mirror of
https://github.com/yt-dlp/yt-dlp.git
synced 2024-11-01 16:57:20 -04:00
22a6f15061
From now on, test_download will run these tests. That means we benefit not only from the networking setup in there, but also from the other tests (for example test_all_urls to find problems with _VALID_URLs).
99 lines
3.1 KiB
Python
99 lines
3.1 KiB
Python
from __future__ import unicode_literals
|
|
|
|
import re
|
|
import json
|
|
import itertools
|
|
|
|
from .common import InfoExtractor
|
|
from ..utils import unified_strdate
|
|
|
|
|
|
class VineIE(InfoExtractor):
|
|
_VALID_URL = r'https?://(?:www\.)?vine\.co/v/(?P<id>\w+)'
|
|
_TEST = {
|
|
'url': 'https://vine.co/v/b9KOOWX7HUx',
|
|
'md5': '2f36fed6235b16da96ce9b4dc890940d',
|
|
'info_dict': {
|
|
'id': 'b9KOOWX7HUx',
|
|
'ext': 'mp4',
|
|
'title': 'Chicken.',
|
|
'description': 'Chicken.',
|
|
'upload_date': '20130519',
|
|
'uploader': 'Jack Dorsey',
|
|
'uploader_id': '76',
|
|
},
|
|
}
|
|
|
|
def _real_extract(self, url):
|
|
mobj = re.match(self._VALID_URL, url)
|
|
video_id = mobj.group('id')
|
|
|
|
webpage = self._download_webpage('https://vine.co/v/' + video_id, video_id)
|
|
|
|
data = json.loads(self._html_search_regex(
|
|
r'window\.POST_DATA = { %s: ({.+?}) }' % video_id, webpage, 'vine data'))
|
|
|
|
formats = [
|
|
{
|
|
'url': data['videoLowURL'],
|
|
'ext': 'mp4',
|
|
'format_id': 'low',
|
|
},
|
|
{
|
|
'url': data['videoUrl'],
|
|
'ext': 'mp4',
|
|
'format_id': 'standard',
|
|
}
|
|
]
|
|
|
|
return {
|
|
'id': video_id,
|
|
'title': self._og_search_title(webpage),
|
|
'description': data['description'],
|
|
'thumbnail': data['thumbnailUrl'],
|
|
'upload_date': unified_strdate(data['created']),
|
|
'uploader': data['username'],
|
|
'uploader_id': data['userIdStr'],
|
|
'like_count': data['likes']['count'],
|
|
'comment_count': data['comments']['count'],
|
|
'repost_count': data['reposts']['count'],
|
|
'formats': formats,
|
|
}
|
|
|
|
|
|
class VineUserIE(InfoExtractor):
|
|
IE_NAME = 'vine:user'
|
|
_VALID_URL = r'(?:https?://)?vine\.co/(?P<user>[^/]+)/?(\?.*)?$'
|
|
_VINE_BASE_URL = "https://vine.co/"
|
|
_TEST = {
|
|
'url': 'https://vine.co/Visa',
|
|
'info_dict': {
|
|
'id': 'Visa',
|
|
},
|
|
'playlist_mincount': 47,
|
|
}
|
|
|
|
def _real_extract(self, url):
|
|
mobj = re.match(self._VALID_URL, url)
|
|
user = mobj.group('user')
|
|
|
|
profile_url = "%sapi/users/profiles/vanity/%s" % (
|
|
self._VINE_BASE_URL, user)
|
|
profile_data = self._download_json(
|
|
profile_url, user, note='Downloading user profile data')
|
|
|
|
user_id = profile_data['data']['userId']
|
|
timeline_data = []
|
|
for pagenum in itertools.count(1):
|
|
timeline_url = "%sapi/timelines/users/%s?page=%s" % (
|
|
self._VINE_BASE_URL, user_id, pagenum)
|
|
timeline_page = self._download_json(
|
|
timeline_url, user, note='Downloading page %d' % pagenum)
|
|
timeline_data.extend(timeline_page['data']['records'])
|
|
if timeline_page['data']['nextPage'] is None:
|
|
break
|
|
|
|
entries = [
|
|
self.url_result(e['permalinkUrl'], 'Vine') for e in timeline_data]
|
|
return self.playlist_result(entries, user)
|