0
0
Fork 0
mirror of https://github.com/yt-dlp/yt-dlp.git synced 2024-11-24 02:35:12 +00:00
yt-dlp/yt_dlp/extractor/googledrive.py

339 lines
14 KiB
Python
Raw Normal View History

2015-06-29 06:53:21 +00:00
import re
2015-06-24 00:13:23 +00:00
from .common import InfoExtractor
2020-11-21 14:50:42 +00:00
from ..compat import compat_parse_qs
from ..utils import (
ExtractorError,
determine_ext,
get_element_by_class,
2015-12-21 02:05:34 +00:00
int_or_none,
lowercase_escape,
2020-11-21 14:50:42 +00:00
try_get,
update_url_query,
)
2015-06-24 00:13:23 +00:00
2015-12-21 02:05:34 +00:00
class GoogleDriveIE(InfoExtractor):
_VALID_URL = r'''(?x)
https?://
(?:
(?:docs|drive)\.google\.com/
(?:
(?:uc|open)\?.*?id=|
file/d/
)|
video\.google\.com/get_player\?.*?docid=
)
(?P<id>[a-zA-Z0-9_-]{28,})
'''
_TESTS = [{
2015-12-21 02:05:34 +00:00
'url': 'https://drive.google.com/file/d/0ByeS4oOUV-49Zzh4R1J6R09zazQ/edit?pli=1',
'md5': '5c602afbbf2c1db91831f5d82f678554',
2015-06-29 06:53:21 +00:00
'info_dict': {
2015-12-21 02:05:34 +00:00
'id': '0ByeS4oOUV-49Zzh4R1J6R09zazQ',
2015-06-29 06:53:21 +00:00
'ext': 'mp4',
2015-12-21 02:05:34 +00:00
'title': 'Big Buck Bunny.mp4',
'duration': 45,
2015-06-29 06:53:21 +00:00
}
}, {
# video can't be watched anonymously due to view count limit reached,
2019-03-09 12:14:41 +00:00
# but can be downloaded (see https://github.com/ytdl-org/youtube-dl/issues/14046)
'url': 'https://drive.google.com/file/d/0B-vUyvmDLdWDcEt4WjBqcmI2XzQ/view',
2020-11-21 14:50:42 +00:00
'only_matching': True,
}, {
# video id is longer than 28 characters
'url': 'https://drive.google.com/file/d/1ENcQ_jeCuj7y19s66_Ou9dRP4GKGsodiDQ/edit',
'only_matching': True,
}, {
'url': 'https://drive.google.com/open?id=0B2fjwgkl1A_CX083Tkowdmt6d28',
'only_matching': True,
}, {
'url': 'https://drive.google.com/uc?id=0B2fjwgkl1A_CX083Tkowdmt6d28',
'only_matching': True,
}]
2015-12-21 02:05:34 +00:00
_FORMATS_EXT = {
'5': 'flv',
'6': 'flv',
'13': '3gp',
'17': '3gp',
'18': 'mp4',
'22': 'mp4',
'34': 'flv',
'35': 'flv',
'36': '3gp',
'37': 'mp4',
'38': 'mp4',
'43': 'webm',
'44': 'webm',
'45': 'webm',
'46': 'webm',
'59': 'mp4',
}
_BASE_URL_CAPTIONS = 'https://drive.google.com/timedtext'
_CAPTIONS_ENTRY_TAG = {
'subtitles': 'track',
'automatic_captions': 'target',
}
_caption_formats_ext = []
_captions_xml = None
2015-06-29 06:53:21 +00:00
@classmethod
def _extract_embed_urls(cls, url, webpage):
2015-06-29 06:53:21 +00:00
mobj = re.search(
r'<iframe[^>]+src="https?://(?:video\.google\.com/get_player\?.*?docid=|(?:docs|drive)\.google\.com/file/d/)(?P<id>[a-zA-Z0-9_-]{28,})',
2015-06-29 06:53:21 +00:00
webpage)
if mobj:
yield 'https://drive.google.com/file/d/%s' % mobj.group('id')
2015-06-29 06:53:21 +00:00
def _download_subtitles_xml(self, video_id, subtitles_id, hl):
if self._captions_xml:
return
self._captions_xml = self._download_xml(
self._BASE_URL_CAPTIONS, video_id, query={
'id': video_id,
'vid': subtitles_id,
'hl': hl,
'v': video_id,
'type': 'list',
'tlangs': '1',
'fmts': '1',
'vssids': '1',
}, note='Downloading subtitles XML',
errnote='Unable to download subtitles XML', fatal=False)
if self._captions_xml:
for f in self._captions_xml.findall('format'):
if f.attrib.get('fmt_code') and not f.attrib.get('default'):
self._caption_formats_ext.append(f.attrib['fmt_code'])
def _get_captions_by_type(self, video_id, subtitles_id, caption_type,
