From 8365f7ba06eae3ecb68e734ceac2deafeb2afd6b Mon Sep 17 00:00:00 2001 From: Paul Hartmann Date: Fri, 11 Nov 2016 22:41:37 +0100 Subject: [PATCH] [zdf] added 2nd method based on simple URL as fallback (idea: flak3) keep complicated method as default as it has more formats and the simple URL method seems to be lacking subtitles in addition: changes to teaser image and subtitle extraction --- youtube_dl/extractor/zdf.py | 274 +++++++++++++++++++++++++----------- 1 file changed, 192 insertions(+), 82 deletions(-) diff --git a/youtube_dl/extractor/zdf.py b/youtube_dl/extractor/zdf.py index 8054df1eb..530a61cdf 100644 --- a/youtube_dl/extractor/zdf.py +++ b/youtube_dl/extractor/zdf.py @@ -8,8 +8,10 @@ from .common import InfoExtractor from ..utils import ( OnDemandPagedList, determine_ext, - parse_iso8601 + parse_iso8601, + ExtractorError ) +from ..compat import compat_str class ZDFIE(InfoExtractor): _VALID_URL = r'https?://www\.zdf\.de/.*?/(?P[^/?]*?)\.html' @@ -25,23 +27,124 @@ class ZDFIE(InfoExtractor): def _real_extract(self, url): video_id = self._match_id(url) - webpage = self._download_webpage(url, video_id) + try: + extr_player = ZDFExtractorPlayer(self, url, video_id) + formats = extr_player._real_extract() + except (ExtractorError, KeyError) as e: + self._downloader.report_warning('%s: %s\nusing fallback method (mobile url)' % (type(e).__name__, compat_str(e))) + extr_mobile = ZDFExtractorMobile(self, url, video_id) + formats = extr_mobile._real_extract() + return formats - jsb = self._search_regex(r"data-zdfplayer-jsb='([^']*)'", webpage, 'zdfplayer jsb data') - jsb_json = self._parse_json(jsb, video_id) +class ZDFExtractor: + """Super class for the 2 extraction methods""" + def __init__(self, parent, url, video_id): + self.parent = parent + self.url = url + self.video_id = video_id + + def _real_extract(self): + formats = [] + for entry in self._fetch_entries(): + video_url = self._get_video_url(entry) + if not video_url: + continue + format_id = self._get_format_id(entry) + ext = determine_ext(video_url, None) + if ext == 'meta': + continue + if ext == 'm3u8': + formats.extend(self.parent._extract_m3u8_formats( + video_url, self.video_id, 'mp4', m3u8_id=format_id, fatal=False)) + elif ext == 'f4m': + formats.extend(self.parent._extract_f4m_formats( + video_url, self.video_id, f4m_id=format_id, fatal=False)) + else: + formats.append({ + 'format_id': format_id, + 'url': video_url, + 'format_note': self._get_format_note(entry) + }) + self.parent._sort_formats(formats) + + return { + 'id': self.video_id, + 'title': self._get_title(), + 'formats': formats, + 'subtitles': self._get_subtitles(), + 'thumbnail': self._get_thumbnail(), + 'description': self._get_description(), + 'timestamp': self._get_timestamp() + } + +class ZDFExtractorMobile(ZDFExtractor): + """Simple URL extraction method. Disadvantage: fewer formats, no subtitles""" + def __init__(self, parent, url, video_id): + ZDFExtractor.__init__(self, parent, url, video_id) + + def _fetch_entries(self): + meta_data_url = 'https://zdf-cdn.live.cellular.de/mediathekV2/document/' + self.video_id + self.meta_data = self.parent._download_json(meta_data_url, self.video_id, note='Downloading meta data') + return self.meta_data['document']['formitaeten'] + + def _get_title(self): + return self.meta_data['document']['titel'] + + def _get_video_url(self, entry): + return entry['url'] + + def _get_format_id(self, entry): + format_id = entry['type'] + if 'quality' in entry: + format_id += '-' + entry['quality'] + return format_id + + def _get_format_note(self, entry): + return None + + def _get_subtitles(self): + return None + + def _get_description(self): + return self.meta_data['document'].get('beschreibung') + + def _get_timestamp(self): + meta = self.meta_data['meta'] + if meta: + return parse_iso8601(meta.get('editorialDate')) + + def _get_thumbnail(self): + teaser_images = self.meta_data['document'].get('teaserBild') + if teaser_images: + max_res = max(teaser_images, key=int) + return teaser_images[max_res].get('url') + +class ZDFExtractorPlayer(ZDFExtractor): + """Extraction method that requires downloads of several pages. + + Follows the requests of the website.""" + def __init__(self, parent, url, video_id): + ZDFExtractor.