mirror of
https://github.com/ytdl-org/youtube-dl.git
synced 2024-11-22 02:01:50 +00:00
[NPORadio] Added extractor for live radio
This commit is contained in:
parent
08b38d5401
commit
7c24ce225d
2 changed files with 41 additions and 0 deletions
|
@ -320,6 +320,7 @@ from .npo import (
|
|||
NPOLiveIE,
|
||||
TegenlichtVproIE,
|
||||
)
|
||||
from .nporadio import NPORadioIE
|
||||
from .nrk import (
|
||||
NRKIE,
|
||||
NRKTVIE,
|
||||
|
|
40
youtube_dl/extractor/nporadio.py
Normal file
40
youtube_dl/extractor/nporadio.py
Normal file
|
@ -0,0 +1,40 @@
|
|||
# coding: utf-8
|
||||
from __future__ import unicode_literals
|
||||
|
||||
import json
|
||||
|
||||
from .common import InfoExtractor
|
||||
|
||||
|
||||
class NPORadioIE(InfoExtractor):
|
||||
_VALID_URL = r'https?://(?:www\.)?npo\.nl/radio/(?P<id>.*)'
|
||||
_TEST = {
|
||||
'url': 'http://www.npo.nl/radio/radio-1',
|
||||
'info_dict': {
|
||||
'id': 'radio-1',
|
||||
'ext': 'mp3',
|
||||
'title': 'NPO Radio 1',
|
||||
}
|
||||
}
|
||||
|
||||
def _real_extract(self, url):
|
||||
video_id = self._match_id(url)
|
||||
webpage = self._download_webpage(url, video_id)
|
||||
|
||||
title = self._html_search_regex(
|
||||
self._html_get_attribute_regex('data-channel'), webpage, 'title')
|
||||
|
||||
json_data = json.loads(
|
||||
self._html_search_regex(
|
||||
self._html_get_attribute_regex('data-streams'), webpage, 'data-streams'))
|
||||
|
||||
return {
|
||||
'id': video_id,
|
||||
'title': title,
|
||||
'ext': json_data['codec'],
|
||||
'url': json_data['url']
|
||||
}
|
||||
|
||||
def _html_get_attribute_regex(self, attribute):
|
||||
return r'{0}\s*=\s*\'([^\']+)\''.format(attribute)
|
||||
|
Loading…
Reference in a new issue