1
0
Fork 0
mirror of https://github.com/ytdl-org/youtube-dl.git synced 2024-11-14 21:35:39 +00:00
youtube-dl/youtube_dl/extractor/hrti.py

209 lines
7 KiB
Python
Raw Normal View History

# coding: utf-8
from __future__ import unicode_literals
import json
import re
from .common import InfoExtractor
from ..compat import compat_HTTPError
from ..utils import (
clean_html,
ExtractorError,
int_or_none,
parse_age_limit,
sanitized_Request,
try_get,
)
class HRTiBaseIE(InfoExtractor):
"""
Base Information Extractor for Croatian Radiotelevision
video on demand site https://hrti.hrt.hr
Reverse engineered from the JavaScript app in app.min.js
"""
_NETRC_MACHINE = 'hrti'
_APP_LANGUAGE = 'hr'
_APP_VERSION = '1.1'
_APP_PUBLICATION_ID = 'all_in_one'
_API_URL = 'http://clientapi.hrt.hr/client_api.php/config/identify/format/json'
def _initialize_api(self):
init_data = {
'application_publication_id': self._APP_PUBLICATION_ID
}
uuid = self._download_json(
self._API_URL, None, note='Downloading uuid',
errnote='Unable to download uuid',
data=json.dumps(init_data).encode('utf-8'))['uuid']
app_data = {
'uuid': uuid,
'application_publication_id': self._APP_PUBLICATION_ID,
'application_version': self._APP_VERSION
}
req = sanitized_Request(self._API_URL, data=json.dumps(app_data).encode('utf-8'))
req.get_method = lambda: 'PUT'
resources = self._download_json(
req, None, note='Downloading session information',
errnote='Unable to download session information')
self._session_id = resources['session_id']
modules = resources['modules']
self._search_url = modules['vod_catalog']['resources']['search']['uri'].format(
language=self._APP_LANGUAGE,
application_id=self._APP_PUBLICATION_ID)
self._login_url = (modules['user']['resources']['login']['uri']
+ '/format/json').format(session_id=self._session_id)
self._logout_url = modules['user']['resources']['logout']['uri']
def _login(self):
username, password = self._get_login_info()
# TODO: figure out authentication with cookies
if username is None or password is None:
self.raise_login_required()
auth_data = {
'username': username,
'password': password,
}
try:
auth_info = self._download_json(
self._login_url, None, note='Logging in', errnote='Unable to log in',
data=json.dumps(auth_data).encode('utf-8'))
except ExtractorError as e:
if isinstance(e.cause, compat_HTTPError) and e.cause.code == 406:
auth_info = self._parse_json(e.cause.read().encode('utf-8'), None)
else:
raise
error_message = auth_info.get('error', {}).get('message')
if error_message:
raise ExtractorError(
'%s said: %s' % (self.IE_NAME, error_message),
expected=True)
self._token = auth_info['secure_streaming_token']
def _real_initialize(self):
self._initialize_api()
self._login()
class HRTiIE(HRTiBaseIE):
_VALID_URL = r'''(?x)
(?:
hrti:(?P<short_id>[0-9]+)|
https?://
2017-10-07 22:40:08 +00:00
hrti\.hrt\.hr/(?:\#/)?video/show/(?P<id>[0-9]+)/(?P<display_id>[^/]+)?
)
'''
_TESTS = [{
'url': 'https://hrti.hrt.hr/#/video/show/2181385/republika-dokumentarna-serija-16-hd',
'info_dict': {
'id': '2181385',
'display_id': 'republika-dokumentarna-serija-16-hd',
'ext': 'mp4',
'title': 'REPUBLIKA, dokumentarna serija (1/6) (HD)',
'description': 'md5:48af85f620e8e0e1df4096270568544f',
'duration': 2922,
'view_count': int,
'average_rating': int,
'episode_number': int,
'season_number': int,
'age_limit': 12,
},
'skip': 'Requires account credentials',
}, {
'url': 'https://hrti.hrt.hr/#/video/show/2181385/',
'only_matching': True,
}, {
'url': 'hrti:2181385',
'only_matching': True,
2017-10-07 22:40:08 +00:00
}, {
'url': 'https://hrti.hrt.hr/video/show/3873068/cuvar-dvorca-dramska-serija-14',
'only_matching': True,
}]
def _real_extract(self, url):
mobj = re.match(self._VALID_URL, url)
video_id = mobj.group('short_id') or mobj.group('id')
display_id = mobj.group('display_id') or video_id
video = self._download_json(
'%s/video_id/%s/format/json' % (self._search_url, video_id),
display_id, 'Downloading video metadata JSON')['video'][0]
title_info = video['title']
title = title_info['title_long']
movie = video['video_assets']['movie'][0]
m3u8_url = movie['url'].format(TOKEN=self._token)
formats = self._extract_m3u8_formats(
m3u8_url, display_id, 'mp4', entry_protocol='m3u8_native',
m3u8_id='hls')
self._sort_formats(formats)
description = clean_html(title_info.get('summary_long'))
age_limit = parse_age_limit(video.get('parental_control', {}).get('rating'))
view_count = int_or_none(video.get('views'))
average_rating = int_or_none(video.get('user_rating'))
duration = int_or_none(movie.get('duration'))
return {
'id': video_id,
'display_id': display_id,
'title': title,
'description': description,
'duration': duration,
'view_count': view_count,
'average_rating': average_rating,
'age_limit': age_limit,
'formats': formats,
}
class HRTiPlaylistIE(HRTiBaseIE):
2017-10-09 16:52:39 +00:00
_VALID_URL = r'https?://hrti\.hrt\.hr/(?:#/)?video/list/category/(?P<id>[0-9]+)/(?P<display_id>[^/]+)?'
_TESTS = [{
'url': 'https://hrti.hrt.hr/#/video/list/category/212/ekumena',
'info_dict': {
'id': '212',
'title': 'ekumena',
},
'playlist_mincount': 8,
'skip': 'Requires account credentials',
}, {
'url': 'https://hrti.hrt.hr/#/video/list/category/212/',
'only_matching': True,
2017-10-09 16:52:39 +00:00
}, {
'url': 'https://hrti.hrt.hr/video/list/category/212/ekumena',
'only_matching': True,
}]
def _real_extract(self, url):
mobj = re.match(self._VALID_URL, url)
category_id = mobj.group('id')
display_id = mobj.group('display_id') or category_id
response = self._download_json(
'%s/category_id/%s/format/json' % (self._search_url, category_id),
display_id, 'Downloading video metadata JSON')
video_ids = try_get(
response, lambda x: x['video_listings'][0]['alternatives'][0]['list'],
list) or [video['id'] for video in response.get('videos', []) if video.get('id')]
entries = [self.url_result('hrti:%s' % video_id) for video_id in video_ids]
return self.playlist_result(entries, category_id, display_id)