mirror of
https://codeberg.org/polarisfm/youtube-dl.git
synced 2024-12-21 23:32:21 +00:00
[twitch] Switch streams to GraphQL and refactor (closes #26535)
This commit is contained in:
parent
82ef02e936
commit
defc820b70
1 changed files with 81 additions and 61 deletions
|
@ -24,7 +24,6 @@ from ..utils import (
|
|||
parse_duration,
|
||||
parse_iso8601,
|
||||
qualities,
|
||||
str_or_none,
|
||||
try_get,
|
||||
unified_timestamp,
|
||||
update_url_query,
|
||||
|
@ -337,19 +336,27 @@ def _make_video_result(node):
|
|||
class TwitchGraphQLBaseIE(TwitchBaseIE):
|
||||
_PAGE_LIMIT = 100
|
||||
|
||||
def _download_gql(self, video_id, op, variables, sha256_hash, note, fatal=True):
|
||||
_OPERATION_HASHES = {
|
||||
'CollectionSideBar': '27111f1b382effad0b6def325caef1909c733fe6a4fbabf54f8d491ef2cf2f14',
|
||||
'FilterableVideoTower_Videos': 'a937f1d22e269e39a03b509f65a7490f9fc247d7f83d6ac1421523e3b68042cb',
|
||||
'ClipsCards__User': 'b73ad2bfaecfd30a9e6c28fada15bd97032c83ec77a0440766a56fe0bd632777',
|
||||
'ChannelCollectionsContent': '07e3691a1bad77a36aba590c351180439a40baefc1c275356f40fc7082419a84',
|
||||
'StreamMetadata': '1c719a40e481453e5c48d9bb585d971b8b372f8ebb105b17076722264dfa5b3e',
|
||||
'ComscoreStreamingQuery': 'e1edae8122517d013405f237ffcc124515dc6ded82480a88daef69c83b53ac01',
|
||||
'VideoPreviewOverlay': '3006e77e51b128d838fa4e835723ca4dc9a05c5efd4466c1085215c6e437e65c',
|
||||
}
|
||||
|
||||
def _download_gql(self, video_id, ops, note, fatal=True):
|
||||
for op in ops:
|
||||
op['extensions'] = {
|
||||
'persistedQuery': {
|
||||
'version': 1,
|
||||
'sha256Hash': self._OPERATION_HASHES[op['operationName']],
|
||||
}
|
||||
}
|
||||
return self._download_json(
|
||||
'https://gql.twitch.tv/gql', video_id, note,
|
||||
data=json.dumps({
|
||||
'operationName': op,
|
||||
'variables': variables,
|
||||
'extensions': {
|
||||
'persistedQuery': {
|
||||
'version': 1,
|
||||
'sha256Hash': sha256_hash,
|
||||
}
|
||||
}
|
||||
}).encode(),
|
||||
data=json.dumps(ops).encode(),
|
||||
headers={
|
||||
'Content-Type': 'text/plain;charset=UTF-8',
|
||||
'Client-ID': self._CLIENT_ID,
|
||||
|
@ -369,14 +376,15 @@ class TwitchCollectionIE(TwitchGraphQLBaseIE):
|
|||
}]
|
||||
|
||||
_OPERATION_NAME = 'CollectionSideBar'
|
||||
_SHA256_HASH = '27111f1b382effad0b6def325caef1909c733fe6a4fbabf54f8d491ef2cf2f14'
|
||||
|
||||
def _real_extract(self, url):
|
||||
collection_id = self._match_id(url)
|
||||
collection = self._download_gql(
|
||||
collection_id, self._OPERATION_NAME,
|
||||
{'collectionID': collection_id}, self._SHA256_HASH,
|
||||
'Downloading collection GraphQL')['data']['collection']
|
||||
collection_id, [{
|
||||
'operationName': self._OPERATION_NAME,
|
||||
'variables': {'collectionID': collection_id},
|
||||
}],
|
||||
'Downloading collection GraphQL')[0]['data']['collection']
|
||||
title = collection.get('title')
|
||||
entries = []
|
||||
for edge in collection['items']['edges']:
|
||||
|
@ -403,14 +411,16 @@ class TwitchPlaylistBaseIE(TwitchGraphQLBaseIE):
|
|||
if cursor:
|
||||
variables['cursor'] = cursor
|
||||
page = self._download_gql(
|
||||
channel_name, self._OPERATION_NAME, variables,
|
||||
self._SHA256_HASH,
|
||||
channel_name, [{
|
||||
'operationName': self._OPERATION_NAME,
