|
|
@ -1,149 +0,0 @@ |
|
|
|
# coding: utf-8 |
|
|
|
from __future__ import unicode_literals |
|
|
|
|
|
|
|
import re |
|
|
|
|
|
|
|
from .common import InfoExtractor |
|
|
|
from ..compat import compat_HTTPError |
|
|
|
from ..utils import ( |
|
|
|
determine_ext, |
|
|
|
ExtractorError, |
|
|
|
int_or_none, |
|
|
|
parse_age_limit, |
|
|
|
parse_iso8601, |
|
|
|
) |
|
|
|
|
|
|
|
|
|
|
|
class Go90IE(InfoExtractor): |
|
|
|
_VALID_URL = r'https?://(?:www\.)?go90\.com/(?:videos|embed)/(?P<id>[0-9a-zA-Z]+)' |
|
|
|
_TESTS = [{ |
|
|
|
'url': 'https://www.go90.com/videos/84BUqjLpf9D', |
|
|
|
'md5': 'efa7670dbbbf21a7b07b360652b24a32', |
|
|
|
'info_dict': { |
|
|
|
'id': '84BUqjLpf9D', |
|
|
|
'ext': 'mp4', |
|
|
|
'title': 'Daily VICE - Inside The Utah Coalition Against Pornography Convention', |
|
|
|
'description': 'VICE\'s Karley Sciortino meets with activists who discuss the state\'s strong anti-porn stance. Then, VICE Sports explains NFL contracts.', |
|
|
|
'timestamp': 1491868800, |
|
|
|
'upload_date': '20170411', |
|
|
|
'age_limit': 14, |
|
|
|
} |
|
|
|
}, { |
|
|
|
'url': 'https://www.go90.com/embed/261MflWkD3N', |
|
|
|
'only_matching': True, |
|
|
|
}] |
|
|
|
_GEO_BYPASS = False |
|
|
|
|
|
|
|
def _real_extract(self, url): |
|
|
|
video_id = self._match_id(url) |
|
|
|
|
|
|
|
try: |
|
|
|
headers = self.geo_verification_headers() |
|
|
|
headers.update({ |
|
|
|
'Content-Type': 'application/json; charset=utf-8', |
|
|
|
}) |
|
|
|
video_data = self._download_json( |
|
|
|
'https://www.go90.com/api/view/items/' + video_id, video_id, |
|
|
|
headers=headers, data=b'{"client":"web","device_type":"pc"}') |
|
|
|
except ExtractorError as e: |
|
|
|
if isinstance(e.cause, compat_HTTPError) and e.cause.code == 400: |
|
|
|
message = self._parse_json(e.cause.read().decode(), None)['error']['message'] |
|
|
|
if 'region unavailable' in message: |
|
|
|
self.raise_geo_restricted(countries=['US']) |
|
|
|
raise ExtractorError(message, expected=True) |
|
|
|
raise |
|
|
|
|
|
|
|
if video_data.get('requires_drm'): |
|
|
|
raise ExtractorError('This video is DRM protected.', expected=True) |
|
|
|
main_video_asset = video_data['main_video_asset'] |
|
|
|
|
|
|
|
episode_number = int_or_none(video_data.get('episode_number')) |
|
|
|
series = None |
|
|
|
season = None |
|
|
|
season_id = None |
|
|
|
season_number = None |
|
|
|
for metadata in video_data.get('__children', {}).get('Item', {}).values(): |
|
|
|
if metadata.get('type') == 'show': |
|
|
|
series = metadata.get('title') |
|
|
|
elif metadata.get('type') == 'season': |
|
|
|
season = metadata.get('title') |
|
|
|
season_id = metadata.get('id') |
|
|
|
season_number = int_or_none(metadata.get('season_number')) |
|
|
|
|
|
|
|
title = episode = video_data.get('title') or series |
|
|
|
if series and series != title: |
|
|
|
title = '%s - %s' % (series, title) |
|
|
|
|
|
|
|
thumbnails = [] |
|
|
|
formats = [] |
|
|
|
subtitles = {} |
|
|
|
for asset in video_data.get('assets'): |
|
|
|
if asset.get('id') == main_video_asset: |
|
|
|
for source in asset.get('sources', []): |
|
|
|
source_location = source.get('location') |
|
|
|
if not source_location: |
|
|
|
continue |
|
|
|
source_type = source.get('type') |
|
|
|
if source_type == 'hls': |
|
|
|
m3u8_formats = self._extract_m3u8_formats( |
|
|
|
source_location, video_id, 'mp4', |
|
|
|
'm3u8_native', m3u8_id='hls', fatal=False) |
|
|
|
for f in m3u8_formats: |
|
|
|
mobj = re.search(r'/hls-(\d+)-(\d+)K', f['url']) |
|
|
|
if mobj: |
|
|
|
height, tbr = mobj.groups() |
|
|
|
height = int_or_none(height) |
|
|
|
f.update({ |
|
|
|
'height': f.get('height') or height, |
|
|
|
'width': f.get('width') or int_or_none(height / 9.0 * 16.0 if height else None), |
|
|
|
'tbr': f.get('tbr') or int_or_none(tbr), |
|
|
|
}) |
|
|
|
formats.extend(m3u8_formats) |
|
|
|
elif source_type == 'dash': |
|
|
|
formats.extend(self._extract_mpd_formats( |
|
|
|
source_location, video_id, mpd_id='dash', fatal=False)) |
|
|
|
else: |
|
|
|
formats.append({ |
|
|
|
'format_id': source.get('name'), |
|
|
|
'url': source_location, |
|
|
|
'width': int_or_none(source.get('width')), |
|
|
|
'height': int_or_none(source.get('height')), |
|
|
|
'tbr': int_or_none(source.get('bitrate')), |
|
|
|
}) |
|
|
|
|
|
|
|
for caption in asset.get('caption_metadata', []): |
|
|
|
caption_url = caption.get('source_url') |
|
|
|
if not caption_url: |
|
|
|
continue |
|
|
|
subtitles.setdefault(caption.get('language', 'en'), []).append({ |
|
|
|
'url': caption_url, |
|
|
|
'ext': determine_ext(caption_url, 'vtt'), |
|
|
|
}) |
|
|
|
elif asset.get('type') == 'image': |
|
|
|
asset_location = asset.get('location') |
|
|
|
if not asset_location: |
|
|
|
continue |
|
|
|
thumbnails.append({ |
|
|
|
'url': asset_location, |
|
|
|
'width': int_or_none(asset.get('width')), |
|
|
|
'height': int_or_none(asset.get('height')), |
|
|
|
}) |
|
|
|
self._sort_formats(formats) |
|
|
|
|
|
|
|
return { |
|
|
|
'id': video_id, |
|
|
|
'title': title, |
|
|
|
'formats': formats, |
|
|
|
'thumbnails': thumbnails, |
|
|
|
'description': video_data.get('short_description'), |
|
|
|
'like_count': int_or_none(video_data.get('like_count')), |
|
|
|
'timestamp': parse_iso8601(video_data.get('released_at')), |
|
|
|
'series': series, |
|
|
|
'episode': episode, |
|
|
|
'season': season, |
|
|
|
'season_id': season_id, |
|
|
|
'season_number': season_number, |
|
|
|
'episode_number': episode_number, |
|
|
|
'subtitles': subtitles, |
|
|
|
'age_limit': parse_age_limit(video_data.get('rating')), |
|
|
|
} |