mirror of
https://github.com/yt-dlp/yt-dlp.git
synced 2024-11-30 12:20:23 +00:00
15707c7e02
encode_dict functionality has been improved and moved directly into compat_urllib_parse_urlencode All occurrences of compat_urllib_parse.urlencode throughout the codebase have been replaced by compat_urllib_parse_urlencode Closes #8974
191 lines
8.0 KiB
Python
191 lines
8.0 KiB
Python
# coding: utf-8
|
|
from __future__ import unicode_literals
|
|
|
|
from .common import InfoExtractor
|
|
from ..utils import (
|
|
clean_html,
|
|
determine_ext,
|
|
int_or_none,
|
|
sanitized_Request,
|
|
ExtractorError,
|
|
urlencode_postdata
|
|
)
|
|
|
|
|
|
class FunimationIE(InfoExtractor):
|
|
_VALID_URL = r'https?://(?:www\.)?funimation\.com/shows/[^/]+/videos/(?:official|promotional)/(?P<id>[^/?#&]+)'
|
|
|
|
_NETRC_MACHINE = 'funimation'
|
|
|
|
_TESTS = [{
|
|
'url': 'http://www.funimation.com/shows/air/videos/official/breeze',
|
|
'info_dict': {
|
|
'id': '658',
|
|
'display_id': 'breeze',
|
|
'ext': 'mp4',
|
|
'title': 'Air - 1 - Breeze',
|
|
'description': 'md5:1769f43cd5fc130ace8fd87232207892',
|
|
'thumbnail': 're:https?://.*\.jpg',
|
|
},
|
|
}, {
|
|
'url': 'http://www.funimation.com/shows/hacksign/videos/official/role-play',
|
|
'info_dict': {
|
|
'id': '31128',
|
|
'display_id': 'role-play',
|
|
'ext': 'mp4',
|
|
'title': '.hack//SIGN - 1 - Role Play',
|
|
'description': 'md5:b602bdc15eef4c9bbb201bb6e6a4a2dd',
|
|
'thumbnail': 're:https?://.*\.jpg',
|
|
},
|
|
}, {
|
|
'url': 'http://www.funimation.com/shows/attack-on-titan-junior-high/videos/promotional/broadcast-dub-preview',
|
|
'info_dict': {
|
|
'id': '9635',
|
|
'display_id': 'broadcast-dub-preview',
|
|
'ext': 'mp4',
|
|
'title': 'Attack on Titan: Junior High - Broadcast Dub Preview',
|
|
'description': 'md5:f8ec49c0aff702a7832cd81b8a44f803',
|
|
'thumbnail': 're:https?://.*\.(?:jpg|png)',
|
|
},
|
|
}]
|
|
|
|
def _login(self):
|
|
(username, password) = self._get_login_info()
|
|
if username is None:
|
|
return
|
|
data = urlencode_postdata({
|
|
'email_field': username,
|
|
'password_field': password,
|
|
})
|
|
login_request = sanitized_Request('http://www.funimation.com/login', data, headers={
|
|
'User-Agent': 'Mozilla/5.0 (Windows NT 5.2; WOW64; rv:42.0) Gecko/20100101 Firefox/42.0',
|
|
'Content-Type': 'application/x-www-form-urlencoded'
|
|
})
|
|
login_page = self._download_webpage(
|
|
login_request, None, 'Logging in as %s' % username)
|
|
if any(p in login_page for p in ('funimation.com/logout', '>Log Out<')):
|
|
return
|
|
error = self._html_search_regex(
|
|
r'(?s)<div[^>]+id=["\']errorMessages["\'][^>]*>(.+?)</div>',
|
|
login_page, 'error messages', default=None)
|
|
if error:
|
|
raise ExtractorError('Unable to login: %s' % error, expected=True)
|
|
raise ExtractorError('Unable to log in')
|
|
|
|
def _real_initialize(self):
|
|
self._login()
|
|
|
|
def _real_extract(self, url):
|
|
display_id = self._match_id(url)
|
|
|
|
errors = []
|
|
formats = []
|
|
|
|
ERRORS_MAP = {
|
|
'ERROR_MATURE_CONTENT_LOGGED_IN': 'matureContentLoggedIn',
|
|
'ERROR_MATURE_CONTENT_LOGGED_OUT': 'matureContentLoggedOut',
|
|
'ERROR_SUBSCRIPTION_LOGGED_OUT': 'subscriptionLoggedOut',
|
|
'ERROR_VIDEO_EXPIRED': 'videoExpired',
|
|
'ERROR_TERRITORY_UNAVAILABLE': 'territoryUnavailable',
|
|
'SVODBASIC_SUBSCRIPTION_IN_PLAYER': 'basicSubscription',
|
|
'SVODNON_SUBSCRIPTION_IN_PLAYER': 'nonSubscription',
|
|
'ERROR_PLAYER_NOT_RESPONDING': 'playerNotResponding',
|
|
'ERROR_UNABLE_TO_CONNECT_TO_CDN': 'unableToConnectToCDN',
|
|
'ERROR_STREAM_NOT_FOUND': 'streamNotFound',
|
|
}
|
|
|
|
USER_AGENTS = (
|
|
# PC UA is served with m3u8 that provides some bonus lower quality formats
|
|
('pc', 'Mozilla/5.0 (Windows NT 5.2; WOW64; rv:42.0) Gecko/20100101 Firefox/42.0'),