origin_lang_code=None):
if not subtitles_id or not caption_type:
return
captions = {}
for caption_entry in self._captions_xml.findall(
self._CAPTIONS_ENTRY_TAG[caption_type]):
caption_lang_code = caption_entry.attrib.get('lang_code')
if not caption_lang_code:
continue
caption_format_data = []
for caption_format in self._caption_formats_ext:
query = {
'vid': subtitles_id,
'v': video_id,
'fmt': caption_format,
'lang': (caption_lang_code if origin_lang_code is None
else origin_lang_code),
'type': 'track',
'name': '',
'kind': '',
}
if origin_lang_code is not None:
query.update({'tlang': caption_lang_code})
caption_format_data.append({
'url': update_url_query(self._BASE_URL_CAPTIONS, query),
'ext': caption_format,
})
captions[caption_lang_code] = caption_format_data
return captions
def _get_subtitles(self, video_id, subtitles_id, hl):
if not subtitles_id or not hl:
return
self._download_subtitles_xml(video_id, subtitles_id, hl)
if not self._captions_xml:
return
return self._get_captions_by_type(video_id, subtitles_id, 'subtitles')
def _get_automatic_captions(self, video_id, subtitles_id, hl):
if not subtitles_id or not hl:
return
self._download_subtitles_xml(video_id, subtitles_id, hl)
if not self._captions_xml:
return
track = self._captions_xml.find('track')
if track is None:
return
origin_lang_code = track.attrib.get('lang_code')
if not origin_lang_code:
return
return self._get_captions_by_type(
video_id, subtitles_id, 'automatic_captions', origin_lang_code)
2015-06-29 06:53:21 +00:00
def _real_extract(self, url):
video_id = self._match_id(url)
2020-11-21 14:50:42 +00:00
video_info = compat_parse_qs(self._download_webpage(
'https://drive.google.com/get_video_info',
video_id, 'Downloading video webpage', query={'docid': video_id}))
2020-11-21 14:50:42 +00:00
def get_value(key):
return try_get(video_info, lambda x: x[key][0])
2015-06-29 06:53:21 +00:00
2020-11-21 14:50:42 +00:00
reason = get_value('reason')
title = get_value('title')
formats = []
2020-11-21 14:50:42 +00:00
fmt_stream_map = (get_value('fmt_stream_map') or '').split(',')
fmt_list = (get_value('fmt_list') or '').split(',')
if fmt_stream_map and fmt_list:
resolutions = {}
for fmt in fmt_list:
mobj = re.search(
r'^(?P<format_id>\d+)/(?P<width>\d+)[xX](?P<height>\d+)', fmt)
if mobj:
resolutions[mobj.group('format_id')] = (
int(mobj.group('width')), int(mobj.group('height')))
2015-06-24 00:13:23 +00:00
for fmt_stream in fmt_stream_map:
fmt_stream_split = fmt_stream.split('|')
if len(fmt_stream_split) < 2:
continue
format_id, format_url = fmt_stream_split[:2]
f = {
'url': lowercase_escape(format_url),
'format_id': format_id,
'ext': self._FORMATS_EXT[format_id],
}
resolution = resolutions.get(format_id)
if resolution:
f.update({
'width': resolution[0],
'height': resolution[1],
})
formats.append(f)
source_url = update_url_query(
'https://drive.google.com/uc', {
'id': video_id,
'export': 'download',
})
def request_source_file(source_url, kind):
return self._request_webpage(
source_url, video_id, note='Requesting %s file' % kind,
errnote='Unable to request %s file' % kind, fatal=False)
urlh = request_source_file(source_url, 'source')
if urlh:
def add_source_format(urlh):
nonlocal title
if not title:
title = self._search_regex(
r'\bfilename="([^"]+)"', urlh.headers.get('Content-Disposition'),
'title', default=None)
