X-Git-Url: https://git.rapsys.eu/youtubedl/blobdiff_plain/ca4d08063804fb264eb0ae9cc57894198f66e1fb..8593410c28e395b68b410169356663541005c3ae:/youtube_dl/extractor/pornhd.py?ds=inline diff --git a/youtube_dl/extractor/pornhd.py b/youtube_dl/extractor/pornhd.py index 71abd50..c6052ac 100644 --- a/youtube_dl/extractor/pornhd.py +++ b/youtube_dl/extractor/pornhd.py @@ -1,38 +1,121 @@ +from __future__ import unicode_literals + import re from .common import InfoExtractor -from ..utils import compat_urllib_parse +from ..utils import ( + determine_ext, + ExtractorError, + int_or_none, + js_to_json, + merge_dicts, + urljoin, +) class PornHdIE(InfoExtractor): - _VALID_URL = r'(?:http://)?(?:www\.)?pornhd\.com/videos/(?P<video_id>[0-9]+)/(?P<video_title>.+)' - _TEST = { - u'url': u'http://www.pornhd.com/videos/1962/sierra-day-gets-his-cum-all-over-herself-hd-porn-video', - u'file': u'1962.flv', - u'md5': u'35272469887dca97abd30abecc6cdf75', - u'info_dict': { - u"title": u"sierra-day-gets-his-cum-all-over-herself-hd-porn-video", - u"age_limit": 18, - } - } + _VALID_URL = r'https?://(?:www\.)?pornhd\.com/(?:[a-z]{2,4}/)?videos/(?P<id>\d+)(?:/(?P<display_id>.+))?' + _TESTS = [{ + 'url': 'http://www.pornhd.com/videos/9864/selfie-restroom-masturbation-fun-with-chubby-cutie-hd-porn-video', + 'md5': '87f1540746c1d32ec7a2305c12b96b25', + 'info_dict': { + 'id': '9864', + 'display_id': 'selfie-restroom-masturbation-fun-with-chubby-cutie-hd-porn-video', + 'ext': 'mp4', + 'title': 'Restroom selfie masturbation', + 'description': 'md5:3748420395e03e31ac96857a8f125b2b', + 'thumbnail': r're:^https?://.*\.jpg', + 'view_count': int, + 'like_count': int, + 'age_limit': 18, + }, + 'skip': 'HTTP Error 404: Not Found', + }, { + 'url': 'http://www.pornhd.com/videos/1962/sierra-day-gets-his-cum-all-over-herself-hd-porn-video', + 'md5': '1b7b3a40b9d65a8e5b25f7ab9ee6d6de', + 'info_dict': { + 'id': '1962', + 'display_id': 'sierra-day-gets-his-cum-all-over-herself-hd-porn-video', + 'ext': 'mp4', + 'title': 'md5:98c6f8b2d9c229d0f0fde47f61a1a759', + 'description': 'md5:8ff0523848ac2b8f9b065ba781ccf294', + 'thumbnail': r're:^https?://.*\.jpg', + 'view_count': int, + 'like_count': int, + 'age_limit': 18, + }, + }] def _real_extract(self, url): mobj = re.match(self._VALID_URL, url) + video_id = mobj.group('id') + display_id = mobj.group('display_id') + + webpage = self._download_webpage(url, display_id or video_id) + + title = self._html_search_regex( + [r'<span[^>]+class=["\']video-name["\'][^>]*>([^<]+)', + r'<title>(.+?) - .*?[Pp]ornHD.*?</title>'], webpage, 'title') + + sources = self._parse_json(js_to_json(self._search_regex( + r"(?s)sources'?\s*[:=]\s*(\{.+?\})", + webpage, 'sources', default='{}')), video_id) + + info = {} + if not sources: + entries = self._parse_html5_media_entries(url, webpage, video_id) + if entries: + info = entries[0] + + if not sources and not info: + message = self._html_search_regex( + r'(?s)<(div|p)[^>]+class="no-video"[^>]*>(?P<value>.+?)</\1', + webpage, 'error message', group='value') + raise ExtractorError('%s said: %s' % (self.IE_NAME, message), expected=True) - video_id = mobj.group('video_id') - video_title = mobj.group('video_title') + formats = [] + for format_id, video_url in sources.items(): + video_url = urljoin(url, video_url) + if not video_url: + continue + height = int_or_none(self._search_regex( + r'^(\d+)[pP]', format_id, 'height', default=None)) + formats.append({ + 'url': video_url, + 'ext': determine_ext(video_url, 'mp4'), + 'format_id': format_id, + 'height': height, + }) + if formats: + info['formats'] = formats + self._sort_formats(info['formats']) - webpage = self._download_webpage(url, video_id) + description = self._html_search_regex( + (r'(?s)<section[^>]+class=["\']video-description[^>]+>(?P<value>.+?)</section>', + r'<(div|p)[^>]+class="description"[^>]*>(?P<value>[^<]+)</\1'), + webpage, 'description', fatal=False, + group='value') or self._html_search_meta( + 'description', webpage, default=None) or self._og_search_description(webpage) + view_count = int_or_none(self._html_search_regex( + r'(\d+) views\s*<', webpage, 'view count', fatal=False)) + thumbnail = self._search_regex( + r"poster'?\s*:\s*([\"'])(?P<url>(?:(?!\1).)+)\1", webpage, + 'thumbnail', default=None, group='url') - video_url = self._html_search_regex( - r'&hd=(http.+?)&', webpage, u'video URL') - video_url = compat_urllib_parse.unquote(video_url) - age_limit = 18 + like_count = int_or_none(self._search_regex( + (r'(\d+)</span>\s*likes', + r'(\d+)\s*</11[^>]+>(?: |\s)*\blikes', + r'class=["\']save-count["\'][^>]*>\s*(\d+)'), + webpage, 'like count', fatal=False)) - return { + return merge_dicts(info, { 'id': video_id, - 'url': video_url, - 'ext': 'flv', - 'title': video_title, - 'age_limit': age_limit, - } + 'display_id': display_id, + 'title': title, + 'description': description, + 'thumbnail': thumbnail, + 'view_count': view_count, + 'like_count': like_count, + 'formats': formats, + 'age_limit': 18, + })