1
0
Fork 0
mirror of https://github.com/ytdl-org/youtube-dl.git synced 2024-11-18 08:16:10 +00:00
youtube-dl/youtube_dl/extractor/urort.py
Philipp Hagemeister 0dae5083f1 [urort] Add date
2014-03-27 02:56:23 +01:00

61 lines
1.8 KiB
Python

# coding: utf-8
from __future__ import unicode_literals
import re
from .common import InfoExtractor
from ..utils import (
compat_urllib_parse,
unified_strdate,
)
class UrortIE(InfoExtractor):
IE_DESC = 'NRK P3 Urørt'
_VALID_URL = r'https?://(?:www\.)?urort\.p3\.no/#!/Band/(?P<id>[^/]+)$'
_TEST = {
'url': 'https://urort.p3.no/#!/Band/Gerilja',
'md5': '5ed31a924be8a05e47812678a86e127b',
'info_dict': {
'id': '33124-4',
'ext': 'mp3',
'title': 'The Bomb',
'thumbnail': 're:^https?://.+\.jpg',
'like_count': int,
'uploader': 'Gerilja',
'uploader_id': 'Gerilja',
'upload_date': '20100323',
},
'params': {
'matchtitle': '^The Bomb$', # To test, we want just one video
}
}
def _real_extract(self, url):
mobj = re.match(self._VALID_URL, url)
playlist_id = mobj.group('id')
fstr = compat_urllib_parse.quote("InternalBandUrl eq '%s'" % playlist_id)
json_url = 'http://urort.p3.no/breeze/urort/TrackDtos?$filter=' + fstr
songs = self._download_json(json_url, playlist_id)
print(songs[0])
entries = [{
'id': '%d-%s' % (s['BandId'], s['$id']),
'title': s['Title'],
'url': s['TrackUrl'],
'ext': 'mp3',
'uploader_id': playlist_id,
'uploader': s.get('BandName', playlist_id),
'like_count': s.get('LikeCount'),
'thumbnail': 'http://urort.p3.no/cloud/images/%s' % s['Image'],
'upload_date': unified_strdate(s.get('Released')),
} for s in songs]
return {
'_type': 'playlist',
'id': playlist_id,
'title': playlist_id,
'entries': entries,
}