from __future__ import unicode_literals import re from .common import InfoExtractor from ..utils import ( int_or_none, unified_strdate, xpath_text, determine_ext, ExtractorError, ) class PhoenixIE(InfoExtractor): IE_NAME = 'phoenix.de' _VALID_URL = r'''(?x)https?://(?:www\.)?phoenix\.de/content/ (?: phoenix/die_sendungen/(?:[^/]+/)? )? (?P[0-9]+)''' _TESTS = [ { 'url': 'http://www.phoenix.de/content/884301', 'md5': 'ed249f045256150c92e72dbb70eadec6', 'info_dict': { 'id': '884301', 'ext': 'mp4', 'title': 'Michael Krons mit Hans-Werner Sinn', 'description': 'Im Dialog - Sa. 25.10.14, 00.00 - 00.35 Uhr', 'upload_date': '20141025', 'uploader': 'Im Dialog', } }, { 'url': 'http://www.phoenix.de/content/phoenix/die_sendungen/869815', 'only_matching': True, }, { 'url': 'http://www.phoenix.de/content/phoenix/die_sendungen/diskussionen/928234', 'only_matching': True, }, ] def extract_from_xml_url(self, video_id, xml_url): doc = self._download_xml( xml_url, video_id, note='Downloading video info', errnote='Failed to download video info') status_code = xpath_text(doc, './status/statuscode') if status_code and status_code != 'ok': if status_code == 'notVisibleAnymore': message = 'Video %s is not available' % video_id else: message = '%s returned error: %s' % (self.IE_NAME, status_code) raise ExtractorError(message, expected=True) title = xpath_text(doc, './/information/title', 'title', True) urls = [] formats = [] for fnode in doc.findall('.//formitaeten/formitaet'): video_url = xpath_text(fnode, 'url') if not video_url or video_url in urls: continue urls.append(video_url) is_available = 'http://www.metafilegenerator' not in video_url geoloced = 'static_geoloced_online' in video_url if not is_available or geoloced: continue format_id = fnode.attrib['basetype'] format_m = re.match(r'''(?x) (?P[^_]+)_(?P[^_]+)_(?P[^_]+)_ (?P[^_]+)_(?P[^_]+)_(?P[^_]+) ''', format_id) ext = determine_ext(video_url, None) or format_m.group('container') if ext == 'meta': continue elif ext == 'smil': formats.extend(self._extract_smil_formats( video_url, video_id, fatal=False)) elif ext == 'm3u8': # the certificates are misconfigured (see # https://github.com/ytdl-org/youtube-dl/issues/8665) if video_url.startswith('https://'): continue formats.extend(self._extract_m3u8_formats( video_url, video_id, 'mp4', 'm3u8_native', m3u8_id=format_id, fatal=False)) elif ext == 'f4m': formats.extend(self._extract_f4m_formats( video_url, video_id, f4m_id=format_id, fatal=False)) else: quality = xpath_text(fnode, './quality') if quality: format_id += '-' + quality abr = int_or_none(xpath_text(fnode, './audioBitrate'), 1000) vbr = int_or_none(xpath_text(fnode, './videoBitrate'), 1000) tbr = int_or_none(self._search_regex( r'_(\d+)k', video_url, 'bitrate', None)) if tbr and vbr and not abr: abr = tbr - vbr formats.append({ 'format_id': format_id, 'url': video_url, 'ext': ext, 'acodec': format_m.group('acodec'), 'vcodec': format_m.group('vcodec'), 'abr': abr, 'vbr': vbr, 'tbr': tbr, 'width': int_or_none(xpath_text(fnode, './width')), 'height': int_or_none(xpath_text(fnode, './height')), 'filesize': int_or_none(xpath_text(fnode, './filesize')), 'protocol': format_m.group('proto').lower(), }) geolocation = xpath_text(doc, './/details/geolocation') if not formats and geolocation and geolocation != 'none': self.raise_geo_restricted(countries=self._GEO_COUNTRIES) self._sort_formats(formats) thumbnails = [] for node in doc.findall('.//teaserimages/teaserimage'): thumbnail_url = node.text if not thumbnail_url: continue thumbnail = { 'url': thumbnail_url, } thumbnail_key = node.get('key') if thumbnail_key: m = re.match('^([0-9]+)x([0-9]+)$', thumbnail_key) if m: thumbnail['width'] = int(m.group(1)) thumbnail['height'] = int(m.group(2)) thumbnails.append(thumbnail) upload_date = unified_strdate(xpath_text(doc, './/details/airtime')) return { 'id': video_id, 'title': title, 'description': xpath_text(doc, './/information/detail'), 'duration': int_or_none(xpath_text(doc, './/details/lengthSec')), 'thumbnails': thumbnails, 'uploader': xpath_text(doc, './/details/originChannelTitle'), 'uploader_id': xpath_text(doc, './/details/originChannelId'), 'upload_date': upload_date, 'formats': formats, } def _real_extract(self, url): video_id = self._match_id(url) webpage = self._download_webpage(url, video_id) internal_id = self._search_regex( r'