Added yt_dlp directly, added rename format options, added xclip clipboard subproc, added copy name context menu option
This commit is contained in:
145
plugins/youtube_download/yt_dlp/extractor/tubitv.py
Normal file
145
plugins/youtube_download/yt_dlp/extractor/tubitv.py
Normal file
@@ -0,0 +1,145 @@
|
||||
# coding: utf-8
|
||||
from __future__ import unicode_literals
|
||||
|
||||
import re
|
||||
|
||||
from .common import InfoExtractor
|
||||
from ..utils import (
|
||||
ExtractorError,
|
||||
int_or_none,
|
||||
js_to_json,
|
||||
sanitized_Request,
|
||||
urlencode_postdata,
|
||||
)
|
||||
|
||||
|
||||
class TubiTvIE(InfoExtractor):
|
||||
_VALID_URL = r'''(?x)
|
||||
(?:
|
||||
tubitv:|
|
||||
https?://(?:www\.)?tubitv\.com/(?:video|movies|tv-shows)/
|
||||
)
|
||||
(?P<id>[0-9]+)'''
|
||||
_LOGIN_URL = 'http://tubitv.com/login'
|
||||
_NETRC_MACHINE = 'tubitv'
|
||||
_GEO_COUNTRIES = ['US']
|
||||
_TESTS = [{
|
||||
'url': 'http://tubitv.com/video/283829/the_comedian_at_the_friday',
|
||||
'md5': '43ac06be9326f41912dc64ccf7a80320',
|
||||
'info_dict': {
|
||||
'id': '283829',
|
||||
'ext': 'mp4',
|
||||
'title': 'The Comedian at The Friday',
|
||||
'description': 'A stand up comedian is forced to look at the decisions in his life while on a one week trip to the west coast.',
|
||||
'uploader_id': 'bc168bee0d18dd1cb3b86c68706ab434',
|
||||
},
|
||||
}, {
|
||||
'url': 'http://tubitv.com/tv-shows/321886/s01_e01_on_nom_stories',
|
||||
'only_matching': True,
|
||||
}, {
|
||||
'url': 'http://tubitv.com/movies/383676/tracker',
|
||||
'only_matching': True,
|
||||
}, {
|
||||
'url': 'https://tubitv.com/movies/560057/penitentiary?start=true',
|
||||
'info_dict': {
|
||||
'id': '560057',
|
||||
'ext': 'mp4',
|
||||
'title': 'Penitentiary',
|
||||
'description': 'md5:8d2fc793a93cc1575ff426fdcb8dd3f9',
|
||||
'uploader_id': 'd8fed30d4f24fcb22ec294421b9defc2',
|
||||
'release_year': 1979,
|
||||
},
|
||||
'params': {
|
||||
'skip_download': True,
|
||||
},
|
||||
}]
|
||||
|
||||
def _login(self):
|
||||
username, password = self._get_login_info()
|
||||
if username is None:
|
||||
return
|
||||
self.report_login()
|
||||
form_data = {
|
||||
'username': username,
|
||||
'password': password,
|
||||
}
|
||||
payload = urlencode_postdata(form_data)
|
||||
request = sanitized_Request(self._LOGIN_URL, payload)
|
||||
request.add_header('Content-Type', 'application/x-www-form-urlencoded')
|
||||
login_page = self._download_webpage(
|
||||
request, None, False, 'Wrong login info')
|
||||
if not re.search(r'id="tubi-logout"', login_page):
|
||||
raise ExtractorError(
|
||||
'Login failed (invalid username/password)', expected=True)
|
||||
|
||||
def _real_initialize(self):
|
||||
self._login()
|
||||
|
||||
def _real_extract(self, url):
|
||||
video_id = self._match_id(url)
|
||||
video_data = self._download_json(
|
||||
'http://tubitv.com/oz/videos/%s/content' % video_id, video_id)
|
||||
title = video_data['title']
|
||||
|
||||
formats = []
|
||||
url = video_data['url']
|
||||
# URL can be sometimes empty. Does this only happen when there is DRM?
|
||||
if url:
|
||||
formats = self._extract_m3u8_formats(
|
||||
self._proto_relative_url(url),
|
||||
video_id, 'mp4', 'm3u8_native')
|
||||
self._sort_formats(formats)
|
||||
|
||||
thumbnails = []
|
||||
for thumbnail_url in video_data.get('thumbnails', []):
|
||||
if not thumbnail_url:
|
||||
continue
|
||||
thumbnails.append({
|
||||
'url': self._proto_relative_url(thumbnail_url),
|
||||
})
|
||||
|
||||
subtitles = {}
|
||||
for sub in video_data.get('subtitles', []):
|
||||
sub_url = sub.get('url')
|
||||
if not sub_url:
|
||||
continue
|
||||
subtitles.setdefault(sub.get('lang', 'English'), []).append({
|
||||
'url': self._proto_relative_url(sub_url),
|
||||
})
|
||||
|
||||
return {
|
||||
'id': video_id,
|
||||
'title': title,
|
||||
'formats': formats,
|
||||
'subtitles': subtitles,
|
||||
'thumbnails': thumbnails,
|
||||
'description': video_data.get('description'),
|
||||
'duration': int_or_none(video_data.get('duration')),
|
||||
'uploader_id': video_data.get('publisher_id'),
|
||||
'release_year': int_or_none(video_data.get('year')),
|
||||
}
|
||||
|
||||
|
||||
class TubiTvShowIE(InfoExtractor):
|
||||
_VALID_URL = r'https?://(?:www\.)?tubitv\.com/series/[0-9]+/(?P<show_name>[^/?#]+)'
|
||||
_TESTS = [{
|
||||
'url': 'https://tubitv.com/series/3936/the-joy-of-painting-with-bob-ross?start=true',
|
||||
'playlist_mincount': 390,
|
||||
'info_dict': {
|
||||
'id': 'the-joy-of-painting-with-bob-ross',
|
||||
}
|
||||
}]
|
||||
|
||||
def _entries(self, show_url, show_name):
|
||||
show_webpage = self._download_webpage(show_url, show_name)
|
||||
show_json = self._parse_json(self._search_regex(
|
||||
r"window\.__data\s*=\s*({.+?});\s*</script>",
|
||||
show_webpage, 'data',), show_name, transform_source=js_to_json)['video']
|
||||
for episode_id in show_json['fullContentById'].keys():
|
||||
yield self.url_result(
|
||||
'tubitv:%s' % episode_id,
|
||||
ie=TubiTvIE.ie_key(), video_id=episode_id)
|
||||
|
||||
def _real_extract(self, url):
|
||||
show_name = self._match_valid_url(url).group('show_name')
|
||||
return self.playlist_result(self._entries(url, show_name), playlist_id=show_name)
|
Reference in New Issue
Block a user