X-Git-Url: https://git.rapsys.eu/youtubedl/blobdiff_plain/ca4d08063804fb264eb0ae9cc57894198f66e1fb..0fef6c070f2a2ec2c197d440b3e9bb8646c4f9ce:/youtube_dl/extractor/liveleak.py diff --git a/youtube_dl/extractor/liveleak.py b/youtube_dl/extractor/liveleak.py index 5ae57a7..22a067e 100644 --- a/youtube_dl/extractor/liveleak.py +++ b/youtube_dl/extractor/liveleak.py @@ -1,52 +1,185 @@ +from __future__ import unicode_literals + import re from .common import InfoExtractor -from ..utils import ( - ExtractorError, -) +from ..utils import int_or_none class LiveLeakIE(InfoExtractor): - - _VALID_URL = r'^(?:http://)?(?:\w+\.)?liveleak\.com/view\?(?:.*?)i=(?P[\w_]+)(?:.*)' - IE_NAME = u'liveleak' - _TEST = { - u'url': u'http://www.liveleak.com/view?i=757_1364311680', - u'file': u'757_1364311680.mp4', - u'md5': u'0813c2430bea7a46bf13acf3406992f4', - u'info_dict': { - u"description": u"extremely bad day for this guy..!", - u"uploader": u"ljfriel2", - u"title": u"Most unlucky car accident" + _VALID_URL = r'https?://(?:\w+\.)?liveleak\.com/view\?.*?\b[it]=(?P[\w_]+)' + _TESTS = [{ + 'url': 'http://www.liveleak.com/view?i=757_1364311680', + 'md5': '0813c2430bea7a46bf13acf3406992f4', + 'info_dict': { + 'id': '757_1364311680', + 'ext': 'mp4', + 'description': 'extremely bad day for this guy..!', + 'uploader': 'ljfriel2', + 'title': 'Most unlucky car accident', + 'thumbnail': r're:^https?://.*\.jpg$' + } + }, { + 'url': 'http://www.liveleak.com/view?i=f93_1390833151', + 'md5': 'd3f1367d14cc3c15bf24fbfbe04b9abf', + 'info_dict': { + 'id': 'f93_1390833151', + 'ext': 'mp4', + 'description': 'German Television Channel NDR does an exclusive interview with Edward Snowden.\r\nUploaded on LiveLeak cause German Television thinks the rest of the world isn\'t intereseted in Edward Snowden.', + 'uploader': 'ARD_Stinkt', + 'title': 'German Television does first Edward Snowden Interview (ENGLISH)', + 'thumbnail': r're:^https?://.*\.jpg$' + } + }, { + # Prochan embed + 'url': 'http://www.liveleak.com/view?i=4f7_1392687779', + 'md5': '42c6d97d54f1db107958760788c5f48f', + 'info_dict': { + 'id': '4f7_1392687779', + 'ext': 'mp4', + 'description': "The guy with the cigarette seems amazingly nonchalant about the whole thing... I really hope my friends' reactions would be a bit stronger.\r\n\r\nAction-go to 0:55.", + 'uploader': 'CapObveus', + 'title': 'Man is Fatally Struck by Reckless Car While Packing up a Moving Truck', + 'age_limit': 18, + }, + 'skip': 'Video is dead', + }, { + # Covers https://github.com/rg3/youtube-dl/pull/5983 + # Multiple resolutions + 'url': 'http://www.liveleak.com/view?i=801_1409392012', + 'md5': 'c3a449dbaca5c0d1825caecd52a57d7b', + 'info_dict': { + 'id': '801_1409392012', + 'ext': 'mp4', + 'description': 'Happened on 27.7.2014. \r\nAt 0:53 you can see people still swimming at near beach.', + 'uploader': 'bony333', + 'title': 'Crazy Hungarian tourist films close call waterspout in Croatia', + 'thumbnail': r're:^https?://.*\.jpg$' } - } + }, { + # Covers https://github.com/rg3/youtube-dl/pull/10664#issuecomment-247439521 + 'url': 'http://m.liveleak.com/view?i=763_1473349649', + 'add_ie': ['Youtube'], + 'info_dict': { + 'id': '763_1473349649', + 'ext': 'mp4', + 'title': 'Reporters and public officials ignore epidemic of black on asian violence in Sacramento | Colin Flaherty', + 'description': 'Colin being the warrior he is and showing the injustice Asians in Sacramento are being subjected to.', + 'uploader': 'Ziz', + 'upload_date': '20160908', + 'uploader_id': 'UCEbta5E_jqlZmEJsriTEtnw' + }, + 'params': { + 'skip_download': True, + }, + }, { + 'url': 'https://www.liveleak.com/view?i=677_1439397581', + 'info_dict': { + 'id': '677_1439397581', + 'title': 'Fuel Depot in China Explosion caught on video', + }, + 'playlist_count': 3, + }, { + 'url': 'https://www.liveleak.com/view?t=HvHi_1523016227', + 'only_matching': True, + }] + + @staticmethod + def _extract_urls(webpage): + return re.findall( + r']+src="(https?://(?:\w+\.)