1 # -*- coding: utf-8 -*-
2 from __future__ import unicode_literals
6 from .common import InfoExtractor
7 from ..utils import parse_filesize, ExtractorError
10 class TagesschauIE(InfoExtractor):
11 _VALID_URL = r'https?://(?:www\.)?tagesschau\.de/multimedia/(?:sendung/(ts|tsg|tt|nm|bab/bab)|video/video|tsvorzwanzig)(?P<id>-?[0-9]+)(?:~[-_a-zA-Z0-9]*)?\.html'
14 'url': 'http://www.tagesschau.de/multimedia/video/video1399128.html',
15 'md5': 'bcdeac2194fb296d599ce7929dfa4009',
19 'title': 'Harald Range, Generalbundesanwalt, zu den Ermittlungen',
20 'description': 'md5:69da3c61275b426426d711bde96463ab',
21 'thumbnail': 're:^http:.*\.jpg$',
24 'url': 'http://www.tagesschau.de/multimedia/sendung/ts-5727.html',
25 'md5': '3c54c1f6243d279b706bde660ceec633',
29 'description': 'md5:695c01bfd98b7e313c501386327aea59',
30 'title': 'Sendung: tagesschau \t04.12.2014 20:00 Uhr',
31 'thumbnail': 're:^http:.*\.jpg$',
34 'url': 'http://www.tagesschau.de/multimedia/sendung/tsg-3771.html',
35 'md5': '90757268b49ef56deae90c7b48928d58',
40 'title': 'Sendung: tagesschau (mit Gebärdensprache) \t14.07.2015 20:00 Uhr',
41 'thumbnail': 're:^http:.*\.jpg$',
44 'url': 'http://www.tagesschau.de/multimedia/sendung/tt-3827.html',
45 'md5': '6e3ebdc75e8d67da966a8d06721eda71',
49 'description': 'md5:d511d0e278b0ad341a95ad9ab992ce66',
50 'title': 'Sendung: tagesthemen \t14.07.2015 22:15 Uhr',
51 'thumbnail': 're:^http:.*\.jpg$',
54 'url': 'http://www.tagesschau.de/multimedia/sendung/nm-3475.html',
55 'md5': '8a8875a568f0a5ae5ceef93c501a225f',
59 'description': 'md5:ed149f5649cda3dac86813a9d777e131',
60 'title': 'Sendung: nachtmagazin \t15.07.2015 00:15 Uhr',
61 'thumbnail': 're:^http:.*\.jpg$',
64 'url': 'http://www.tagesschau.de/multimedia/tsvorzwanzig-959.html',
65 'md5': 'be4d6f0421f2acd8abe25ea29f6f015b',
70 'title': 'Sendung: tagesschau vor 20 Jahren \t14.07.2015 22:45 Uhr',
71 'thumbnail': 're:^http:.*\.jpg$',
74 'url': 'http://www.tagesschau.de/multimedia/sendung/bab/bab-3299~_bab-sendung-209.html',
75 'md5': '42e3757018d9908581481a80cc1806da',
80 'title': 'Nach dem Referendum: Schaltgespräch nach Athen',
81 'thumbnail': 're:^http:.*\.jpg$',
86 's': {'width': 256, 'height': 144, 'quality': 1},
87 'm': {'width': 512, 'height': 288, 'quality': 2},
88 'l': {'width': 960, 'height': 544, 'quality': 3},
91 def _real_extract(self, url):
92 video_id = self._match_id(url)
93 display_id = video_id.lstrip('-')
94 webpage = self._download_webpage(url, display_id)
96 player_url = self._html_search_meta(
97 'twitter:player', webpage, 'player URL', default=None)
99 playerpage = self._download_webpage(
100 player_url, display_id, 'Downloading player page')
103 r'"(http://media.+?)", type:"video/(.+?)", quality:"(.+?)"',
106 for url, ext, res in medias:
108 'format_id': res + '_' + ext,
112 f.update(self._FORMATS.get(res, {}))
114 thumbnail_fn = re.findall(r'"(/multimedia/.+?\.jpg)"', playerpage)[-1]
115 title = self._og_search_title(webpage).strip()
116 description = self._og_search_description(webpage).strip()
118 download_text = self._search_regex(
119 r'(?s)<p>Wir bieten dieses Video in folgenden Formaten zum Download an:</p>\s*<div class="controls">(.*?)</div>\s*<p>',
120 webpage, 'download links')
122 r'<div class="button" title="(?P<title>[^"]*)"><a href="(?P<url>[^"]+)">(?P<name>.+?)</a></div>',
126 format_id = self._search_regex(
127 r'.*/[^/.]+\.([^/]+)\.[^/.]+', l.group('url'), 'format ID')
129 'format_id': format_id,
130 'url': l.group('url'),
131 'format_name': l.group('name'),
135 Video:\s*(?P<vcodec>[a-zA-Z0-9/._-]+)\s*&\#10;
136 (?P<width>[0-9]+)x(?P<height>[0-9]+)px&\#10;
137 (?P<vbr>[0-9]+)kbps&\#10;
138 Audio:\s*(?P<abr>[0-9]+)kbps,\s*(?P<audio_desc>[A-Za-z\.0-9]+)&\#10;
139 Größe:\s*(?P<filesize_approx>[0-9.,]+\s+[a-zA-Z]*B)''',
143 'format_note': m.group('audio_desc'),
144 'vcodec': m.group('vcodec'),
145 'width': int(m.group('width')),
146 'height': int(m.group('height')),
147 'abr': int(m.group('abr')),
148 'vbr': int(m.group('vbr')),
149 'filesize_approx': parse_filesize(m.group('filesize_approx')),
151 formats.append(format)
152 thumbnail_fn = self._search_regex(
153 r'(?s)<img alt="Sendungsbild".*?src="([^"]+)"',
154 webpage, 'thumbnail', fatal=False)
155 # there are some videos without description
158 description = self._html_search_regex(
159 r'(?s)<p class="teasertext">(.*?)</p>',
160 webpage, 'description', fatal=False)
161 except ExtractorError:
163 title = self._html_search_regex(
164 r'<span class="headline".*?>(.*?)</span>', webpage, 'title')
166 self._sort_formats(formats)
167 thumbnail = 'http://www.tagesschau.de' + thumbnail_fn
172 'thumbnail': thumbnail,
174 'description': description,