1
0
mirror of https://codeberg.org/polarisfm/youtube-dl synced 2024-11-14 05:34:32 +01:00
youtube-dl/youtube_dl/extractor/philharmoniedeparis.py

107 lines
3.7 KiB
Python

# coding: utf-8
from __future__ import unicode_literals
from .common import InfoExtractor
from ..compat import compat_str
from ..utils import (
try_get,
urljoin,
)
class PhilharmonieDeParisIE(InfoExtractor):
IE_DESC = 'Philharmonie de Paris'
_VALID_URL = r'''(?x)
https?://
(?:
live\.philharmoniedeparis\.fr/(?:[Cc]oncert/|embed(?:app)?/|misc/Playlist\.ashx\?id=)|
pad\.philharmoniedeparis\.fr/doc/CIMU/
)
(?P<id>\d+)
'''
_TESTS = [{
'url': 'http://pad.philharmoniedeparis.fr/doc/CIMU/1086697/jazz-a-la-villette-knower',
'md5': 'a0a4b195f544645073631cbec166a2c2',
'info_dict': {
'id': '1086697',
'ext': 'mp4',
'title': 'Jazz à la Villette : Knower',
},
}, {
'url': 'http://live.philharmoniedeparis.fr/concert/1032066.html',
'info_dict': {
'id': '1032066',
'title': 'md5:0a031b81807b3593cffa3c9a87a167a0',
},
'playlist_mincount': 2,
}, {
'url': 'http://live.philharmoniedeparis.fr/Concert/1030324.html',
'only_matching': True,
}, {
'url': 'http://live.philharmoniedeparis.fr/misc/Playlist.ashx?id=1030324&track=&lang=fr',
'only_matching': True,
}, {
'url': 'https://live.philharmoniedeparis.fr/embedapp/1098406/berlioz-fantastique-lelio-les-siecles-national-youth-choir-of.html?lang=fr-FR',
'only_matching': True,
}, {
'url': 'https://live.philharmoniedeparis.fr/embed/1098406/berlioz-fantastique-lelio-les-siecles-national-youth-choir-of.html?lang=fr-FR',
'only_matching': True,
}]
_LIVE_URL = 'https://live.philharmoniedeparis.fr'
def _real_extract(self, url):
video_id = self._match_id(url)
config = self._download_json(
'%s/otoPlayer/config.ashx' % self._LIVE_URL, video_id, query={
'id': video_id,
'lang': 'fr-FR',
})
def extract_entry(source):
if not isinstance(source, dict):
return
title = source.get('title')
if not title:
return
files = source.get('files')
if not isinstance(files, dict):
return
format_urls = set()
formats = []
for format_id in ('mobile', 'desktop'):
format_url = try_get(
files, lambda x: x[format_id]['file'], compat_str)
if not format_url or format_url in format_urls:
continue
format_urls.add(format_url)
m3u8_url = urljoin(self._LIVE_URL, format_url)
formats.extend(self._extract_m3u8_formats(
m3u8_url, video_id, 'mp4', entry_protocol='m3u8_native',
m3u8_id='hls', fatal=False))
if not formats:
return
self._sort_formats(formats)
return {
'title': title,
'formats': formats,
}
thumbnail = urljoin(self._LIVE_URL, config.get('image'))
info = extract_entry(config)
if info:
info.update({
'id': video_id,
'thumbnail': thumbnail,
})
return info
entries = []
for num, chapter in enumerate(config['chapters'], start=1):
entry = extract_entry(chapter)
entry['id'] = '%s-%d' % (video_id, num)
entries.append(entry)
return self.playlist_result(entries, video_id, config.get('title'))