- from __future__ import unicode_literals
-
- from .common import InfoExtractor
- from ..utils import (
- float_or_none,
- parse_iso8601,
- )
-
-
- class ClypIE(InfoExtractor):
- _VALID_URL = r'https?://(?:www\.)?clyp\.it/(?P<id>[a-z0-9]+)'
- _TEST = {
- 'url': 'https://clyp.it/ojz2wfah',
- 'md5': '1d4961036c41247ecfdcc439c0cddcbb',
- 'info_dict': {
- 'id': 'ojz2wfah',
- 'ext': 'mp3',
- 'title': 'Krisson80 - bits wip wip',
- 'description': '#Krisson80BitsWipWip #chiptune\n#wip',
- 'duration': 263.21,
- 'timestamp': 1443515251,
- 'upload_date': '20150929',
- },
- }
-
- def _real_extract(self, url):
- audio_id = self._match_id(url)
-
- metadata = self._download_json(
- 'https://api.clyp.it/%s' % audio_id, audio_id)
-
- formats = []
- for secure in ('', 'Secure'):
- for ext in ('Ogg', 'Mp3'):
- format_id = '%s%s' % (secure, ext)
- format_url = metadata.get('%sUrl' % format_id)
- if format_url:
- formats.append({
- 'url': format_url,
- 'format_id': format_id,
- 'vcodec': 'none',
- })
- self._sort_formats(formats)
-
- title = metadata['Title']
- description = metadata.get('Description')
- duration = float_or_none(metadata.get('Duration'))
- timestamp = parse_iso8601(metadata.get('DateCreated'))
-
- return {
- 'id': audio_id,
- 'title': title,
- 'description': description,
- 'duration': duration,
- 'timestamp': timestamp,
- 'formats': formats,
- }
|