|
|
# Mobile UA allows to extract direct links and also does not fail when
|
|
# PC UA fails with hulu error (e.g.
|
|
# http://www.funimation.com/shows/hacksign/videos/official/role-play)
|
|
('mobile', 'Mozilla/5.0 (Linux; Android 4.4.2; Nexus 4 Build/KOT49H) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/34.0.1847.114 Mobile Safari/537.36'),
|
|
)
|
|
|
|
for kind, user_agent in USER_AGENTS:
|
|
request = sanitized_Request(url)
|
|
request.add_header('User-Agent', user_agent)
|
|
webpage = self._download_webpage(
|
|
request, display_id, 'Downloading %s webpage' % kind)
|
|
|
|
playlist = self._parse_json(
|
|
self._search_regex(
|
|
r'var\s+playersData\s*=\s*(\[.+?\]);\n',
|
|
webpage, 'players data'),
|
|
display_id)[0]['playlist']
|
|
|
|
items = next(item['items'] for item in playlist if item.get('items'))
|
|
item = next(item for item in items if item.get('itemAK') == display_id)
|
|
|
|
error_messages = {}
|
|
video_error_messages = self._search_regex(
|
|
r'var\s+videoErrorMessages\s*=\s*({.+?});\n',
|
|
webpage, 'error messages', default=None)
|
|
if video_error_messages:
|
|
error_messages_json = self._parse_json(video_error_messages, display_id, fatal=False)
|
|
if error_messages_json:
|
|
for _, error in error_messages_json.items():
|
|
type_ = error.get('type')
|
|
description = error.get('description')
|
|
content = error.get('content')
|
|
if type_ == 'text' and description and content:
|
|
error_message = ERRORS_MAP.get(description)
|
|
if error_message:
|
|
error_messages[error_message] = content
|
|
|
|
for video in item.get('videoSet', []):
|
|
auth_token = video.get('authToken')
|
|
if not auth_token:
|
|
continue
|
|
funimation_id = video.get('FUNImationID') or video.get('videoId')
|
|
preference = 1 if video.get('languageMode') == 'dub' else 0
|
|
if not auth_token.startswith('?'):
|
|
auth_token = '?%s' % auth_token
|
|
for quality, height in (('sd', 480), ('hd', 720), ('hd1080', 1080)):
|
|
format_url = video.get('%sUrl' % quality)
|
|
if not format_url:
|
|
continue
|
|
if not format_url.startswith(('http', '//')):
|
|
errors.append(format_url)
|
|
continue
|
|
if determine_ext(format_url) == 'm3u8':
|
|
formats.extend(self._extract_m3u8_formats(
|
|
format_url + auth_token, display_id, 'mp4', entry_protocol='m3u8_native',
|
|
preference=preference, m3u8_id='%s-hls' % funimation_id, fatal=False))
|
|
else:
|
|
tbr = int_or_none(self._search_regex(
|
|
r'-(\d+)[Kk]', format_url, 'tbr', default=None))
|
|
formats.append({
|
|
'url': format_url + auth_token,
|
|
'format_id': '%s-http-%dp' % (funimation_id, height),
|
|
'height': height,
|
|
'tbr': tbr,
|
|
'preference': preference,
|
|
})
|
|
|
|
if not formats and errors:
|
|
raise ExtractorError(
|
|
'%s returned error: %s'
|
|
% (self.IE_NAME, clean_html(error_messages.get(errors[0], errors[0]))),
|
|
expected=True)
|
|
|
|
self._sort_formats(formats)
|
|
|
|
title = item['title']
|
|
artist = item.get('artist')
|
|
if artist:
|
|
title = '%s - %s' % (artist, title)
|
|
description = self._og_search_description(webpage) or item.get('description')
|
|
thumbnail = self._og_search_thumbnail(webpage) or item.get('posterUrl')
|
|
video_id = item.get('itemId') or display_id
|
|
|
|
return {
|
|
'id': video_id,
|
|
'display_id': display_id,
|
|
'title': title,
|
|
'description': description,
|
|
'thumbnail': thumbnail,
|
|
'formats': formats,
|
|
}
|