|
||||
'variables': variables,
|
||||
}],
|
||||
'Downloading %ss GraphQL page %s' % (self._NODE_KIND, page_num),
|
||||
fatal=False)
|
||||
if not page:
|
||||
break
|
||||
edges = try_get(
|
||||
page, lambda x: x['data']['user'][entries_key]['edges'], list)
|
||||
page, lambda x: x[0]['data']['user'][entries_key]['edges'], list)
|
||||
if not edges:
|
||||
break
|
||||
for edge in edges:
|
||||
|
@ -553,7 +563,6 @@ class TwitchVideosIE(TwitchPlaylistBaseIE):
|
|||
'views': 'Popular',
|
||||
}
|
||||
|
||||
_SHA256_HASH = 'a937f1d22e269e39a03b509f65a7490f9fc247d7f83d6ac1421523e3b68042cb'
|
||||
_OPERATION_NAME = 'FilterableVideoTower_Videos'
|
||||
_ENTRY_KIND = 'video'
|
||||
_EDGE_KIND = 'VideoEdge'
|
||||
|
@ -622,7 +631,6 @@ class TwitchVideosClipsIE(TwitchPlaylistBaseIE):
|
|||
# NB: values other than 20 result in skipped videos
|
||||
_PAGE_LIMIT = 20
|
||||
|
||||
_SHA256_HASH = 'b73ad2bfaecfd30a9e6c28fada15bd97032c83ec77a0440766a56fe0bd632777'
|
||||
_OPERATION_NAME = 'ClipsCards__User'
|
||||
_ENTRY_KIND = 'clip'
|
||||
_EDGE_KIND = 'ClipEdge'
|
||||
|
@ -680,7 +688,6 @@ class TwitchVideosCollectionsIE(TwitchPlaylistBaseIE):
|
|||
'playlist_mincount': 3,
|
||||
}]
|
||||
|
||||
_SHA256_HASH = '07e3691a1bad77a36aba590c351180439a40baefc1c275356f40fc7082419a84'
|
||||
_OPERATION_NAME = 'ChannelCollectionsContent'
|
||||
_ENTRY_KIND = 'collection'
|
||||
_EDGE_KIND = 'CollectionsItemEdge'
|
||||
|
@ -717,7 +724,7 @@ class TwitchVideosCollectionsIE(TwitchPlaylistBaseIE):
|
|||
playlist_title='%s - Collections' % channel_name)
|
||||
|
||||
|
||||
class TwitchStreamIE(TwitchBaseIE):
|
||||
class TwitchStreamIE(TwitchGraphQLBaseIE):
|
||||
IE_NAME = 'twitch:stream'
|
||||
_VALID_URL = r'''(?x)
|
||||
https?://
|
||||
|
@ -774,28 +781,43 @@ class TwitchStreamIE(TwitchBaseIE):
|
|||
else super(TwitchStreamIE, cls).suitable(url))
|
||||
|
||||
def _real_extract(self, url):
|
||||
channel_name = self._match_id(url)
|
||||
channel_name = self._match_id(url).lower()
|
||||
|
||||
access_token = self._download_access_token(channel_name)
|
||||
gql = self._download_gql(
|
||||
channel_name, [{
|
||||
'operationName': 'StreamMetadata',
|
||||
'variables': {'channelLogin': channel_name},
|
||||
}, {
|
||||
'operationName': 'ComscoreStreamingQuery',
|
||||
'variables': {
|
||||
'channel': channel_name,
|
||||
'clipSlug': '',
|
||||
'isClip': False,
|
||||
'isLive': True,
|
||||
'isVodOrCollection': False,
|
||||
'vodID': '',
|
||||
},
|
||||
}, {
|
||||
'operationName': 'VideoPreviewOverlay',
|
||||
'variables': {'login': channel_name},
|
||||
}],
|
||||
'Downloading stream GraphQL')
|
||||
|
||||
token = access_token['token']
|
||||
channel_id = self._extract_channel_id(token, channel_name)
|
||||
user = gql[0]['data']['user']
|
||||
|
||||
stream = self._call_api(
|
||||
'kraken/streams/%s?stream_type=all' % channel_id,
|
||||
channel_id, 'Downloading stream JSON').get('stream')
|
||||
if not user:
|
||||
raise ExtractorError(
|
||||
'%s does not exist' % channel_name, expected=True)
|
||||
|
||||
stream = user['stream']
|
||||
|
||||
if not stream:
|
||||
raise ExtractorError('%s is offline' % channel_id, expected=True)
|
||||
raise ExtractorError('%s is offline' % channel_name, expected=True)
