+# coding: utf-8
+from __future__ import unicode_literals
+
+import functools
import re
from .common import InfoExtractor
from ..utils import (
- determine_ext,
- ExtractorError,
+ int_or_none,
+ unified_strdate,
+ OnDemandPagedList,
)
+def extract_from_xml_url(ie, video_id, xml_url):
+ doc = ie._download_xml(
+ xml_url, video_id,
+ note='Downloading video info',
+ errnote='Failed to download video info')
+
+ title = doc.find('.//information/title').text
+ description = doc.find('.//information/detail').text
+ duration = int(doc.find('.//details/lengthSec').text)
+ uploader_node = doc.find('.//details/originChannelTitle')
+ uploader = None if uploader_node is None else uploader_node.text
+ uploader_id_node = doc.find('.//details/originChannelId')
+ uploader_id = None if uploader_id_node is None else uploader_id_node.text
+ upload_date = unified_strdate(doc.find('.//details/airtime').text)
+
+ def xml_to_format(fnode):
+ video_url = fnode.find('url').text
+ is_available = 'http://www.metafilegenerator' not in video_url
+
+ format_id = fnode.attrib['basetype']
+ format_m = re.match(r'''(?x)
+ (?P<vcodec>[^_]+)_(?P<acodec>[^_]+)_(?P<container>[^_]+)_
+ (?P<proto>[^_]+)_(?P<index>[^_]+)_(?P<indexproto>[^_]+)
+ ''', format_id)
+
+ ext = format_m.group('container')
+ proto = format_m.group('proto').lower()
+
+ quality = fnode.find('./quality').text
+ abr = int(fnode.find('./audioBitrate').text) // 1000
+ vbr_node = fnode.find('./videoBitrate')
+ vbr = None if vbr_node is None else int(vbr_node.text) // 1000
+
+ width_node = fnode.find('./width')
+ width = None if width_node is None else int_or_none(width_node.text)
+ height_node = fnode.find('./height')
+ height = None if height_node is None else int_or_none(height_node.text)
+
+ format_note = ''
+ if not format_note:
+ format_note = None
+
+ return {
+ 'format_id': format_id + '-' + quality,
+ 'url': video_url,
+ 'ext': ext,
+ 'acodec': format_m.group('acodec'),
+ 'vcodec': format_m.group('vcodec'),
+ 'abr': abr,
+ 'vbr': vbr,
+ 'width': width,
+ 'height': height,
+ 'filesize': int_or_none(fnode.find('./filesize').text),
+ 'format_note': format_note,
+ 'protocol': proto,
+ '_available': is_available,
+ }
+
+ format_nodes = doc.findall('.//formitaeten/formitaet')
+ formats = list(filter(
+ lambda f: f['_available'],
+ map(xml_to_format, format_nodes)))
+ ie._sort_formats(formats)
+
+ return {
+ 'id': video_id,
+ 'title': title,
+ 'description': description,
+ 'duration': duration,
+ 'uploader': uploader,
+ 'uploader_id': uploader_id,
+ 'upload_date': upload_date,
+ 'formats': formats,
+ }
+
+
class ZDFIE(InfoExtractor):
- _VALID_URL = r'^http://www\.zdf\.de\/ZDFmediathek(?P<hash>#)?\/(.*beitrag\/video\/)(?P<video_id>[^/\?]+)(?:\?.*)?'
- _MEDIA_STREAM = r'<a href="(?P<video_url>.+(?P<media_type>.streaming).+/zdf/(?P<quality>[^\/]+)/[^"]*)".+class="play".+>'
+ _VALID_URL = r'(?:zdf:|zdf:video:|https?://www\.zdf\.de/ZDFmediathek(?:#)?/(.*beitrag/(?:video/)?))(?P<id>[0-9]+)(?:/[^/?]+)?(?:\?.*)?'
