mirror of
https://github.com/ytdl-org/youtube-dl.git
synced 2024-12-05 01:48:48 +01:00
142 lines
5.1 KiB
Python
142 lines
5.1 KiB
Python
# coding: utf-8
|
|
from __future__ import unicode_literals
|
|
|
|
import re
|
|
import itertools
|
|
|
|
from .common import InfoExtractor
|
|
|
|
|
|
class VierIE(InfoExtractor):
|
|
IE_NAME = 'vier'
|
|
IE_DESC = 'vier.be and vijf.be'
|
|
_VALID_URL = r'https?://(?:www\.)?(?P<site>vier|vijf)\.be/(?:[^/]+/videos/(?P<display_id>[^/]+)(?:/(?P<id>\d+))?|video/v3/embed/(?P<embed_id>\d+))'
|
|
_TESTS = [{
|
|
'url': 'http://www.vier.be/planb/videos/het-wordt-warm-de-moestuin/16129',
|
|
'info_dict': {
|
|
'id': '16129',
|
|
'display_id': 'het-wordt-warm-de-moestuin',
|
|
'ext': 'mp4',
|
|
'title': 'Het wordt warm in De Moestuin',
|
|
'description': 'De vele uren werk eisen hun tol. Wim droomt van assistentie...',
|
|
},
|
|
'params': {
|
|
# m3u8 download
|
|
'skip_download': True,
|
|
},
|
|
}, {
|
|
'url': 'http://www.vijf.be/temptationisland/videos/zo-grappig-temptation-island-hosts-moeten-kiezen-tussen-onmogelijke-dilemmas/2561614',
|
|
'info_dict': {
|
|
'id': '2561614',
|
|
'display_id': 'zo-grappig-temptation-island-hosts-moeten-kiezen-tussen-onmogelijke-dilemmas',
|
|
'ext': 'mp4',
|
|
'title': 'ZO grappig: Temptation Island hosts moeten kiezen tussen onmogelijke dilemma\'s',
|
|
'description': 'Het spel is simpel: Annelien Coorevits en Rick Brandsteder krijgen telkens 2 dilemma\'s voorgeschoteld en ze MOETEN een keuze maken.',
|
|
},
|
|
'params': {
|
|
# m3u8 download
|
|
'skip_download': True,
|
|
},
|
|
}, {
|
|
'url': 'http://www.vier.be/planb/videos/mieren-herders-van-de-bladluizen',
|
|
'only_matching': True,
|
|
}, {
|
|
'url': 'http://www.vier.be/video/v3/embed/16129',
|
|
'only_matching': True,
|
|
}]
|
|
|
|
def _real_extract(self, url):
|
|
mobj = re.match(self._VALID_URL, url)
|
|
embed_id = mobj.group('embed_id')
|
|
display_id = mobj.group('display_id') or embed_id
|
|
site = mobj.group('site')
|
|
|
|
webpage = self._download_webpage(url, display_id)
|
|
|
|
video_id = self._search_regex(
|
|
[r'data-nid="(\d+)"', r'"nid"\s*:\s*"(\d+)"'],
|
|
webpage, 'video id')
|
|
application = self._search_regex(
|
|
[r'data-application="([^"]+)"', r'"application"\s*:\s*"([^"]+)"'],
|
|
webpage, 'application', default=site + '_vod')
|
|
filename = self._search_regex(
|
|
[r'data-filename="([^"]+)"', r'"filename"\s*:\s*"([^"]+)"'],
|
|
webpage, 'filename')
|
|
|
|
playlist_url = 'http://vod.streamcloud.be/%s/_definst_/mp4:%s.mp4/playlist.m3u8' % (application, filename)
|
|
formats = self._extract_wowza_formats(playlist_url, display_id, skip_protocols=['dash'])
|
|
self._sort_formats(formats)
|
|
|
|
title = self._og_search_title(webpage, default=display_id)
|
|
description = self._og_search_description(webpage, default=None)
|
|
thumbnail = self._og_search_thumbnail(webpage, default=None)
|
|
|
|
return {
|
|
'id': video_id,
|
|
'display_id': display_id,
|
|
'title': title,
|
|
'description': description,
|
|
'thumbnail': thumbnail,
|
|
'formats': formats,
|
|
}
|
|
|
|
|
|
class VierVideosIE(InfoExtractor):
|
|
IE_NAME = 'vier:videos'
|
|
_VALID_URL = r'https?://(?:www\.)?(?P<site>vier|vijf)\.be/(?P<program>[^/]+)/videos(?:\?.*\bpage=(?P<page>\d+)|$)'
|
|
_TESTS = [{
|
|
'url': 'http://www.vier.be/demoestuin/videos',
|
|
'info_dict': {
|
|
'id': 'demoestuin',
|
|
},
|
|
'playlist_mincount': 153,
|
|
}, {
|
|
'url': 'http://www.vijf.be/temptationisland/videos',
|
|
'info_dict': {
|
|
'id': 'temptationisland',
|
|
},
|
|
'playlist_mincount': 159,
|
|
}, {
|
|
'url': 'http://www.vier.be/demoestuin/videos?page=6',
|
|
'info_dict': {
|
|
'id': 'demoestuin-page6',
|
|
},
|
|
'playlist_mincount': 20,
|
|
}, {
|
|
'url': 'http://www.vier.be/demoestuin/videos?page=7',
|
|
'info_dict': {
|
|
'id': 'demoestuin-page7',
|
|
},
|
|
'playlist_mincount': 13,
|
|
}]
|
|
|
|
def _real_extract(self, url):
|
|
mobj = re.match(self._VALID_URL, url)
|
|
program = mobj.group('program')
|
|
site = mobj.group('site')
|
|
|
|
page_id = mobj.group('page')
|
|
if page_id:
|
|
page_id = int(page_id)
|
|
start_page = page_id
|
|
playlist_id = '%s-page%d' % (program, page_id)
|
|
else:
|
|
start_page = 0
|
|
playlist_id = program
|
|
|
|
entries = []
|
|
for current_page_id in itertools.count(start_page):
|
|
current_page = self._download_webpage(
|
|
'http://www.%s.be/%s/videos?page=%d' % (site, program, current_page_id),
|
|
program,
|
|
'Downloading page %d' % (current_page_id + 1))
|
|
page_entries = [
|
|
self.url_result('http://www.' + site + '.be' + video_url, 'Vier')
|
|
for video_url in re.findall(
|
|
r'<h[23]><a href="(/[^/]+/videos/[^/]+(?:/\d+)?)">', current_page)]
|
|
entries.extend(page_entries)
|
|
if page_id or '>Meer<' not in current_page:
|
|
break
|
|
|
|
return self.playlist_result(entries, playlist_id)
|