n', webpage, 'initials',
+                default='{}'),
+            video_id, fatal=False)
+        if initials:
+            video = initials['videoModel']
+            title = video['title']
+            formats = []
+            for format_id, formats_dict in video['sources'].items():
+                if not isinstance(formats_dict, dict):
+                    continue
+                for quality, format_item in formats_dict.items():
+                    if format_id == 'download':
+                        # Download link takes some time to be generated,
+                        # skipping for now
+                        continue
+                        if not isinstance(format_item, dict):
+                            continue
+                        format_url = format_item.get('link')
+                        filesize = int_or_none(
+                            format_item.get('size'), invscale=1000000)
+                    else:
+                        format_url = format_item
+                        filesize = None
+                    format_url = url_or_none(format_url)
+                    if not format_url:
+                        continue
+                    formats.append({
+                        'format_id': '%s-%s' % (format_id, quality),
+                        'url': format_url,
+                        'ext': determine_ext(format_url, 'mp4'),
+                        'height': get_height(quality),
+                        'filesize': filesize,
+                    })
+            self._sort_formats(formats)
+
+            categories_list = video.get('categories')
+            if isinstance(categories_list, list):
+                categories = []
+                for c in categories_list:
+                    if not isinstance(c, dict):
+                        continue
+                    c_name = c.get('name')
+                    if isinstance(c_name, compat_str):
+                        categories.append(c_name)
+            else:
+                categories = None
+
+            return {
+                'id': video_id,
+                'display_id': display_id,
+                'title': title,
+                'description': video.get('description'),
+                'timestamp': int_or_none(video.get('created')),
+                'uploader': try_get(
+                    video, lambda x: x['author']['name'], compat_str),
+                'thumbnail': video.get('thumbURL'),
+                'duration': int_or_none(video.get('duration')),
+                'view_count': int_or_none(video.get('views')),
+                'like_count': int_or_none(try_get(
+                    video, lambda x: x['rating']['likes'], int)),
+                'dislike_count': int_or_none(try_get(
+                    video, lambda x: x['rating']['dislikes'], int)),
+                'comment_count': int_or_none(video.get('views')),
+                'age_limit': age_limit,
+                'categories': categories,
+                'formats': formats,
+            }
+
+        # Old layout fallback
+
+        title = self._html_search_regex(
+            [r'<h1[^>]*>([^<]+)</h1>',
+             r'<meta[^>]+itemprop=".*?caption.*?"[^>]+content="(.+?)"',
+             r'<title[^>]*>(.+?)(?:,\s*[^,]*?\s*Porn\s*[^,]*?:\s*xHamster[^<]*| - xHamster\.com)</title>'],
+            webpage, 'title')
+
+        formats = []
+        format_urls = set()
+
+        sources = self._parse_json(
+            self._search_regex(
+                r'sources\s*:\s*({.+?})\s*,?\s*\n', webpage, 'sources',
+                default='{}'),
+            video_id, fatal=False)
+        for format_id, format_url in sources.items():
+            format_url = url_or_none(format_url)
+            if not format_url:
+                continue
+            if format_url in format_urls:
+                continue
+            format_urls.add(format_url)
+            formats.append({
+                'format_id': format_id,
+                'url': format_url,
+                'height': get_height(format_id),
+            })
+
+        video_url = self._search_regex(
+            [r'''file\s*:\s*(?P<q>["'])(?P<mp4>.+?)(?P=q)''',
+             r'''<a\s+href=(?P<q>["'])(?P<mp4>.+?)(?P=q)\s+class=["']mp4Thumb''',
+             r'''<video[^>]+file=(?P<q>["'])(?P<mp4>.+?)(?P=q)[^>]*>'''],
