1
0
Fork 0
mirror of https://github.com/ytdl-org/youtube-dl.git synced 2024-11-25 19:52:11 +00:00
youtube-dl/youtube_dl/extractor/appleconnect.py

54 lines
1.9 KiB
Python
Raw Normal View History

# coding: utf-8
from __future__ import unicode_literals
from .common import InfoExtractor
from ..utils import (
str_to_int,
ExtractorError
)
class AppleConnectIE(InfoExtractor):
2021-06-16 21:12:13 +00:00
_VALID_URL = r'https?://itunes\.apple\.com/\w{0,2}/?post/(?:id)?sa\.(?P<id>[\w-]+)'
_TESTS = [{
'url': 'https://itunes.apple.com/us/post/idsa.4ab17a39-2720-11e5-96c5-a5b38f6c42d3',
2021-06-16 21:12:13 +00:00
'md5': 'c1d41f72c8bcaf222e089434619316e4',
'info_dict': {
'id': '4ab17a39-2720-11e5-96c5-a5b38f6c42d3',
'ext': 'm4v',
'title': 'Energy',
'uploader': 'Drake',
2017-04-27 20:47:25 +00:00
'thumbnail': r're:^https?://.*\.jpg$',
'upload_date': '20150710',
'timestamp': 1436545535,
},
2021-06-16 21:12:13 +00:00
}, {
'url': 'https://itunes.apple.com/us/post/sa.0fe0229f-2457-11e5-9f40-1bb645f2d5d9',
'only_matching': True,
}]
def _real_extract(self, url):
video_id = self._match_id(url)
webpage = self._download_webpage(url, video_id)
try:
video_json = self._html_search_regex(
r'class="auc-video-data">(\{.*?\})', webpage, 'json')
except ExtractorError:
raise ExtractorError('This post doesn\'t contain a video', expected=True)
video_data = self._parse_json(video_json, video_id)
timestamp = str_to_int(self._html_search_regex(r'data-timestamp="(\d+)"', webpage, 'timestamp'))
2021-06-16 21:12:13 +00:00
like_count = str_to_int(self._html_search_regex(r'(\d+) Loves', webpage, 'like count', default=None))
return {
'id': video_id,
'url': video_data['sslSrc'],
'title': video_data['title'],
'description': video_data['description'],
'uploader': video_data['artistName'],
'thumbnail': video_data['artworkUrl'],
'timestamp': timestamp,
'like_count': like_count,
}