mirror of
https://github.com/yt-dlp/yt-dlp.git
synced 2024-11-24 00:01:54 +01:00
Compare commits
2 Commits
4cee264e7c
...
77839103ec
Author | SHA1 | Date | |
---|---|---|---|
|
77839103ec | ||
|
04417e89de |
|
@ -1718,6 +1718,7 @@ from .rinsefm import (
|
||||||
RinseFMIE,
|
RinseFMIE,
|
||||||
)
|
)
|
||||||
from .rmcdecouverte import RMCDecouverteIE
|
from .rmcdecouverte import RMCDecouverteIE
|
||||||
|
from .roblox import RobloxIE
|
||||||
from .rockstargames import RockstarGamesIE
|
from .rockstargames import RockstarGamesIE
|
||||||
from .rokfin import (
|
from .rokfin import (
|
||||||
RokfinChannelIE,
|
RokfinChannelIE,
|
||||||
|
|
129
yt_dlp/extractor/roblox.py
Normal file
129
yt_dlp/extractor/roblox.py
Normal file
|
@ -0,0 +1,129 @@
|
||||||
|
from .common import InfoExtractor
|
||||||
|
from ..postprocessor.ffmpeg import FFmpegPostProcessor
|
||||||
|
from ..utils import (
|
||||||
|
ExtractorError,
|
||||||
|
extract_attributes,
|
||||||
|
float_or_none,
|
||||||
|
get_element_by_id,
|
||||||
|
get_element_by_class,
|
||||||
|
get_element_html_by_class,
|
||||||
|
get_element_html_by_id,
|
||||||
|
int_or_none,
|
||||||
|
PostProcessingError,
|
||||||
|
strip_or_none,
|
||||||
|
str_or_none,
|
||||||
|
str_to_int,
|
||||||
|
traverse_obj,
|
||||||
|
try_call,
|
||||||
|
unified_timestamp,
|
||||||
|
)
|
||||||
|
|
||||||
|
|
||||||
|
class RobloxIE(InfoExtractor):
|
||||||
|
_VALID_URL = r'https?://(?:www\.)?roblox\.com/library/(?P<id>\d+)'
|
||||||
|
_TESTS = [{
|
||||||
|
# UGC Audio
|
||||||
|
'url': 'https://www.roblox.com/library/7910582982/Backrooms-Ambiance-High-Quality',
|
||||||
|
'md5': '',
|
||||||
|
'info_dict': {
|
||||||
|
'id': '7910582982',
|
||||||
|
'ext': 'ogg',
|
||||||
|
'title': 'Backrooms Ambiance (High Quality)',
|
||||||
|
'description': 'Found an actual higher quality of the sound.',
|
||||||
|
'uploader': 'ChaseDJ549',
|
||||||
|
'uploader_id': '412014916',
|
||||||
|
'categories': ['Horror'],
|
||||||
|
'like_count': int,
|
||||||
|
'timestamp': 1636142127,
|
||||||
|
'modified_timestamp': 1656694893
|
||||||
|
},
|
||||||
|
}]
|
||||||
|
|
||||||
|
def _real_extract(self, url):
|
||||||
|
asset_id = self._match_id(url)
|
||||||
|
webpage = self._download_webpage(url, asset_id)
|
||||||
|
|
||||||
|
item_container_div = get_element_html_by_id('item-container', webpage)
|
||||||
|
item_container_attrs = extract_attributes(item_container_div[:item_container_div.find('>')+1])
|
||||||
|
asset_type = item_container_attrs.get('data-asset-type')
|
||||||
|
if asset_type and (asset_type not in ('Audio', 'Video')):
|
||||||
|
raise ExtractorError('This asset is not an audio/video', expected=True)
|
||||||
|
asset_uploader_id, asset_uploader_name = self._search_regex(
|
||||||
|
r'>By <a.+href=["\']https?://(?:www\.)?roblox\.com/users/(?P<id>\d+)[^"\']*["\'][^>]*>@?(?P<name>\w+)</a',
|
||||||
|
webpage, 'asset creator', fatal=False, group=('id', 'name'))
|
||||||
|
|
||||||
|
is_logged_out = not self._get_cookies('https://roblox.com').get('.ROBLOSECURITY')
|
||||||
|
toolbox_result = traverse_obj(
|
||||||
|
self._download_json(f'https://apis.roblox.com/toolbox-service/v1/items/details', asset_id, query={'assetIds': asset_id},
|
||||||
|
note='Downloading extra metadata JSON', errnote=False if is_logged_out else 'Unable to download extra metadata JSON',
|
||||||
|
fatal=False),
|
||||||
|
('data', ...), default={}, expected_type=dict, get_all=False)
|
||||||
|
toolbox_asset_data = toolbox_result.get('asset') or {}
|
||||||
|
toolbox_creator_data = toolbox_result.get('creator') or {}
|
||||||
|
toolbox_audio_data = toolbox_asset_data.get('audioDetails') or {}
|
||||||
