1
1
mirror of https://github.com/ytdl-org/youtube-dl synced 2024-11-25 02:36:50 +01:00
youtube-dl/youtube_dl/extractor/vidme.py

298 lines
9.7 KiB
Python
Raw Normal View History

2014-07-31 15:26:52 +02:00
from __future__ import unicode_literals
import itertools
from .common import InfoExtractor
2017-08-04 19:00:21 +02:00
from ..compat import (
compat_HTTPError,
compat_str,
)
2014-07-31 15:26:52 +02:00
from ..utils import (
ExtractorError,
2014-07-31 15:26:52 +02:00
int_or_none,
float_or_none,
2015-09-09 17:02:28 +02:00
parse_iso8601,
2014-07-31 15:26:52 +02:00
)
class VidmeIE(InfoExtractor):
2016-02-05 18:58:26 +01:00
IE_NAME = 'vidme'
_VALID_URL = r'https?://vid\.me/(?:e/)?(?P<id>[\da-zA-Z]{,5})(?:[^\da-zA-Z]|$)'
_TESTS = [{
2014-07-31 15:26:52 +02:00
'url': 'https://vid.me/QNB',
2015-10-20 16:18:23 +02:00
'md5': 'f42d05e7149aeaec5c037b17e5d3dc82',
2014-07-31 15:26:52 +02:00
'info_dict': {
'id': 'QNB',
'ext': 'mp4',
'title': 'Fishing for piranha - the easy way',
'description': 'source: https://www.facebook.com/photo.php?v=312276045600871',
'thumbnail': r're:^https?://.*\.jpg',
2014-07-31 15:26:52 +02:00
'timestamp': 1406313244,
'upload_date': '20140725',
'age_limit': 0,
'duration': 119.92,
'view_count': int,
'like_count': int,
2015-09-09 17:02:28 +02:00
'comment_count': int,
},
}, {
'url': 'https://vid.me/Gc6M',
'md5': 'f42d05e7149aeaec5c037b17e5d3dc82',
'info_dict': {
'id': 'Gc6M',
'ext': 'mp4',
'title': 'O Mere Dil ke chain - Arnav and Khushi VM',
'thumbnail': r're:^https?://.*\.jpg',
2015-09-09 17:02:28 +02:00
'timestamp': 1441211642,
'upload_date': '20150902',
'uploader': 'SunshineM',
'uploader_id': '3552827',
'age_limit': 0,
'duration': 223.72,
2015-09-09 17:02:28 +02:00
'view_count': int,
'like_count': int,
'comment_count': int,
},
'params': {
'skip_download': True,
},
}, {
# tests uploader field
'url': 'https://vid.me/4Iib',
'info_dict': {
'id': '4Iib',
'ext': 'mp4',
'title': 'The Carver',
'description': 'md5:e9c24870018ae8113be936645b93ba3c',
'thumbnail': r're:^https?://.*\.jpg',
'timestamp': 1433203629,
'upload_date': '20150602',
'uploader': 'Thomas',
'uploader_id': '109747',
'age_limit': 0,
'duration': 97.859999999999999,
'view_count': int,
'like_count': int,
2015-09-09 17:02:28 +02:00
'comment_count': int,
},
'params': {
'skip_download': True,
2014-07-31 15:26:52 +02:00
},
}, {
# nsfw test from http://naked-yogi.tumblr.com/post/118312946248/naked-smoking-stretching
'url': 'https://vid.me/e/Wmur',
'info_dict': {
'id': 'Wmur',
'ext': 'mp4',
'title': 'naked smoking & stretching',
'thumbnail': r're:^https?://.*\.jpg',
'timestamp': 1430931613,
'upload_date': '20150506',
'uploader': 'naked-yogi',
'uploader_id': '1638622',
'age_limit': 18,
'duration': 653.26999999999998,
'view_count': int,
'like_count': int,
'comment_count': int,
},
'params': {
'skip_download': True,
},
2015-10-17 19:01:24 +02:00
}, {
# nsfw, user-disabled
'url': 'https://vid.me/dzGJ',
'only_matching': True,
}, {
# suspended
'url': 'https://vid.me/Ox3G',
'only_matching': True,
2015-10-26 14:42:17 +01:00
}, {
# deleted
'url': 'https://vid.me/KTPm',
'only_matching': True,
}, {
# no formats in the API response
'url': 'https://vid.me/e5g',
'info_dict': {
'id': 'e5g',
'ext': 'mp4',
'title': 'Video upload (e5g)',
'thumbnail': r're:^https?://.*\.jpg',
'timestamp': 1401480195,
'upload_date': '20140530',
'uploader': None,
'uploader_id': None,
'age_limit': 0,
'duration': 483,
'view_count': int,
'like_count': int,
'comment_count': int,
},
'params': {
'skip_download': True,
},
}]
2014-07-31 15:26:52 +02:00
def _real_extract(self, url):
2015-03-05 22:34:56 +01:00
video_id = self._match_id(url)
2015-09-09 17:02:28 +02:00
try:
response = self._download_json(
'https://api.vid.me/videoByUrl/%s' % video_id, video_id)
except ExtractorError as e:
if isinstance(e.cause, compat_HTTPError) and e.cause.code == 400:
response = self._parse_json(e.cause.read(), video_id)
else:
raise
2015-09-09 17:02:28 +02:00
error = response.get('error')
if error:
raise ExtractorError(
'%s returned error: %s' % (self.IE_NAME, error), expected=True)
2014-07-31 15:26:52 +02:00
video = response['video']
2014-07-31 15:26:52 +02:00
2015-10-26 14:42:17 +01:00
if video.get('state') == 'deleted':
raise ExtractorError(
'Vidme said: Sorry, this video has been deleted.',
expected=True)
if video.get('state') in ('user-disabled', 'suspended'):
raise ExtractorError(
'Vidme said: This video has been suspended either due to a copyright claim, '