formats.append({
# Use redirect URLs as download URLs in order to calculate
# correct cookies in _calc_cookies.
# Using original URLs may result in redirect loop due to
# google.com's cookies mistakenly used for googleusercontent.com
# redirect URLs (see #23919).
'url': urlh.geturl(),
'ext': determine_ext(title, 'mp4').lower(),
'format_id': 'source',
'quality': 1,
})
if urlh.headers.get('Content-Disposition'):
add_source_format(urlh)
else:
confirmation_webpage = self._webpage_read_content(
urlh, url, video_id, note='Downloading confirmation page',
errnote='Unable to confirm download', fatal=False)
if confirmation_webpage:
confirm = self._search_regex(
r'confirm=([^&"\']+)', confirmation_webpage,
'confirmation code', default=None)
if confirm:
confirmed_source_url = update_url_query(source_url, {
'confirm': confirm,
})
urlh = request_source_file(confirmed_source_url, 'confirmed source')
if urlh and urlh.headers.get('Content-Disposition'):
add_source_format(urlh)
else:
self.report_warning(
get_element_by_class('uc-error-subcaption', confirmation_webpage)
or get_element_by_class('uc-error-caption', confirmation_webpage)
or 'unable to extract confirmation code')
2020-11-21 14:50:42 +00:00
if not formats and reason:
if title:
self.raise_no_formats(reason, expected=True)
else:
raise ExtractorError(reason, expected=True)
2020-11-21 14:50:42 +00:00
hl = get_value('hl')
subtitles_id = None
2020-11-21 14:50:42 +00:00
ttsurl = get_value('ttsurl')
if ttsurl:
# the video Id for subtitles will be the last value in the ttsurl
# query string
subtitles_id = ttsurl.encode('utf-8').decode(
'unicode_escape').split('=')[-1]
self.cookiejar.clear(domain='.google.com', path='/', name='NID')
2015-06-24 00:13:23 +00:00
return {
'id': video_id,
'title': title,
2020-11-21 14:50:42 +00:00
'thumbnail': 'https://drive.google.com/thumbnail?id=' + video_id,
'duration': int_or_none(get_value('length_seconds')),
2015-12-21 02:05:34 +00:00
'formats': formats,
'subtitles': self.extract_subtitles(video_id, subtitles_id, hl),
'automatic_captions': self.extract_automatic_captions(
video_id, subtitles_id, hl),
2015-06-24 00:13:23 +00:00
}
class GoogleDriveFolderIE(InfoExtractor):
IE_NAME = 'GoogleDrive:Folder'
_VALID_URL = r'https?://(?:docs|drive)\.google\.com/drive/folders/(?P<id>[\w-]{28,})'
_TESTS = [{
'url': 'https://drive.google.com/drive/folders/1dQ4sx0-__Nvg65rxTSgQrl7VyW_FZ9QI',
'info_dict': {
'id': '1dQ4sx0-__Nvg65rxTSgQrl7VyW_FZ9QI',
'title': 'Forrest'
},
'playlist_count': 3,
}]
_BOUNDARY = '=====vc17a3rwnndj====='
_REQUEST = "/drive/v2beta/files?openDrive=true&reason=102&syncType=0&errorRecovery=false&q=trashed%20%3D%20false%20and%20'{folder_id}'%20in%20parents&fields=kind%2CnextPageToken%2Citems(kind%2CmodifiedDate%2CmodifiedByMeDate%2ClastViewedByMeDate%2CfileSize%2Cowners(kind%2CpermissionId%2Cid)%2ClastModifyingUser(kind%2CpermissionId%2Cid)%2ChasThumbnail%2CthumbnailVersion%2Ctitle%2Cid%2CresourceKey%2Cshared%2CsharedWithMeDate%2CuserPermission(role)%2CexplicitlyTrashed%2CmimeType%2CquotaBytesUsed%2Ccopyable%2CfileExtension%2CsharingUser(kind%2CpermissionId%2Cid)%2Cspaces%2Cversion%2CteamDriveId%2ChasAugmentedPermissions%2CcreatedDate%2CtrashingUser(kind%2CpermissionId%2Cid)%2CtrashedDate%2Cparents(id)%2CshortcutDetails(targetId%2CtargetMimeType%2CtargetLookupStatus)%2Ccapabilities(canCopy%2CcanDownload%2CcanEdit%2CcanAddChildren%2CcanDelete%2CcanRemoveChildren%2CcanShare%2CcanTrash%2CcanRename%2CcanReadTeamDrive%2CcanMoveTeamDriveItem)%2Clabels(starred%2Ctrashed%2Crestricted%2Cviewed))%2CincompleteSearch&appDataFilter=NO_APP_DATA&spaces=drive&pageToken={page_token}&maxResults=50&supportsTeamDrives=true&includeItemsFromAllDrives=true&corpora=default&orderBy=folder%2Ctitle_natural%20asc&retryCount=0&key={key} HTTP/1.1"
_DATA = f'''--{_BOUNDARY}
content-type: application/http
content-transfer-encoding: binary
GET %s
--{_BOUNDARY}
'''
def _call_api(self, folder_id, key, data, **kwargs):
response = self._download_webpage(
'https://clients6.google.com/batch/drive/v2beta',
folder_id, data=data.encode('utf-8'),
headers={
'Content-Type': 'text/plain;charset=UTF-8;',
'Origin': 'https://drive.google.com',
}, query={
'$ct': f'multipart/mixed; boundary="{self._BOUNDARY}"',
'key': key
}, **kwargs)
return self._search_json('', response, 'api response', folder_id, **kwargs) or {}
def _get_folder_items(self, folder_id, key):
page_token = ''
while page_token is not None:
request = self._REQUEST.format(folder_id=folder_id, page_token=page_token, key=key)
page = self._call_api(folder_id, key, self._DATA % request)
yield from page['items']
page_token = page.get('nextPageToken')
def _real_extract(self, url):
folder_id = self._match_id(url)
webpage = self._download_webpage(url, folder_id)
key = self._search_regex(r'"(\w{39})"', webpage, 'key')
folder_info = self._call_api(folder_id, key, self._DATA % f'/drive/v2beta/files/{folder_id} HTTP/1.1', fatal=False)
return self.playlist_from_matches(
self._get_folder_items(folder_id, key), folder_id, folder_info.get('title'),
ie=GoogleDriveIE, getter=lambda item: f'https://drive.google.com/file/d/{item["id"]}')