|
|
||||||
|
info_dict = {
|
||||||
|
'id': asset_id,
|
||||||
|
'title': toolbox_asset_data.get('name') or item_container_attrs.get('data-item-name'),
|
||||||
|
'uploader': toolbox_creator_data.get('name') or asset_uploader_name,
|
||||||
|
'uploader_id': str_or_none(toolbox_creator_data.get('id')) or asset_uploader_id,
|
||||||
|
# TODO: Sound effects have separate kinds of categories
|
||||||
|
'categories': toolbox_asset_data.get('assetGenres') or [strip_or_none(get_element_by_class('item-genre', webpage))],
|
||||||
|
'like_count': str_to_int(extract_attributes(get_element_html_by_class('favoriteCount', webpage)).get('title')),
|
||||||
|
'timestamp': unified_timestamp(toolbox_asset_data.get('createdUtc')),
|
||||||
|
'modified_timestamp': unified_timestamp(toolbox_asset_data.get('updatedUtc')), # TODO: Extract from webpage
|
||||||
|
'track': toolbox_audio_data.get('title'),
|
||||||
|
'artist': toolbox_audio_data.get('artist'),
|
||||||
|
'genre': try_call(lambda: toolbox_audio_data['musicGenre'].capitalize())
|
||||||
|
}
|
||||||
|
|
||||||
|
cdn_result = self._download_json(
|
||||||
|
f'https://assetdelivery.roblox.com/v1/assetId/{asset_id}',
|
||||||
|
asset_id, note='Downloading file data JSON', headers={
|
||||||
|
'Accept': 'application/json',
|
||||||
|
'roblox-browser-asset-request': 'true'
|
||||||
|
})
|
||||||
|
asset_file_url = cdn_result.get('location')
|
||||||
|
if not asset_file_url:
|
||||||
|
if asset_type == 'Audio':
|
||||||
|
media_play_icon_div = get_element_html_by_class('MediaPlayerIcon')
|
||||||
|
if not media_play_icon_div:
|
||||||
|
self.raise_no_formats('This audio is unavailable', expected=True, video_id=asset_id)
|
||||||
|
asset_file_url = self._search_regex(r'data-mediathumb-url=["\']https?://[^"\']+["\']', media_play_icon_div, 'audio preview URL')
|
||||||
|
elif is_logged_out: # assetdelivery API randomly requires auth cookies
|
||||||
|
self.raise_login_required(metadata_available=True)
|
||||||
|
else:
|
||||||
|
self.raise_no_formats(
|
||||||
|
traverse_obj(cdn_result, ('errors', ..., 'message'), default='Unable to fetch asset', expected_type=str, get_all=False),
|
||||||
|
video_id=asset_id)
|
||||||
|
|
||||||
|
if asset_file_url:
|
||||||
|
# Assets have no file extension and use binary/octet-stream as Content-Type
|
||||||
|
pp = FFmpegPostProcessor(self._downloader)
|
||||||
|
self.to_screen(f'{asset_id}: Checking file format with ffprobe')
|
||||||
|
try:
|
||||||
|
metadata = pp.get_metadata_object(asset_file_url)
|
||||||
|
except PostProcessingError as err:
|
||||||
|
raise ExtractorError(err.msg, expected=True)
|
||||||
|
|
||||||
|
v_stream = a_stream = {}
|
||||||
|
for stream in metadata['streams']:
|
||||||
|
if stream['codec_type'] == 'video':
|
||||||
|
v_stream = stream
|
||||||
|
elif stream['codec_type'] == 'audio':
|
||||||
|
a_stream = stream
|
||||||
|
|
||||||
|
info_dict['formats'] = [{
|
||||||
|
'url': asset_file_url,
|
||||||
|
'ext': 'mp4' if 'mp4' in metadata['format']['format_name'] else metadata['format']['format_name'].split(',')[-1],
|
||||||
|
'vcodec': v_stream.get('codec_name'),
|
||||||
|
'acodec': a_stream.get('codec_name'),
|
||||||
|
'tbr': int_or_none(metadata['format'].get('bit_rate'), scale=1000),
|
||||||
|
'vbr': int_or_none(v_stream.get('bit_rate'), scale=1000),
|
||||||
|
'abr': int_or_none(a_stream.get('bit_rate'), scale=1000),
|
||||||
|
'height': int_or_none(v_stream.get('height')),
|
||||||
|
'width': int_or_none(v_stream.get('width')),
|
||||||
|
'filesize': float_or_none(metadata['format'].get('size'))
|
||||||
|
}]
|
||||||
|
|
||||||
|
return info_dict
|
Loading…
Reference in New Issue
Block a user