youtube-dl/youtube_dl/extractor/mdr.py

196 lines
6.9 KiB
Python
Raw Permalink Normal View History

2015-10-31 16:17:09 +00:00
# coding: utf-8
from __future__ import unicode_literals
2013-12-10 17:40:50 +00:00
from .common import InfoExtractor
from ..compat import (
compat_str,
compat_urlparse,
)
2015-10-31 16:17:09 +00:00
from ..utils import (
determine_ext,
int_or_none,
parse_duration,
parse_iso8601,
url_or_none,
2015-10-31 16:17:09 +00:00
xpath_text,
)
2013-12-10 17:40:50 +00:00
2013-12-16 04:44:34 +00:00
2013-12-10 17:40:50 +00:00
class MDRIE(InfoExtractor):
2015-10-31 16:17:09 +00:00
IE_DESC = 'MDR.DE and KiKA'
_VALID_URL = r'https?://(?:www\.)?(?:mdr|kika)\.de/(?:.*)/[a-z-]+-?(?P<id>\d+)(?:_.+?)?\.html'
2014-11-23 19:41:03 +00:00
2020-12-13 16:43:16 +00:00
_GEO_COUNTRIES = ['DE']
2015-10-31 16:17:09 +00:00
_TESTS = [{
2016-01-10 15:17:47 +00:00
# MDR regularly deletes its videos
'url': 'http://www.mdr.de/fakt/video189002.html',
'only_matching': True,
2015-10-31 17:00:36 +00:00
}, {
2015-10-31 16:24:21 +00:00
# audio
'url': 'http://www.mdr.de/kultur/audio1312272_zc-15948bad_zs-86171fdd.html',
'md5': '64c4ee50f0a791deb9479cd7bbe9d2fa',
'info_dict': {
'id': '1312272',
'ext': 'mp3',
'title': 'Feuilleton vom 30. Oktober 2015',
'duration': 250,
'uploader': 'MITTELDEUTSCHER RUNDFUNK',
},
'skip': '404 not found',
2015-10-31 16:17:09 +00:00
}, {
'url': 'http://www.kika.de/baumhaus/videos/video19636.html',
'md5': '4930515e36b06c111213e80d1e4aad0e',
'info_dict': {
'id': '19636',
'ext': 'mp4',
'title': 'Baumhaus vom 30. Oktober 2015',
'duration': 134,
'uploader': 'KIKA',
},
'skip': '404 not found',
2015-10-31 16:17:09 +00:00
}, {
'url': 'http://www.kika.de/sendungen/einzelsendungen/weihnachtsprogramm/videos/video8182.html',
'md5': '5fe9c4dd7d71e3b238f04b8fdd588357',
'info_dict': {
'id': '8182',
'ext': 'mp4',
'title': 'Beutolomäus und der geheime Weihnachtswunsch',
'description': 'md5:b69d32d7b2c55cbe86945ab309d39bbd',
'timestamp': 1482541200,
'upload_date': '20161224',
2015-10-31 16:17:09 +00:00
'duration': 4628,
'uploader': 'KIKA',
},
}, {
# audio with alternative playerURL pattern
'url': 'http://www.mdr.de/kultur/videos-und-audios/audio-radio/operation-mindfuck-robert-wilson100.html',
'info_dict': {
'id': '100',
'ext': 'mp4',
'title': 'Feature: Operation Mindfuck - Robert Anton Wilson',
'duration': 3239,
'uploader': 'MITTELDEUTSCHER RUNDFUNK',
},
}, {
# empty bitrateVideo and bitrateAudio
'url': 'https://www.kika.de/filme/sendung128372_zc-572e3f45_zs-1d9fb70e.html',
'info_dict': {
'id': '128372',
'ext': 'mp4',
'title': 'Der kleine Wichtel kehrt zurück',
'description': 'md5:f77fafdff90f7aa1e9dca14f662c052a',
'duration': 4876,
'timestamp': 1607823300,
'upload_date': '20201213',
'uploader': 'ZDF',
},
'params': {
'skip_download': True,
},
2015-10-31 16:17:09 +00:00
}, {
'url': 'http://www.kika.de/baumhaus/sendungen/video19636_zc-fea7f8a0_zs-4bf89c60.html',
'only_matching': True,
}, {
'url': 'http://www.kika.de/sendungen/einzelsendungen/weihnachtsprogramm/einzelsendung2534.html',
'only_matching': True,
2016-02-29 19:24:26 +00:00
}, {
'url': 'http://www.mdr.de/mediathek/mdr-videos/a/video-1334.html',
'only_matching': True,
2015-10-31 16:17:09 +00:00
}]
2013-12-10 17:40:50 +00:00
def _real_extract(self, url):