__init__(self, parent, url, video_id) + + def _fetch_entries(self): + webpage = self.parent._download_webpage(self.url, self.video_id) + + jsb = self.parent._search_regex(r"data-zdfplayer-jsb='([^']*)'", webpage, 'zdfplayer jsb data') + jsb_json = self.parent._parse_json(jsb, self.video_id) configuration_url = 'https://www.zdf.de' + jsb_json['config'] - configuration_json = self._download_json(configuration_url, video_id, note='Downloading player configuration') + configuration_json = self.parent._download_json(configuration_url, self.video_id, note='Downloading player configuration') api_token = configuration_json['apiToken'] - player_js = self._download_webpage('https://www.zdf.de/ZDFplayer/latest-v2/skins/zdf/zdf-player.js', video_id, fatal=False, note='Downloading player script') + player_js = self.parent._download_webpage('https://www.zdf.de/ZDFplayer/latest-v2/skins/zdf/zdf-player.js', self.video_id, fatal=False, note='Downloading player script') if player_js: - player_id = self._search_regex(r'this\.ptmd_player_id="([^"]*)"', player_js, 'player id', fatal=False) + player_id = self.parent._search_regex(r'this\.ptmd_player_id="([^"]*)"', player_js, 'player id', fatal=False) else: player_id = None - content_json = self._download_json(jsb_json['content'], video_id, headers={'Api-Auth': 'Bearer %s' % api_token}, note='Downloading content description') - main_video_content = content_json['mainVideoContent']['http://zdf.de/rels/target'] + self.content_json = self.parent._download_json(jsb_json['content'], self.video_id, headers={'Api-Auth': 'Bearer %s' % api_token}, note='Downloading content description') + + main_video_content = self.content_json['mainVideoContent']['http://zdf.de/rels/target'] meta_data_url = None if not player_id: # could not determine player_id => try alternativ generic URL @@ -55,89 +158,96 @@ class ZDFIE(InfoExtractor): meta_data_url_template = main_video_content['http://zdf.de/rels/streams/ptmd-template'] meta_data_url = 'https://api.zdf.de' + meta_data_url_template.replace('{playerId}', player_id) - title = content_json['title'] - - meta_data = self._download_json(meta_data_url, video_id, note='Downloading meta data') + self.meta_data = self.parent._download_json(meta_data_url, self.video_id, note='Downloading meta data') formats = [] - for p_list_entry in meta_data['priorityList']: + for p_list_entry in self.meta_data['priorityList']: for formitaet in p_list_entry['formitaeten']: - # mime = formitaet.get('mimeType') - facets = formitaet.get('facets') or [] - add = '' - if formitaet.get('isAdaptive'): - add += 'a' - facets.append('adaptive') - if 'restriction_useragent' in facets: - add += 'b' - if 'progressive' in facets: - add += 'p' - type_ = formitaet['type'] for entry in formitaet['qualities']: - tracks = entry['audio']['tracks'] - if not tracks: - continue - if len(tracks) > 1: - self._downloader.report_warning('unexpected input: multiple tracks') - track = tracks[0] - video_url = track['uri'] - format_id = type_ + '-' - if add: - format_id += add + '-' - # named qualities are not very useful for sorting the formats: - # a 'high' m3u8 entry can be better quality