+            webpage, 'video url', group='mp4', default=None)
+        if video_url and video_url not in format_urls:
+            formats.append({
+                'url': video_url,
+            })
+
+        self._sort_formats(formats)
+
+        # Only a few videos have an description
+        mobj = re.search(r'<span>Description: </span>([^<]+)', webpage)
+        description = mobj.group(1) if mobj else None
+
+        upload_date = unified_strdate(self._search_regex(
+            r'hint=["\'](\d{4}-\d{2}-\d{2}) \d{2}:\d{2}:\d{2} [A-Z]{3,4}',
+            webpage, 'upload date', fatal=False))
+
+        uploader = self._html_search_regex(
+            r'<span[^>]+itemprop=["\']author[^>]+><a[^>]+><span[^>]+>([^<]+)',
+            webpage, 'uploader', default='anonymous')
+
+        thumbnail = self._search_regex(
+            [r'''["']thumbUrl["']\s*:\s*(?P<q>["'])(?P<thumbnail>.+?)(?P=q)''',
+             r'''<video[^>]+"poster"=(?P<q>["'])(?P<thumbnail>.+?)(?P=q)[^>]*>'''],
+            webpage, 'thumbnail', fatal=False, group='thumbnail')
+
+        duration = parse_duration(self._search_regex(
+            [r'<[^<]+\bitemprop=["\']duration["\'][^<]+\bcontent=["\'](.+?)["\']',
+             r'Runtime:\s*</span>\s*([\d:]+)'], webpage,
+            'duration', fatal=False))
+
+        view_count = int_or_none(self._search_regex(
+            r'content=["\']User(?:View|Play)s:(\d+)',
+            webpage, 'view count', fatal=False))
+
+        mobj = re.search(r'hint=[\'"](?P<likecount>\d+) Likes / (?P<dislikecount>\d+) Dislikes', webpage)
+        (like_count, dislike_count) = (mobj.group('likecount'), mobj.group('dislikecount')) if mobj else (None, None)
+
+        mobj = re.search(r'</label>Comments \((?P<commentcount>\d+)\)</div>', webpage)
+        comment_count = mobj.group('commentcount') if mobj else 0
+
+        categories_html = self._search_regex(
+            r'(?s)<table.+?(<span>Categories:.+?)</table>', webpage,
+            'categories', default=None)
+        categories = [clean_html(category) for category in re.findall(
+            r'<a[^>]+>(.+?)</a>', categories_html)] if categories_html else None
+
+        return {
+            'id': video_id,
+            'display_id': display_id,
+            'title': title,
+            'description': description,
+            'upload_date': upload_date,
+            'uploader': uploader,
+            'thumbnail': thumbnail,
+            'duration': duration,
+            'view_count': view_count,
+            'like_count': int_or_none(like_count),
+            'dislike_count': int_or_none(dislike_count),
+            'comment_count': int_or_none(comment_count),
+            'age_limit': age_limit,
+            'categories': categories,
+            'formats': formats,
+        }
+
+
+class XHamsterEmbedIE(InfoExtractor):
+    _VALID_URL = r'https?://(?:.+?\.)?%s/xembed\.php\?video=(?P<id>\d+)' % XHamsterIE._DOMAINS
     _TEST = {
-        u'url': u'http://xhamster.com/movies/1509445/femaleagent_shy_beauty_takes_the_bait.html',
-        u'file': u'1509445.flv',
-        u'md5': u'9f48e0e8d58e3076bb236ff412ab62fa',
-        u'info_dict': {
-            u"upload_date": u"20121014", 
-            u"uploader_id": u"Ruseful2011", 
-            u"title": u"FemaleAgent Shy beauty takes the bait"
+        'url': 'http://xhamster.com/xembed.php?video=3328539',
+        'info_dict': {
+            'id': '3328539',
+            'ext': 'mp4',
+            'title': 'Pen Masturbation',
+            'timestamp': 1406581861,
+            'upload_date': '20140728',
+            'uploader': 'ManyakisArt',
+            'duration': 5,
+            'age_limit': 18,
         }
     }
 
-    def _real_extract(self,url):
-        mobj = re.match(self._VALID_URL, url)
+    @staticmethod
+    def _extract_urls(webpage):
+        return [url for _, url in re.findall(
+            r'<iframe[^>]+?src=(["\'])(?P<url>(?:https?:)?//(?:www\.)?xhamster\.com/xembed\.php\?video=\d+)\1',
+            webpage)]
+
+    def _real_extract(self, url):
+        video_id = self._match_id(url)
+
+        webpage = self._download_webpage(url, video_id)
+
+        video_url = self._search_regex(
+            r'href="(https?://xhamster\.com/(?:movies/{0}/[^"]*\.html|videos/[^/]*-{0})[^"]*)"'.format(video_id),
+            webpage, 'xhamster url', default=None)
+
