forked from minhngoc25a/yt-dlc
[revision3] Add new extractor(closes #6388)
- revision3.com - testtube.com - animalist.com
This commit is contained in:
parent
ca227c8698
commit
054479754c
|
@ -557,6 +557,7 @@ from .redtube import RedTubeIE
|
|||
from .regiotv import RegioTVIE
|
||||
from .restudy import RestudyIE
|
||||
from .reverbnation import ReverbNationIE
|
||||
from .revision3 import Revision3IE
|
||||
from .ringtv import RingTVIE
|
||||
from .ro220 import Ro220IE
|
||||
from .rottentomatoes import RottenTomatoesIE
|
||||
|
|
|
@ -0,0 +1,103 @@
|
|||
# coding: utf-8
|
||||
from __future__ import unicode_literals
|
||||
|
||||
import re
|
||||
|
||||
from .common import InfoExtractor
|
||||
from ..compat import compat_str
|
||||
from ..utils import (
|
||||
int_or_none,
|
||||
parse_iso8601,
|
||||
unescapeHTML,
|
||||
)
|
||||
|
||||
|
||||
class Revision3IE(InfoExtractor):
|
||||
_VALID_URL = r'https?://(?:www\.)?(?P<domain>(?:revision3|testtube|animalist)\.com)/(?P<id>[^/]+(?:/[^/?#]+)?)'
|
||||
_TESTS = [{
|
||||
'url': 'http://www.revision3.com/technobuffalo/5-google-predictions-for-2016',
|
||||
'md5': 'd94a72d85d0a829766de4deb8daaf7df',
|
||||
'info_dict': {
|
||||
'id': '73034',
|
||||
'ext': 'webm',
|
||||
'title': '5 Google Predictions for 2016',
|
||||
'description': 'Google had a great 2015, but it\'s already time to look ahead. Here are our five predictions for 2016.',
|
||||
'upload_date': '20151228',
|
||||
'timestamp': 1451325600,
|
||||
'duration': 187,
|
||||
}
|
||||
}, {
|
||||
'url': 'http://testtube.com/brainstuff',
|
||||
'info_dict': {
|
||||
'id': '251',
|
||||
'title': 'BrainStuff',
|
||||
'description': 'Whether the topic is popcorn or particle physics, you can count on the HowStuffWorks team to explore-and explain-the everyday science in the world around us on BrainStuff.',
|
||||
},
|
||||
'playlist_mincount': 93,
|
||||
}]
|
||||
_PAGE_DATA_TEMPLATE = 'http://www.%s/apiProxy/ddn/%s?domain=%s'
|
||||
_API_KEY = 'ba9c741bce1b9d8e3defcc22193f3651b8867e62'
|
||||
|
||||
def _real_extract(self, url):
|
||||
domain, display_id = re.match(self._VALID_URL, url).groups()
|
||||
page_info = self._download_json(
|
||||
self._PAGE_DATA_TEMPLATE % (domain, display_id, domain), display_id)
|
||||
|
||||
if page_info['data']['type'] == 'episode':
|
||||
episode_data = page_info['data']
|
||||
video_id = compat_str(episode_data['video']['data']['id'])
|
||||
video_data = self._download_json(
|
||||
'http://revision3.com/api/getPlaylist.json?api_key=%s&codecs=h264,vp8,theora&video_id=%s' % (self._API_KEY, video_id),
|
||||
video_id)['items'][0]
|
||||
|
||||
formats = []
|
||||
for media_type, media in video_data['media'].items():
|
||||
for quality_id, quality in media.items():
|
||||
if quality_id == 'hls':
|
||||
formats.extend(self._extract_m3u8_formats(
|
||||
quality['url'], video_id, 'mp4',
|
||||
'm3u8_native', m3u8_id='hls', fatal=False))
|
||||
else:
|
||||
formats.append({
|
||||
'url': quality['url'],
|
||||
'format_id': '%s-%s' % (media_type, quality_id),
|
||||
'tbr': int_or_none(quality.get('bitrate')),
|
||||
})
|
||||
self._sort_formats(formats)
|
||||
|
||||
thumbnails = [{
|
||||
'url': image_url,
|
||||
'id': image_id,
|
||||
} for image_id, image_url in video_data.get('images', {}).items()]
|
||||
|
||||
return {
|
||||
'id': video_id,
|
||||
'title': unescapeHTML(video_data['title']),
|
||||
'description': unescapeHTML(video_data.get('summary')),
|
||||
'timestamp': parse_iso8601(episode_data.get('publishTime'), ' '),
|
||||
'author': episode_data.get('author'),
|
||||
'duration': int_or_none(video_data.get('duration')),
|
||||
'thumbnails': thumbnails,
|
||||
'formats': formats,
|
||||
}
|
||||
else:
|
||||
show_data = page_info['show']['data']
|
||||
episodes_data = page_info['episodes']['data']
|
||||
num_episodes = page_info['meta']['totalEpisodes']
|
||||
processed_episodes = 0
|
||||
entries = []
|
||||
page_num = 1
|
||||
while True:
|
||||
entries.extend([self.url_result(
|
||||
url + '/%s' % episode['slug']) for episode in episodes_data])
|
||||
processed_episodes += len(episodes_data)
|
||||
if processed_episodes == num_episodes:
|
||||
break
|
||||
page_num += 1
|
||||
episodes_data = self._download_json(self._PAGE_DATA_TEMPLATE % (
|
||||
domain, display_id + '/' + compat_str(page_num), domain),
|
||||
display_id)['episodes']['data']
|
||||
|
||||
return self.playlist_result(
|
||||
entries, compat_str(show_data['id']),
|
||||
show_data.get('name'), show_data.get('summary'))
|
Loading…
Reference in New Issue