2015-10-31 16:17:09 +00:00
video_id = self._match_id(url)
webpage = self._download_webpage(url, video_id)
2013-12-10 17:40:50 +00:00
2015-10-31 16:17:09 +00:00
data_url = self._search_regex(
r'(?:dataURL|playerXml(?:["\'])?)\s*:\s*(["\'])(?P<url>.+?-avCustom\.xml)\1',
webpage, 'data url', group='url').replace(r'\/', '/')
2013-12-10 17:40:50 +00:00
2015-10-31 16:17:09 +00:00
doc = self._download_xml(
compat_urlparse.urljoin(url, data_url), video_id)
2015-10-31 16:45:45 +00:00
title = xpath_text(doc, ['./title', './broadcast/broadcastName'], 'title', fatal=True)
2013-12-16 04:44:34 +00:00
type_ = xpath_text(doc, './type', default=None)
2013-12-16 04:44:34 +00:00
formats = []
2015-10-31 16:17:09 +00:00
processed_urls = []
for asset in doc.findall('./assets/asset'):
for source in (
'download',
2015-10-31 16:17:09 +00:00
'progressiveDownload',
'dynamicHttpStreamingRedirector',
'adaptiveHttpStreamingRedirector'):
url_el = asset.find('./%sUrl' % source)
if url_el is None:
continue
video_url = url_or_none(url_el.text)
if not video_url or video_url in processed_urls:
2015-10-31 16:17:09 +00:00
continue
processed_urls.append(video_url)
ext = determine_ext(video_url)
2015-10-31 16:17:09 +00:00
if ext == 'm3u8':
formats.extend(self._extract_m3u8_formats(
2015-10-31 16:17:09 +00:00
video_url, video_id, 'mp4', entry_protocol='m3u8_native',
preference=0, m3u8_id='HLS', fatal=False))
2015-10-31 16:17:09 +00:00
elif ext == 'f4m':
formats.extend(self._extract_f4m_formats(
2015-10-31 16:17:09 +00:00
video_url + '?hdcore=3.7.0&plugin=aasp-3.7.0.39.44', video_id,
preference=0, f4m_id='HDS', fatal=False))
2015-10-31 16:17:09 +00:00
else:
media_type = xpath_text(asset, './mediaType', 'media type', default='MP4')
vbr = int_or_none(xpath_text(asset, './bitrateVideo', 'vbr'), 1000)
abr = int_or_none(xpath_text(asset, './bitrateAudio', 'abr'), 1000)
filesize = int_or_none(xpath_text(asset, './fileSize', 'file size'))
format_id = [media_type]
if vbr or abr:
format_id.append(compat_str(vbr or abr))
2015-10-31 16:17:09 +00:00
f = {
'url': video_url,
'format_id': '-'.join(format_id),
2015-10-31 16:17:09 +00:00
'filesize': filesize,
'abr': abr,
'vbr': vbr,
2015-10-31 16:17:09 +00:00
}
if vbr:
f.update({
'width': int_or_none(xpath_text(asset, './frameWidth', 'width')),
'height': int_or_none(xpath_text(asset, './frameHeight', 'height')),
2015-10-31 16:17:09 +00:00
})
if type_ == 'audio':
f['vcodec'] = 'none'
2015-10-31 16:17:09 +00:00
formats.append(f)
2015-10-31 18:01:08 +00:00
2013-12-24 22:34:11 +00:00
self._sort_formats(formats)
2015-10-31 16:17:09 +00:00
description = xpath_text(doc, './broadcast/broadcastDescription', 'description')
timestamp = parse_iso8601(
2015-10-31 16:45:45 +00:00
xpath_text(
doc, [
'./broadcast/broadcastDate',
'./broadcast/broadcastStartDate',
'./broadcast/broadcastEndDate'],
'timestamp', default=None))
2015-10-31 16:17:09 +00:00
duration = parse_duration(xpath_text(doc, './duration', 'duration'))
uploader = xpath_text(doc, './rights', 'uploader')
2013-12-16 04:44:34 +00:00
return {
'id': video_id,
'title': title,
2015-10-31 16:17:09 +00:00
'description': description,
'timestamp': timestamp,
'duration': duration,
'uploader': uploader,
2013-12-16 04:44:34 +00:00
'formats': formats,
}