mirror of
https://github.com/yt-dlp/yt-dlp.git
synced 2024-12-29 06:35:50 +01:00
Merge remote-tracking branch 'tobidope/gameone'
This commit is contained in:
commit
fa19dfccf9
@ -105,6 +105,7 @@
|
||||
from .freespeech import FreespeechIE
|
||||
from .funnyordie import FunnyOrDieIE
|
||||
from .gamekings import GamekingsIE
|
||||
from .gameone import GameOneIE
|
||||
from .gamespot import GameSpotIE
|
||||
from .gametrailers import GametrailersIE
|
||||
from .gdcvault import GDCVaultIE
|
||||
|
89
youtube_dl/extractor/gameone.py
Normal file
89
youtube_dl/extractor/gameone.py
Normal file
@ -0,0 +1,89 @@
|
||||
# coding: utf-8
|
||||
from __future__ import unicode_literals
|
||||
|
||||
import re
|
||||
|
||||
from .common import InfoExtractor
|
||||
from ..utils import (
|
||||
xpath_with_ns,
|
||||
parse_iso8601
|
||||
)
|
||||
|
||||
NAMESPACE_MAP = {
|
||||
'media': 'http://search.yahoo.com/mrss/',
|
||||
}
|
||||
|
||||
# URL prefix to download the mp4 files directly instead of streaming via rtmp
|
||||
# Credits go to XBox-Maniac
|
||||
# http://board.jdownloader.org/showpost.php?p=185835&postcount=31
|
||||
RAW_MP4_URL = 'http://cdn.riptide-mtvn.com/'
|
||||
|
||||
|
||||
class GameOneIE(InfoExtractor):
|
||||
_VALID_URL = r'https?://(?:www\.)?gameone\.de/tv/(?P<id>\d+)'
|
||||
_TEST = {
|
||||
'url': 'http://www.gameone.de/tv/288',
|
||||
'md5': '136656b7fb4c9cb4a8e2d500651c499b',
|
||||
'info_dict': {
|
||||
'id': '288',
|
||||
'ext': 'mp4',
|
||||
'title': 'Game One - Folge 288',
|
||||
'duration': 1238,
|
||||
'thumbnail': 'http://s3.gameone.de/gameone/assets/video_metas/teaser_images/000/643/636/big/640x360.jpg',
|
||||
'description': 'FIFA-Pressepokal 2014, Star Citizen, Kingdom Come: Deliverance, Project Cars, Schöner Trants Nerdquiz Folge 2 Runde 1',
|
||||
'age_limit': 16,
|
||||
'upload_date': '20140513',
|
||||
'timestamp': 1399980122,
|
||||
}
|
||||
}
|
||||
|
||||
def _real_extract(self, url):
|
||||
mobj = re.match(self._VALID_URL, url)
|
||||
video_id = mobj.group('id')
|
||||
|
||||
webpage = self._download_webpage(url, video_id)
|
||||
og_video = self._og_search_video_url(webpage, secure=False)
|
||||
description = self._html_search_meta('description', webpage)
|
||||
age_limit = int(
|
||||
self._search_regex(
|
||||
r'age=(\d+)',
|
||||
self._html_search_meta(
|
||||
'age-de-meta-label',
|
||||
webpage),
|
||||
'age_limit',
|
||||
'0'))
|
||||
mrss_url = self._search_regex(r'mrss=([^&]+)', og_video, 'mrss')
|
||||
|
||||
mrss = self._download_xml(mrss_url, video_id, 'Downloading mrss')
|
||||
title = mrss.find('.//item/title').text
|
||||
thumbnail = mrss.find('.//item/image').get('url')
|
||||
timestamp = parse_iso8601(mrss.find('.//pubDate').text, delimiter=' ')
|
||||
content = mrss.find(xpath_with_ns('.//media:content', NAMESPACE_MAP))
|
||||
content_url = content.get('url')
|
||||
|
||||
content = self._download_xml(
|
||||
content_url,
|
||||
video_id,
|
||||
'Downloading media:content')
|
||||
rendition_items = content.findall('.//rendition')
|
||||
duration = int(rendition_items[0].get('duration'))
|
||||
formats = [
|
||||
{
|
||||
'url': re.sub(r'.*/(r2)', RAW_MP4_URL + r'\1', r.find('./src').text),
|
||||
'width': int(r.get('width')),
|
||||
'height': int(r.get('height')),
|
||||
'tbr': int(r.get('bitrate')),
|
||||
}
|
||||
for r in rendition_items
|
||||
]
|
||||
|
||||
return {
|
||||
'id': video_id,
|
||||
'title': title,
|
||||
'thumbnail': thumbnail,
|
||||
'duration': duration,
|
||||
'formats': formats,
|
||||
'description': description,
|
||||
'age_limit': age_limit,
|
||||
'timestamp': timestamp,
|
||||
}
|
@ -775,7 +775,7 @@ def http_response(self, req, resp):
|
||||
https_response = http_response
|
||||
|
||||
|
||||
def parse_iso8601(date_str):
|
||||
def parse_iso8601(date_str, delimiter='T'):
|
||||
""" Return a UNIX timestamp from the given date """
|
||||
|
||||
if date_str is None:
|
||||
@ -795,8 +795,8 @@ def parse_iso8601(date_str):
|
||||
timezone = datetime.timedelta(
|
||||
hours=sign * int(m.group('hours')),
|
||||
minutes=sign * int(m.group('minutes')))
|
||||
|
||||
dt = datetime.datetime.strptime(date_str, '%Y-%m-%dT%H:%M:%S') - timezone
|
||||
date_format = '%Y-%m-%d{0}%H:%M:%S'.format(delimiter)
|
||||
dt = datetime.datetime.strptime(date_str, date_format) - timezone
|
||||
return calendar.timegm(dt.timetuple())
|
||||
|
||||
|
||||
|
Loading…
Reference in New Issue
Block a user