diff options
Diffstat (limited to 'youtube_dl/extractor/dreisat.py')
-rw-r--r-- | youtube_dl/extractor/dreisat.py | 59 |
1 files changed, 3 insertions, 56 deletions
diff --git a/youtube_dl/extractor/dreisat.py b/youtube_dl/extractor/dreisat.py index 8ac8587be..028144f20 100644 --- a/youtube_dl/extractor/dreisat.py +++ b/youtube_dl/extractor/dreisat.py @@ -2,14 +2,10 @@ from __future__ import unicode_literals import re -from .common import InfoExtractor -from ..utils import ( - ExtractorError, - unified_strdate, -) +from .zdf import ZDFIE -class DreiSatIE(InfoExtractor): +class DreiSatIE(ZDFIE): IE_NAME = '3sat' _VALID_URL = r'(?:http://)?(?:www\.)?3sat\.de/mediathek/(?:index\.php|mediathek\.php)?\?(?:(?:mode|display)=[^&]+&)*obj=(?P<id>[0-9]+)$' _TESTS = [ @@ -35,53 +31,4 @@ class DreiSatIE(InfoExtractor): mobj = re.match(self._VALID_URL, url) video_id = mobj.group('id') details_url = 'http://www.3sat.de/mediathek/xmlservice/web/beitragsDetails?ak=web&id=%s' % video_id - details_doc = self._download_xml(details_url, video_id, 'Downloading video details') - - status_code = details_doc.find('./status/statuscode') - if status_code is not None and status_code.text != 'ok': - code = status_code.text - if code == 'notVisibleAnymore': - message = 'Video %s is not available' % video_id - else: - message = '%s returned error: %s' % (self.IE_NAME, code) - raise ExtractorError(message, expected=True) - - thumbnail_els = details_doc.findall('.//teaserimage') - thumbnails = [{ - 'width': int(te.attrib['key'].partition('x')[0]), - 'height': int(te.attrib['key'].partition('x')[2]), - 'url': te.text, - } for te in thumbnail_els] - - information_el = details_doc.find('.//information') - video_title = information_el.find('./title').text - video_description = information_el.find('./detail').text - - details_el = details_doc.find('.//details') - video_uploader = details_el.find('./channel').text - upload_date = unified_strdate(details_el.find('./airtime').text) - - format_els = details_doc.findall('.//formitaet') - formats = [{ - 'format_id': fe.attrib['basetype'], - 'width': int(fe.find('./width').text), - 'height': int(fe.find('./height').text), - 'url': fe.find('./url').text, - 'filesize': int(fe.find('./filesize').text), - 'video_bitrate': int(fe.find('./videoBitrate').text), - } for fe in format_els - if not fe.find('./url').text.startswith('http://www.metafilegenerator.de/')] - - self._sort_formats(formats) - - return { - '_type': 'video', - 'id': video_id, - 'title': video_title, - 'formats': formats, - 'description': video_description, - 'thumbnails': thumbnails, - 'thumbnail': thumbnails[-1]['url'], - 'uploader': video_uploader, - 'upload_date': upload_date, - } + return self.extract_from_xml_url(video_id, details_url) |