?liveleak\.com/ll_embed\?[^"]*[ift]=[\w_]+[^"]+)"', + webpage) def _real_extract(self, url): - mobj = re.match(self._VALID_URL, url) - if mobj is None: - raise ExtractorError(u'Invalid URL: %s' % url) + video_id = self._match_id(url) + webpage = self._download_webpage(url, video_id) + + video_title = self._og_search_title(webpage).replace('LiveLeak.com -', '').strip() + video_description = self._og_search_description(webpage) + video_uploader = self._html_search_regex( + r'By:.*?(\w+)', webpage, 'uploader', fatal=False) + age_limit = int_or_none(self._search_regex( + r'you confirm that you are ([0-9]+) years and over.', + webpage, 'age limit', default=None)) + video_thumbnail = self._og_search_thumbnail(webpage) - video_id = mobj.group('video_id') + entries = self._parse_html5_media_entries(url, webpage, video_id) + if not entries: + # Maybe an embed? + embed_url = self._search_regex( + r']+src="((?:https?:)?//(?:www\.)?(?:prochan|youtube)\.com/embed[^"]+)"', + webpage, 'embed URL') + return { + '_type': 'url_transparent', + 'url': embed_url, + 'id': video_id, + 'title': video_title, + 'description': video_description, + 'uploader': video_uploader, + 'age_limit': age_limit, + } - webpage = self._download_webpage(url, video_id) + for idx, info_dict in enumerate(entries): + formats = [] + for a_format in info_dict['formats']: + if not a_format.get('height'): + a_format['height'] = int_or_none(self._search_regex( + r'([0-9]+)p\.mp4', a_format['url'], 'height label', + default=None)) + formats.append(a_format) - video_url = self._search_regex(r'file: "(.*?)",', - webpage, u'video URL') + # Removing '.*.mp4' gives the raw video, which is essentially + # the same video without the LiveLeak logo at the top (see + # https://github.com/rg3/youtube-dl/pull/4768) + orig_url = re.sub(r'\.mp4\.[^.]+', '', a_format['url']) + if a_format['url'] != orig_url: + format_id = a_format.get('format_id') + formats.append({ + 'format_id': 'original' + ('-' + format_id if format_id else ''), + 'url': orig_url, + 'preference': 1, + }) + self._sort_formats(formats) + info_dict['formats'] = formats - video_title = self._og_search_title(webpage).replace('LiveLeak.com -', '').strip() + # Don't append entry ID for one-video pages to keep backward compatibility + if len(entries) > 1: + info_dict['id'] = '%s_%s' % (video_id, idx + 1) + else: + info_dict['id'] = video_id - video_description = self._og_search_description(webpage) + info_dict.update({ + 'title': video_title, + 'description': video_description, + 'uploader': video_uploader, + 'age_limit': age_limit, + 'thumbnail': video_thumbnail, + }) - video_uploader = self._html_search_regex(r'By:.*?(\w+)', - webpage, u'uploader', fatal=False) + return self.playlist_result(entries, video_id, video_title) - info = { - 'id': video_id, - 'url': video_url, - 'ext': 'mp4', - 'title': video_title, - 'description': video_description, - 'uploader': video_uploader - } - return [info] +class LiveLeakEmbedIE(InfoExtractor): + _VALID_URL = r'https?://(?:www\.)?liveleak\.com/ll_embed\?.*?\b(?P[ift])=(?P[\w_]+)' + + # See generic.py for actual test cases + _TESTS = [{ + 'url': 'https://www.liveleak.com/ll_embed?i=874_1459135191', + 'only_matching': True, + }, { + 'url': 'https://www.liveleak.com/ll_embed?f=ab065df993c1', + 'only_matching': True, + }] + + def _real_extract(self, url): + kind, video_id = re.match(self._VALID_URL, url).groups() + + if kind == 'f': + webpage = self._download_webpage(url, video_id) + liveleak_url = self._search_regex( + r'(?:logourl\s*:\s*|window\.open\()(?P[\'"])(?P%s)(?P=q1)' % LiveLeakIE._VALID_URL, + webpage, 'LiveLeak URL', group='url') + else: + liveleak_url = 'http://www.liveleak.com/view?%s=%s' % (kind, video_id) + + return self.url_result(liveleak_url, ie=LiveLeakIE.ie_key())