mirror of
https://github.com/ytdl-org/youtube-dl
synced 2024-12-19 07:57:18 +00:00
339b1944e7
Closes #7690.
49 lines
1.5 KiB
Python
49 lines
1.5 KiB
Python
# coding: utf-8
|
|
from __future__ import unicode_literals
|
|
|
|
from .common import InfoExtractor
|
|
from ..utils import (
|
|
parse_duration,
|
|
remove_end,
|
|
)
|
|
|
|
|
|
class LRTIE(InfoExtractor):
|
|
IE_NAME = 'lrt.lt'
|
|
_VALID_URL = r'https?://(?:www\.)?lrt\.lt/mediateka/irasas/(?P<id>[0-9]+)'
|
|
_TEST = {
|
|
'url': 'http://www.lrt.lt/mediateka/irasas/54391/',
|
|
'info_dict': {
|
|
'id': '54391',
|
|
'ext': 'mp4',
|
|
'title': 'Septynios Kauno dienos',
|
|
'description': 'md5:24d84534c7dc76581e59f5689462411a',
|
|
'duration': 1783,
|
|
},
|
|
'params': {
|
|
'skip_download': True, # m3u8 download
|
|
},
|
|
}
|
|
|
|
def _real_extract(self, url):
|
|
video_id = self._match_id(url)
|
|
webpage = self._download_webpage(url, video_id)
|
|
|
|
title = remove_end(self._og_search_title(webpage), ' - LRT')
|
|
thumbnail = self._og_search_thumbnail(webpage)
|
|
description = self._og_search_description(webpage)
|
|
duration = parse_duration(self._search_regex(
|
|
r"var record_len = '([0-9]+:[0-9]+:[0-9]+)';", webpage, 'record_len', fatal=False, default=None))
|
|
|
|
link = self._search_regex(r'file: "(.*)" \+ location\.hash\.substring\(1\)', webpage, 'link to m3u8')
|
|
formats = self._extract_m3u8_formats(link, video_id, "mp4")
|
|
|
|
return {
|
|
'id': video_id,
|
|
'title': title,
|
|
'formats': formats,
|
|
'thumbnail': thumbnail,
|
|
'description': description,
|
|
'duration': duration,
|
|
}
|