mirror of
https://github.com/ytdl-org/youtube-dl.git
synced 2024-11-23 19:35:19 +01:00
[flipagram] Add extractor
This commit is contained in:
parent
89e2fff2b7
commit
3fee7f636c
|
@ -837,6 +837,13 @@ class InfoExtractor(object):
|
|||
'title': unescapeHTML(json_ld.get('headline')),
|
||||
'description': unescapeHTML(json_ld.get('articleBody')),
|
||||
})
|
||||
elif item_type == 'VideoObject':
|
||||
info.update({
|
||||
'title': unescapeHTML(json_ld.get('name')),
|
||||
'description': unescapeHTML(json_ld.get('description')),
|
||||
'upload_date': unified_strdate(json_ld.get('upload_date')),
|
||||
'url': unescapeHTML(json_ld.get('contentUrl')),
|
||||
})
|
||||
return dict((k, v) for k, v in info.items() if v is not None)
|
||||
|
||||
@staticmethod
|
||||
|
|
|
@ -256,6 +256,7 @@ from .fivemin import FiveMinIE
|
|||
from .fivetv import FiveTVIE
|
||||
from .fktv import FKTVIE
|
||||
from .flickr import FlickrIE
|
||||
from .flipagram import FlipagramIE
|
||||
from .folketinget import FolketingetIE
|
||||
from .footyroom import FootyRoomIE
|
||||
from .formula1 import Formula1IE
|
||||
|
|
103
youtube_dl/extractor/flipagram.py
Normal file
103
youtube_dl/extractor/flipagram.py
Normal file
|
@ -0,0 +1,103 @@
|
|||
# coding: utf-8
|
||||
from __future__ import unicode_literals
|
||||
|
||||
from .common import InfoExtractor
|
||||
|
||||
from ..utils import (
|
||||
int_or_none,
|
||||
parse_iso8601,
|
||||
unified_strdate,
|
||||
unified_timestamp,
|
||||
)
|
||||
|
||||
|
||||
class FlipagramIE(InfoExtractor):
|
||||
_VALID_URL = r'https?://(?:www\.)?flipagram\.com/f/(?P<id>[^/?_]+)'
|
||||
_TESTS = [{
|
||||
'url': 'https://flipagram.com/f/myrWjW9RJw',
|
||||
'md5': '541988fb6c4c7c375215ea22a4a21841',
|
||||
'info_dict': {
|
||||
'id': 'myrWjW9RJw',
|
||||
'title': 'Flipagram by crystaldolce featuring King and Lionheart by Of Monsters and Men',
|
||||
'description': 'Herbie\'s first bannana🍌🐢🍌. #animals #pets #reptile #tortoise #sulcata #tort #justatreat #snacktime #bannanas #rescuepets #ofmonstersandmen @animals',
|
||||
'ext': 'mp4',
|
||||
'uploader': 'Crystal Dolce',
|
||||
'creator': 'Crystal Dolce',
|
||||
'uploader_id': 'crystaldolce',
|
||||
}
|
||||
}, {
|
||||
'url': 'https://flipagram.com/f/nyvTSJMKId',
|
||||
'only_matching': True,
|
||||
}]
|
||||
|
||||
def _real_extract(self, url):
|
||||
video_id = self._match_id(url)
|
||||
webpage = self._download_webpage(url, video_id)
|
||||
|
||||
self.report_extraction(video_id)
|
||||
user_data = self._parse_json(self._search_regex(r'window.reactH2O\s*=\s*({.+});', webpage, 'user data'), video_id)
|
||||
content_data = self._search_json_ld(webpage, video_id)
|
||||
|
||||
flipagram = user_data.get('flipagram', {})
|
||||
counts = flipagram.get('counts', {})
|
||||
user = flipagram.get('user', {})
|
||||
video = flipagram.get('video', {})
|
||||
|
||||
thumbnails = []
|
||||
for cover in flipagram.get('covers', []):
|
||||
if not cover.get('url'):
|
||||
continue
|
||||
thumbnails.append({
|
||||
'url': self._proto_relative_url(cover.get('url')),
|
||||
'width': int_or_none(cover.get('width')),
|
||||
'height': int_or_none(cover.get('height')),
|
||||
})
|
||||
|
||||
# Note that this only retrieves comments that are initally loaded.
|
||||
# For videos with large amounts of comments, most won't be retrieved.
|
||||
comments = []
|
||||
for comment in user_data.get('comments', {}).get(video_id, {}).get('items', []):
|
||||
text = comment.get('comment', [])
|
||||
comments.append({
|
||||
'author': comment.get('user', {}).get('name'),
|
||||
'author_id': comment.get('user', {}).get('username'),
|
||||
'id': comment.get('id'),
|
||||
'text': text[0] if text else '',
|
||||
'timestamp': unified_timestamp(comment.get('created', '')),
|
||||
})
|
||||
|
||||
tags = [tag for item in flipagram['story'][1:] for tag in item]
|
||||
|
||||
formats = []
|
||||
if flipagram.get('music', {}).get('track', {}).get('previewUrl', {}):
|
||||
formats.append({
|
||||
'url': flipagram.get('music').get('track').get('previewUrl'),
|
||||
'ext': 'm4a',
|
||||
'vcodec': 'none',
|
||||
})
|
||||
|
||||
formats.append({
|
||||
'url': video.get('url'),
|
||||
'ext': 'mp4',
|
||||
'width': int_or_none(video.get('width')),
|
||||
'height': int_or_none(video.get('height')),
|
||||
'filesize': int_or_none(video.get('size')),
|
||||
})
|
||||
|
||||
return {
|
||||
'id': video_id,
|
||||
'title': content_data['title'],
|
||||
'formats': formats,
|
||||
'thumbnails': thumbnails,
|
||||
'description': content_data.get('description'),
|
||||
'uploader': user.get('name'),
|
||||
'creator': user.get('name'),
|
||||
'timestamp': parse_iso8601(flipagram.get('iso801Created')),
|
||||
'upload_date': unified_strdate(flipagram.get('created')),
|
||||
'uploader_id': user.get('username'),
|
||||
'view_count': int_or_none(counts.get('plays')),
|
||||
'repost_count': int_or_none(counts.get('reflips')),
|
||||
'comment_count': int_or_none(counts.get('comments')),
|
||||
'comments': comments,
|
||||
'tags': tags,
|
||||
}
|
Loading…
Reference in New Issue
Block a user