|
|
- from __future__ import unicode_literals
-
- import json
- import re
-
- from .common import InfoExtractor
- from ..utils import (
- ExtractorError,
- int_or_none,
- orderedSet,
- )
-
-
- class DeezerPlaylistIE(InfoExtractor):
- _VALID_URL = r'https?://(?:www\.)?deezer\.com/playlist/(?P<id>[0-9]+)'
- _TEST = {
- 'url': 'http://www.deezer.com/playlist/176747451',
- 'info_dict': {
- 'id': '176747451',
- 'title': 'Best!',
- 'uploader': 'Anonymous',
- 'thumbnail': 're:^https?://cdn-images.deezer.com/images/cover/.*\.jpg$',
- },
- 'playlist_count': 30,
- 'skip': 'Only available in .de',
- }
-
- def _real_extract(self, url):
- if 'test' not in self._downloader.params:
- self._downloader.report_warning('For now, this extractor only supports the 30 second previews. Patches welcome!')
-
- mobj = re.match(self._VALID_URL, url)
- playlist_id = mobj.group('id')
-
- webpage = self._download_webpage(url, playlist_id)
- geoblocking_msg = self._html_search_regex(
- r'<p class="soon-txt">(.*?)</p>', webpage, 'geoblocking message',
- default=None)
- if geoblocking_msg is not None:
- raise ExtractorError(
- 'Deezer said: %s' % geoblocking_msg, expected=True)
-
- data_json = self._search_regex(
- r'naboo\.display\(\'[^\']+\',\s*(.*?)\);\n', webpage, 'data JSON')
- data = json.loads(data_json)
-
- playlist_title = data.get('DATA', {}).get('TITLE')
- playlist_uploader = data.get('DATA', {}).get('PARENT_USERNAME')
- playlist_thumbnail = self._search_regex(
- r'<img id="naboo_playlist_image".*?src="([^"]+)"', webpage,
- 'playlist thumbnail')
-
- preview_pattern = self._search_regex(
- r"var SOUND_PREVIEW_GATEWAY\s*=\s*'([^']+)';", webpage,
- 'preview URL pattern', fatal=False)
- entries = []
- for s in data['SONGS']['data']:
- puid = s['MD5_ORIGIN']
- preview_video_url = preview_pattern.\
- replace('{0}', puid[0]).\
- replace('{1}', puid).\
- replace('{2}', s['MEDIA_VERSION'])
- formats = [{
- 'format_id': 'preview',
- 'url': preview_video_url,
- 'preference': -100, # Only the first 30 seconds
- 'ext': 'mp3',
- }]
- self._sort_formats(formats)
- artists = ', '.join(
- orderedSet(a['ART_NAME'] for a in s['ARTISTS']))
- entries.append({
- 'id': s['SNG_ID'],
- 'duration': int_or_none(s.get('DURATION')),
- 'title': '%s - %s' % (artists, s['SNG_TITLE']),
- 'uploader': s['ART_NAME'],
- 'uploader_id': s['ART_ID'],
- 'age_limit': 16 if s.get('EXPLICIT_LYRICS') == '1' else 0,
- 'formats': formats,
- })
-
- return {
- '_type': 'playlist',
- 'id': playlist_id,
- 'title': playlist_title,
- 'uploader': playlist_uploader,
- 'thumbnail': playlist_thumbnail,
- 'entries': entries,
- }
|