mirror of
https://github.com/yt-dlp/yt-dlp
synced 2024-12-29 22:24:25 +01:00
7a5c1cfe93
* All modules and binary names are changed * All documentation references changed * yt-dlp no longer loads youtube-dlc config files * All URLs changed to point to organization account Co-authored-by: Pccode66 Co-authored-by: pukkandan
171 lines
6.2 KiB
Python
171 lines
6.2 KiB
Python
# coding: utf-8
|
|
from __future__ import unicode_literals
|
|
|
|
import re
|
|
|
|
from .common import InfoExtractor
|
|
from ..compat import compat_HTTPError
|
|
from ..utils import (
|
|
determine_ext,
|
|
ExtractorError,
|
|
int_or_none,
|
|
unified_strdate,
|
|
)
|
|
|
|
|
|
class RadioCanadaIE(InfoExtractor):
|
|
IE_NAME = 'radiocanada'
|
|
_VALID_URL = r'(?:radiocanada:|https?://ici\.radio-canada\.ca/widgets/mediaconsole/)(?P<app_code>[^:/]+)[:/](?P<id>[0-9]+)'
|
|
_TESTS = [
|
|
{
|
|
'url': 'http://ici.radio-canada.ca/widgets/mediaconsole/medianet/7184272',
|
|
'info_dict': {
|
|
'id': '7184272',
|
|
'ext': 'mp4',
|
|
'title': 'Le parcours du tireur capté sur vidéo',
|
|
'description': 'Images des caméras de surveillance fournies par la GRC montrant le parcours du tireur d\'Ottawa',
|
|
'upload_date': '20141023',
|
|
},
|
|
'params': {
|
|
# m3u8 download
|
|
'skip_download': True,
|
|
}
|
|
},
|
|
{
|
|
# empty Title
|
|
'url': 'http://ici.radio-canada.ca/widgets/mediaconsole/medianet/7754998/',
|
|
'info_dict': {
|
|
'id': '7754998',
|
|
'ext': 'mp4',
|
|
'title': 'letelejournal22h',
|
|
'description': 'INTEGRALE WEB 22H-TJ',
|
|
'upload_date': '20170720',
|
|
},
|
|
'params': {
|
|
# m3u8 download
|
|
'skip_download': True,
|
|
},
|
|
},
|
|
{
|
|
# with protectionType but not actually DRM protected
|
|
'url': 'radiocanada:toutv:140872',
|
|
'info_dict': {
|
|
'id': '140872',
|
|
'title': 'Épisode 1',
|
|
'series': 'District 31',
|
|
},
|
|
'only_matching': True,
|
|
}
|
|
]
|
|
_GEO_COUNTRIES = ['CA']
|
|
_access_token = None
|
|
_claims = None
|
|
|
|
def _call_api(self, path, video_id=None, app_code=None, query=None):
|
|
if not query:
|
|
query = {}
|
|
query.update({
|
|
'client_key': '773aea60-0e80-41bb-9c7f-e6d7c3ad17fb',
|
|
'output': 'json',
|
|
})
|
|
if video_id:
|
|
query.update({
|
|
'appCode': app_code,
|
|
'idMedia': video_id,
|
|
})
|
|
if self._access_token:
|
|
query['access_token'] = self._access_token
|
|
try:
|
|
return self._download_json(
|
|
'https://services.radio-canada.ca/media/' + path, video_id, query=query)
|
|
except ExtractorError as e:
|
|
if isinstance(e.cause, compat_HTTPError) and e.cause.code in (401, 422):
|
|
data = self._parse_json(e.cause.read().decode(), None)
|
|
error = data.get('error_description') or data['errorMessage']['text']
|
|
raise ExtractorError(error, expected=True)
|
|
raise
|
|
|
|
def _extract_info(self, app_code, video_id):
|
|
metas = self._call_api('meta/v1/index.ashx', video_id, app_code)['Metas']
|
|
|
|
def get_meta(name):
|
|
for meta in metas:
|
|
if meta.get('name') == name:
|
|
text = meta.get('text')
|
|
if text:
|
|
return text
|
|
|
|
# protectionType does not necessarily mean the video is DRM protected (see
|
|
# https://github.com/ytdl-org/youtube-dl/pull/18609).
|
|
if get_meta('protectionType'):
|
|
self.report_warning('This video is probably DRM protected.')
|
|
|
|
query = {
|
|
'connectionType': 'hd',
|
|
'deviceType': 'ipad',
|
|
'multibitrate': 'true',
|
|
}
|
|
if self._claims:
|
|
query['claims'] = self._claims
|
|
v_data = self._call_api('validation/v2/', video_id, app_code, query)
|
|
v_url = v_data.get('url')
|
|
if not v_url:
|
|
error = v_data['message']
|
|
if error == "Le contenu sélectionné n'est pas disponible dans votre pays":
|
|
raise self.raise_geo_restricted(error, self._GEO_COUNTRIES)
|
|
if error == 'Le contenu sélectionné est disponible seulement en premium':
|
|
self.raise_login_required(error)
|
|
raise ExtractorError(
|
|
'%s said: %s' % (self.IE_NAME, error), expected=True)
|
|
formats = self._extract_m3u8_formats(v_url, video_id, 'mp4')
|
|
self._sort_formats(formats)
|
|
|
|
subtitles = {}
|
|
closed_caption_url = get_meta('closedCaption') or get_meta('closedCaptionHTML5')
|
|
if closed_caption_url:
|
|
subtitles['fr'] = [{
|
|
'url': closed_caption_url,
|
|
'ext': determine_ext(closed_caption_url, 'vtt'),
|
|
}]
|
|
|
|
return {
|
|
'id': video_id,
|
|
'title': get_meta('Title') or get_meta('AV-nomEmission'),
|
|
'description': get_meta('Description') or get_meta('ShortDescription'),
|
|
'thumbnail': get_meta('imageHR') or get_meta('imageMR') or get_meta('imageBR'),
|
|
'duration': int_or_none(get_meta('length')),
|
|
'series': get_meta('Emission'),
|
|
'season_number': int_or_none('SrcSaison'),
|
|
'episode_number': int_or_none('SrcEpisode'),
|
|
'upload_date': unified_strdate(get_meta('Date')),
|
|
'subtitles': subtitles,
|
|
'formats': formats,
|
|
}
|
|
|
|
def _real_extract(self, url):
|
|
return self._extract_info(*re.match(self._VALID_URL, url).groups())
|
|
|
|
|
|
class RadioCanadaAudioVideoIE(InfoExtractor):
|
|
IE_NAME = 'radiocanada:audiovideo'
|
|
_VALID_URL = r'https?://ici\.radio-canada\.ca/([^/]+/)*media-(?P<id>[0-9]+)'
|
|
_TESTS = [{
|
|
'url': 'http://ici.radio-canada.ca/audio-video/media-7527184/barack-obama-au-vietnam',
|
|
'info_dict': {
|
|
'id': '7527184',
|
|
'ext': 'mp4',
|
|
'title': 'Barack Obama au Vietnam',
|
|
'description': 'Les États-Unis lèvent l\'embargo sur la vente d\'armes qui datait de la guerre du Vietnam',
|
|
'upload_date': '20160523',
|
|
},
|
|
'params': {
|
|
# m3u8 download
|
|
'skip_download': True,
|
|
},
|
|
}, {
|
|
'url': 'https://ici.radio-canada.ca/info/videos/media-7527184/barack-obama-au-vietnam',
|
|
'only_matching': True,
|
|
}]
|
|
|
|
def _real_extract(self, url):
|
|
return self.url_result('radiocanada:medianet:%s' % self._match_id(url))
|