[DRBonanza] Add new extractor (fixing #4581)

pull/4687/head
Johan K. Jensen 10 years ago
parent c85f368370
commit 8e7a9016d5

@ -93,6 +93,7 @@ from .deezer import DeezerPlaylistIE
from .dfb import DFBIE
from .dotsub import DotsubIE
from .dreisat import DreiSatIE
from .drbonanza import DRBonanzaIE
from .drtuber import DrTuberIE
from .drtv import DRTVIE
from .dvtv import DVTVIE

@ -0,0 +1,120 @@
from __future__ import unicode_literals
from .common import InfoExtractor
from .common import ExtractorError
from ..utils import parse_iso8601
import json
import re
class DRBonanzaIE(InfoExtractor):
_VALID_URL = r'https?://(?:www\.)?dr\.dk/bonanza/(?:[^/]+/)+(?:[^/])+?(?:assetId=(?P<id>\d+))?(?:[#&]|$)'
_TESTS = [{
'url': 'http://www.dr.dk/bonanza/serie/portraetter/Talkshowet.htm?assetId=65517',
'md5': 'fe330252ddea607635cf2eb2c99a0af3',
'info_dict': {
'id': '65517',
'ext': 'mp4',
'title': 'Talkshowet - Leonard Cohen',
'description': 'md5:8f34194fb30cd8c8a30ad8b27b70c0ca',
'timestamp': 1295537932,
'upload_date': '20110120',
'duration': 3664000,
},
},{
'url': 'http://www.dr.dk/bonanza/radio/serie/sport/fodbold.htm?assetId=59410',
'md5': '6dfe039417e76795fb783c52da3de11d',
'info_dict': {
'id': '59410',
'ext': 'mp3',
'title': 'EM fodbold 1992 Danmark - Tyskland finale Transmission',
'description': 'md5:501e5a195749480552e214fbbed16c4e',
'timestamp': 1223274900,
'upload_date': '20081006',
'duration': 7369000,
},
}]
def _real_extract(self, url):
url_id = self._match_id(url)
webpage = self._download_webpage(url, url_id if url_id else "")
if url_id:
info = json.loads(self._html_search_regex(r'({.*?' + url_id + '.*})', webpage, 'json'))
else:
# Just fetch the first video on that page
info = json.loads(self._html_search_regex(r'bonanzaFunctions.newPlaylist\(({.*})\)', webpage, 'json'))
asset_id = str(info['AssetId'])
title = info['Title'].rstrip(' \'\"-,.:;!?')
duration = info['Duration']
timestamp = parse_iso8601(re.sub(r'\.\d+$', '', info['Created'])) # First published online. "FirstPublished" contains the date for original airing.
def parse_filename_info(url):
match = re.search(r'/\d+_(?P<width>\d+)x(?P<height>\d+)x(?P<bitrate>\d+)K\.(?P<ext>\w+)$', url)
if match:
return {'width': int(match.group(1)), 'height': int(match.group(2)), 'bitrate': int(match.group(3)), 'ext': match.group(4)}
match = re.search(r'/\d+_(?P<bitrate>\d+)K\.(?P<ext>\w+)$', url)
if match:
return {'bitrate': int(match.group(1)), 'ext': match.group(2)}
return {'width': None, 'height': None, 'bitrate': None, 'ext': None}
video_types = ['VideoHigh', 'VideoMid', 'VideoLow']
preferencemap = {
'VideoHigh': -1,
'VideoMid': -2,
'VideoLow': -3,
'Audio': -4,
}
formats = []
for file in info['Files']:
if info['Type'] == "Video":
if file['Type'] in video_types:
fileinfo = parse_filename_info(file['Location'])
formats.append({
'url': file['Location'],
'format_id': file['Type'].replace('Video', ''),
'preference': preferencemap.get(file['Type'], -10),
'width': fileinfo['width'],
'height': fileinfo['height'],
'vbr': fileinfo['bitrate'],
'ext': fileinfo['ext'],
})
elif file['Type'] == "Thumb":
thumbnail = file['Location']
elif info['Type'] == "Audio":
if file['Type'] == "Audio":
fileinfo = parse_filename_info(file['Location'])
formats.append({
'url': file['Location'],
'format_id': file['Type'],
'abr': fileinfo['bitrate'],
'ext': fileinfo['ext'],
'vcodec': 'none',
})
elif file['Type'] == "Thumb":
thumbnail = file['Location']
description = "{}\n{}\n{}\n".format(info['Description'], info['Actors'], info['Colophon'])
for f in formats:
f['url'] = f['url'].replace('rtmp://vod-bonanza.gss.dr.dk/bonanza/', 'http://vodfiles.dr.dk/')
f['url'] = f['url'].replace('mp4:bonanza', 'bonanza')
self._sort_formats(formats)
display_id = re.sub(r'[^\w\d-]', '', re.sub(r' ', '-', title.lower())) + '-' + asset_id
display_id = re.sub(r'-+', '-', display_id)
return {
'id': asset_id,
'display_id': display_id,
'title': title,
'formats': formats,
'description': description,
'thumbnail': thumbnail,
'timestamp': timestamp,
'duration': duration,
}
Loading…
Cancel
Save