than a 'veryhigh' direct mp4 download - format_id += entry['quality'] - ext = determine_ext(video_url, None) - if ext == 'meta': - continue - if ext == 'm3u8': - formats.extend(self._extract_m3u8_formats( - video_url, video_id, 'mp4', 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: - formats.append({ - 'format_id': format_id, - 'url': video_url, - 'format_note': ', '.join(facets) - }) - self._sort_formats(formats) + yield (formitaet, entry) + def _get_title(self): + return self.content_json['title'] + + def _get_video_url(self, entry_tuple): + (formitaet, entry) = entry_tuple + tracks = entry['audio'].get('tracks') + if not tracks: + return + if len(tracks) > 1: + self._downloader.report_warning('unexpected input: multiple tracks') + track = tracks[0] + return track['uri'] + + def _get_format_id(self, entry_tuple): + (formitaet, entry) = entry_tuple + facets = self._get_facets(formitaet) + add = '' + if 'adaptive' in facets: + add += 'a' + if 'restriction_useragent' in facets: + add += 'b' + if 'progressive' in facets: + add += 'p' + type_ = formitaet['type'] + format_id = type_ + '-' + if add: + format_id += add + '-' + # named qualities are not very useful for sorting the formats: + # a 'high' m3u8 entry can be better quality than a 'veryhigh' direct mp4 download + format_id += entry['quality'] + return format_id + + def _get_facets(self, formitaet): + facets = formitaet.get('facets') or [] + if formitaet.get('isAdaptive'): + facets.append('adaptive') + return facets + + def _get_format_note(self, entry_tuple): + (formitaet, entry) = entry_tuple + return ', '.join(self._get_facets(formitaet)) + + def _get_subtitles(self): subtitles = {} - if meta_data.get('captions'): - subtitles['de'] = [] - for caption in meta_data['captions']: - if caption.get('language') == 'deu': - subformat = {'url': caption.get('uri')} - if caption.get('format') == 'webvtt': - subformat['ext'] = 'vtt' - elif caption.get('format') == 'ebu-tt-d-basic-de': - subformat['ext'] = 'ttml' - subtitles['de'].append(subformat) + if 'captions' in self.meta_data: + for caption in self.meta_data['captions']: + lang = caption.get('language') + if not lang: + continue + if lang == 'deu': + lang = 'de' + subformat = {'url': caption.get('uri')} + if caption.get('format') == 'webvtt': + subformat['ext'] = 'vtt' + elif caption.get('format') == 'ebu-tt-d-basic-de': + subformat['ext'] = 'ttml' + if not lang in subtitles: + subtitles[lang] = [] + subtitles[lang].append(subformat) + return subtitles - teaser_images = content_json.get('teaserImageRef') + def _get_description(self): + return self.content_json.get('teasertext') + + def _get_timestamp(self): + return parse_iso8601(self.content_json.get('editorialDate')) + + def _get_thumbnail(self): + teaser_images = self.content_json.get('teaserImageRef') if teaser_images: teaser_images_layouts = teaser_images.get('layouts') if teaser_images_layouts: - thumbnail = teaser_images_layouts.get('original') - else: - thumbnail = None - else: - thumbnail = None - - description = content_json.get('teasertext') - timestamp = parse_iso8601(content_json.get('editorialDate')) - - return { - 'id': video_id, - 'title': title, - 'formats': formats, - 'subtitles': subtitles, - 'thumbnail': thumbnail, - 'description': description, - 'timestamp': timestamp - } + if 'original' in teaser_images_layouts: + return teaser_images_layouts['original'] + teasers = {} + for key in teaser_images_layouts: + width = self.parent._search_regex(r'(\d+)x\d+', key, 'teaser width', fatal=False) + if width: + teasers[int(width)] = teaser_images_layouts[key] + if teasers: + best = max(teasers) + return teasers[best] class ZDFChannelIE(InfoExtractor): _WORKING = False