+        if not video_url:
+            vars = self._parse_json(
+                self._search_regex(r'vars\s*:\s*({.+?})\s*,\s*\n', webpage, 'vars'),
+                video_id)
+            video_url = dict_get(vars, ('downloadLink', 'homepageLink', 'commentsLink', 'shareUrl'))
+
+        return self.url_result(video_url, 'XHamster')
+
+
+class XHamsterUserIE(InfoExtractor):
+    _VALID_URL = r'https?://(?:.+?\.)?%s/users/(?P<id>[^/?#&]+)' % XHamsterIE._DOMAINS
+    _TESTS = [{
+        # Paginated user profile
+        'url': 'https://xhamster.com/users/netvideogirls/videos',
+        'info_dict': {
+            'id': 'netvideogirls',
+        },
+        'playlist_mincount': 267,
+    }, {
+        # Non-paginated user profile
+        'url': 'https://xhamster.com/users/firatkaan/videos',
+        'info_dict': {
+            'id': 'firatkaan',
+        },
+        'playlist_mincount': 1,
+    }]
+
+    def _entries(self, user_id):
+        next_page_url = 'https://xhamster.com/users/%s/videos/1' % user_id
+        for pagenum in itertools.count(1):
+            page = self._download_webpage(
+                next_page_url, user_id, 'Downloading page %s' % pagenum)
+            for video_tag in re.findall(
+                    r'(<a[^>]+class=["\'].*?\bvideo-thumb__image-container[^>]+>)',
+                    page):
+                video = extract_attributes(video_tag)
+                video_url = url_or_none(video.get('href'))
+                if not video_url or not XHamsterIE.suitable(video_url):
+                    continue
+                video_id = XHamsterIE._match_id(video_url)
+                yield self.url_result(
+                    video_url, ie=XHamsterIE.ie_key(), video_id=video_id)
+            mobj = re.search(r'<a[^>]+data-page=["\']next[^>]+>', page)
+            if not mobj:
+                break
+            next_page = extract_attributes(mobj.group(0))
+            next_page_url = url_or_none(next_page.get('href'))
+            if not next_page_url:
+                break
 
-        video_id = mobj.group('id')
-        mrss_url = 'http://xhamster.com/movies/%s/.html' % video_id
-        webpage = self._download_webpage(mrss_url, video_id)
-
-        mobj = re.search(r'\'srv\': \'(?P<server>[^\']*)\',\s*\'file\': \'(?P<file>[^\']+)\',', webpage)
-        if mobj is None:
-            raise ExtractorError(u'Unable to extract media URL')
-        if len(mobj.group('server')) == 0:
-            video_url = compat_urllib_parse.unquote(mobj.group('file'))
-        else:
-            video_url = mobj.group('server')+'/key='+mobj.group('file')
-        video_extension = video_url.split('.')[-1]
-
-        video_title = self._html_search_regex(r'<title>(?P<title>.+?) - xHamster\.com</title>',
-            webpage, u'title')
-
-        # Can't see the description anywhere in the UI
-        # video_description = self._html_search_regex(r'<span>Description: </span>(?P<description>[^<]+)',
-        #     webpage, u'description', fatal=False)
-        # if video_description: video_description = unescapeHTML(video_description)
-
-        mobj = re.search(r'hint=\'(?P<upload_date_Y>[0-9]{4})-(?P<upload_date_m>[0-9]{2})-(?P<upload_date_d>[0-9]{2}) [0-9]{2}:[0-9]{2}:[0-9]{2} [A-Z]{3,4}\'', webpage)
-        if mobj:
-            video_upload_date = mobj.group('upload_date_Y')+mobj.group('upload_date_m')+mobj.group('upload_date_d')
-        else:
-            video_upload_date = None
-            self._downloader.report_warning(u'Unable to extract upload date')
-
-        video_uploader_id = self._html_search_regex(r'<a href=\'/user/[^>]+>(?P<uploader_id>[^<]+)',
-            webpage, u'uploader id', default=u'anonymous')
-
-        video_thumbnail = self._search_regex(r'\'image\':\'(?P<thumbnail>[^\']+)\'',
-            webpage, u'thumbnail', fatal=False)
-
-        return [{
-            'id':       video_id,
-            'url':      video_url,
-            'ext':      video_extension,
-            'title':    video_title,
-            # 'description': video_description,
-            'upload_date': video_upload_date,
-            'uploader_id': video_uploader_id,
-            'thumbnail': video_thumbnail
-        }]
+    def _real_extract(self, url):
+        user_id = self._match_id(url)
+        return self.playlist_result(self._entries(user_id), user_id)