+
+ _TEST = {
+ 'url': 'http://www.zdf.de/ZDFmediathek/beitrag/video/2037704/ZDFspezial---Ende-des-Machtpokers--?bc=sts;stt',
+ 'info_dict': {
+ 'id': '2037704',
+ 'ext': 'webm',
+ 'title': 'ZDFspezial - Ende des Machtpokers',
+ 'description': 'Union und SPD haben sich auf einen Koalitionsvertrag geeinigt. Aber was bedeutet das für die Bürger? Sehen Sie hierzu das ZDFspezial "Ende des Machtpokers - Große Koalition für Deutschland".',
+ 'duration': 1022,
+ 'uploader': 'spezial',
+ 'uploader_id': '225948',
+ 'upload_date': '20131127',
+ },
+ 'skip': 'Videos on ZDF.de are depublicised in short order',
+ }
+
+ def _real_extract(self, url):
+ video_id = self._match_id(url)
+ xml_url = 'http://www.zdf.de/ZDFmediathek/xmlservice/web/beitragsDetails?ak=web&id=%s' % video_id
+ return extract_from_xml_url(self, video_id, xml_url)
+
+
+class ZDFChannelIE(InfoExtractor):
+ _VALID_URL = r'(?:zdf:topic:|https?://www\.zdf\.de/ZDFmediathek(?:#)?/.*kanaluebersicht/)(?P<id>[0-9]+)'
+ _TEST = {
+ 'url': 'http://www.zdf.de/ZDFmediathek#/kanaluebersicht/1586442/sendung/Titanic',
+ 'info_dict': {
+ 'id': '1586442',
+ },
+ 'playlist_count': 3,
+ }
+ _PAGE_SIZE = 50
+
+ def _fetch_page(self, channel_id, page):
+ offset = page * self._PAGE_SIZE
+ xml_url = (
+ 'http://www.zdf.de/ZDFmediathek/xmlservice/web/aktuellste?ak=web&offset=%d&maxLength=%d&id=%s'
+ % (offset, self._PAGE_SIZE, channel_id))
+ doc = self._download_xml(
+ xml_url, channel_id,
+ note='Downloading channel info',
+ errnote='Failed to download channel info')
+
+ title = doc.find('.//information/title').text
+ description = doc.find('.//information/detail').text
+ for asset in doc.findall('.//teasers/teaser'):
+ a_type = asset.find('./type').text
+ a_id = asset.find('./details/assetId').text
+ if a_type not in ('video', 'topic'):
+ continue
+ yield {
+ '_type': 'url',
+ 'playlist_title': title,
+ 'playlist_description': description,
+ 'url': 'zdf:%s:%s' % (a_type, a_id),
+ }
def _real_extract(self, url):
- mobj = re.match(self._VALID_URL, url)
- if mobj is None:
- raise ExtractorError(u'Invalid URL: %s' % url)
- video_id = mobj.group('video_id')
-
- if mobj.group('hash'):
- url = url.replace(u'#', u'', 1)
-
- html = self._download_webpage(url, video_id)
- streams = [m.groupdict() for m in re.finditer(self._MEDIA_STREAM, html)]
- if streams is None:
- raise ExtractorError(u'No media url found.')
-
- # s['media_type'] == 'wstreaming' -> use 'Windows Media Player' and mms url
- # s['media_type'] == 'hstreaming' -> use 'Quicktime' and rtsp url
- # choose first/default media type and highest quality for now
- def stream_pref(s):
- TYPE_ORDER = ['ostreaming', 'hstreaming', 'wstreaming']
- try:
- type_pref = TYPE_ORDER.index(s['media_type'])
- except ValueError:
- type_pref = 999
-
- QUALITY_ORDER = ['veryhigh', '300']
- try:
- quality_pref = QUALITY_ORDER.index(s['quality'])
- except ValueError:
- quality_pref = 999
-
- return (type_pref, quality_pref)
-
- sorted_streams = sorted(streams, key=stream_pref)
- if not sorted_streams:
- raise ExtractorError(u'No stream found.')
- stream = sorted_streams[0]
-
- media_link = self._download_webpage(
- stream['video_url'],
- video_id,
- u'Get stream URL')
-
- MMS_STREAM = r'href="(?P<video_url>mms://[^"]*)"'
- RTSP_STREAM = r'(?P<video_url>rtsp://[^"]*.mp4)'
-
- mobj = re.search(self._MEDIA_STREAM, media_link)
- if mobj is None:
- mobj = re.search(RTSP_STREAM, media_link)
- if mobj is None:
- raise ExtractorError(u'Cannot extract mms:// or rtsp:// URL')
- video_url = mobj.group('video_url')
-
- title = self._html_search_regex(
- r'<h1(?: class="beitragHeadline")?>(.*?)</h1>',
- html, u'title')
+ channel_id = self._match_id(url)
+ entries = OnDemandPagedList(
+ functools.partial(self._fetch_page, channel_id), self._PAGE_SIZE)
return {
- 'id': video_id,
- 'url': video_url,
- 'title': title,
- 'ext': determine_ext(video_url)
+ '_type': 'playlist',
+ 'id': channel_id,
+ 'entries': entries,
}