'or for violating the terms of use.',
expected=True)
2017-08-04 19:00:21 +02:00
formats = []
for f in video.get('formats', []):
format_url = f.get('uri')
if not format_url or not isinstance(format_url, compat_str):
continue
format_type = f.get('type')
if format_type == 'dash':
formats.extend(self._extract_mpd_formats(
format_url, video_id, mpd_id='dash', fatal=False))
elif format_type == 'hls':
formats.extend(self._extract_m3u8_formats(
format_url, video_id, 'mp4', entry_protocol='m3u8_native',
m3u8_id='hls', fatal=False))
else:
formats.append({
'format_id': f.get('type'),
'url': format_url,
'width': int_or_none(f.get('width')),
'height': int_or_none(f.get('height')),
'preference': 0 if f.get('type', '').endswith(
'clip') else 1,
})
if not formats and video.get('complete_url'):
formats.append({
'url': video.get('complete_url'),
'width': int_or_none(video.get('width')),
'height': int_or_none(video.get('height')),
})
2015-09-09 17:02:28 +02:00
self._sort_formats(formats)
2014-07-31 15:26:52 +02:00
title = video['title']
description = video.get('description')
thumbnail = video.get('thumbnail_url')
timestamp = parse_iso8601(video.get('date_created'), ' ')
uploader = video.get('user', {}).get('username')
uploader_id = video.get('user', {}).get('user_id')
age_limit = 18 if video.get('nsfw') is True else 0
duration = float_or_none(video.get('duration'))
view_count = int_or_none(video.get('view_count'))
like_count = int_or_none(video.get('likes_count'))
comment_count = int_or_none(video.get('comment_count'))
2014-07-31 15:26:52 +02:00
return {
'id': video_id,
'title': title or 'Video upload (%s)' % video_id,
2014-07-31 15:26:52 +02:00
'description': description,
'thumbnail': thumbnail,
'uploader': uploader,
'uploader_id': uploader_id,
'age_limit': age_limit,
2014-07-31 15:26:52 +02:00
'timestamp': timestamp,
'duration': duration,
'view_count': view_count,
'like_count': like_count,
2015-09-09 17:02:28 +02:00
'comment_count': comment_count,
'formats': formats,
2014-07-31 15:26:52 +02:00
}
2016-02-05 18:57:35 +01:00
class VidmeListBaseIE(InfoExtractor):
# Max possible limit according to https://docs.vid.me/#api-Videos-List
_LIMIT = 100
def _entries(self, user_id, user_name):
for page_num in itertools.count(1):
page = self._download_json(
2016-02-05 18:57:35 +01:00
'https://api.vid.me/videos/%s?user=%s&limit=%d&offset=%d'
% (self._API_ITEM, user_id, self._LIMIT, (page_num - 1) * self._LIMIT),
user_name, 'Downloading user %s page %d' % (self._API_ITEM, page_num))
videos = page.get('videos', [])
if not videos:
break
for video in videos:
video_url = video.get('full_url') or video.get('embed_url')
if video_url:
yield self.url_result(video_url, VidmeIE.ie_key())
total = int_or_none(page.get('page', {}).get('total'))
if total and self._LIMIT * page_num >= total:
break
def _real_extract(self, url):
user_name = self._match_id(url)
user_id = self._download_json(
'https://api.vid.me/userByUsername?username=%s' % user_name,
user_name)['user']['user_id']
2016-02-05 18:57:35 +01:00
return self.playlist_result(
self._entries(user_id, user_name), user_id,
'%s - %s' % (user_name, self._TITLE))
class VidmeUserIE(VidmeListBaseIE):
2016-02-05 18:58:26 +01:00
IE_NAME = 'vidme:user'
2017-08-28 21:35:57 +02:00
_VALID_URL = r'https?://vid\.me/(?:e/)?(?P<id>[\da-zA-Z_-]{6,})(?!/likes)(?:[^\da-zA-Z_-]|$)'
2016-02-05 18:57:35 +01:00
_API_ITEM = 'list'
_TITLE = 'Videos'
2017-08-28 21:35:57 +02:00
_TESTS = [{
'url': 'https://vid.me/MasakoX',
2016-02-05 18:57:35 +01:00
'info_dict': {
2017-08-28 21:35:57 +02:00
'id': '16112341',
'title': 'MasakoX - %s' % _TITLE,
2016-02-05 18:57:35 +01:00
},
2017-08-28 21:35:57 +02:00
'playlist_mincount': 191,
}, {
'url': 'https://vid.me/unsQuare_netWork',
'only_matching': True,
2017-08-28 21:35:57 +02:00
}]
2016-02-05 18:57:35 +01:00
class VidmeUserLikesIE(VidmeListBaseIE):
2016-02-05 18:58:26 +01:00
IE_NAME = 'vidme:user:likes'
2017-08-28 21:35:57 +02:00
_VALID_URL = r'https?://vid\.me/(?:e/)?(?P<id>[\da-zA-Z_-]{6,})/likes'
2016-02-05 18:57:35 +01:00
_API_ITEM = 'likes'
_TITLE = 'Likes'
2017-08-28 21:35:57 +02:00
_TESTS = [{
2016-02-05 18:57:35 +01:00
'url': 'https://vid.me/ErinAlexis/likes',
'info_dict': {
'id': '6483530',
'title': 'ErinAlexis - %s' % _TITLE,
},
'playlist_mincount': 415,
2017-08-28 21:35:57 +02:00
}, {
'url': 'https://vid.me/Kaleidoscope-Ish/likes',
'only_matching': True,
2017-08-28 21:35:57 +02:00
}]