|
||||
|
||||
# Channel name may be typed if different case than the original channel name
|
||||
# (e.g. http://www.twitch.tv/TWITCHPLAYSPOKEMON) that will lead to constructing
|
||||
# an invalid m3u8 URL. Working around by use of original channel name from stream
|
||||
# JSON and fallback to lowercase if it's not available.
|
||||
channel_name = try_get(
|
||||
stream, lambda x: x['channel']['name'],
|
||||
compat_str) or channel_name.lower()
|
||||
access_token = self._download_access_token(channel_name)
|
||||
token = access_token['token']
|
||||
|
||||
stream_id = stream.get('id') or channel_name
|
||||
query = {
|
||||
'allow_source': 'true',
|
||||
'allow_audio_only': 'true',
|
||||
|
@ -808,41 +830,39 @@ class TwitchStreamIE(TwitchBaseIE):
|
|||
'token': token.encode('utf-8'),
|
||||
}
|
||||
formats = self._extract_m3u8_formats(
|
||||
'%s/api/channel/hls/%s.m3u8?%s'
|
||||
% (self._USHER_BASE, channel_name, compat_urllib_parse_urlencode(query)),
|
||||
channel_id, 'mp4')
|
||||
'%s/api/channel/hls/%s.m3u8' % (self._USHER_BASE, channel_name),
|
||||
stream_id, 'mp4', query=query)
|
||||
self._prefer_source(formats)
|
||||
|
||||
view_count = stream.get('viewers')
|
||||
timestamp = parse_iso8601(stream.get('created_at'))
|
||||
timestamp = unified_timestamp(stream.get('createdAt'))
|
||||
|
||||
channel = stream['channel']
|
||||
title = self._live_title(channel.get('display_name') or channel.get('name'))
|
||||
description = channel.get('status')
|
||||
sq_user = try_get(gql, lambda x: x[1]['data']['user'], dict) or {}
|
||||
uploader = sq_user.get('displayName')
|
||||
description = try_get(
|
||||
sq_user, lambda x: x['broadcastSettings']['title'], compat_str)
|
||||
|
||||
thumbnails = []
|
||||
for thumbnail_key, thumbnail_url in stream['preview'].items():
|
||||
m = re.search(r'(?P<width>\d+)x(?P<height>\d+)\.jpg$', thumbnail_key)
|
||||
if not m:
|
||||
continue
|
||||
thumbnails.append({
|
||||
'url': thumbnail_url,
|
||||
'width': int(m.group('width')),
|
||||
'height': int(m.group('height')),
|
||||
})
|
||||
thumbnail = url_or_none(try_get(
|
||||
gql, lambda x: x[2]['data']['user']['stream']['previewImageURL'],
|
||||
compat_str))
|
||||
|
||||
title = uploader or channel_name
|
||||
stream_type = stream.get('type')
|
||||
if stream_type in ['rerun', 'live']:
|
||||
title += ' (%s)' % stream_type
|
||||
|
||||
return {
|
||||
'id': str_or_none(stream.get('_id')) or channel_id,
|
||||
'id': stream_id,
|
||||
'display_id': channel_name,
|
||||
'title': title,
|
||||
'title': self._live_title(title),
|
||||
'description': description,
|
||||
'thumbnails': thumbnails,
|
||||
'uploader': channel.get('display_name'),
|
||||
'uploader_id': channel.get('name'),
|
||||
'thumbnail': thumbnail,
|
||||
'uploader': uploader,
|
||||
'uploader_id': channel_name,
|
||||
'timestamp': timestamp,
|
||||
'view_count': view_count,
|
||||
'formats': formats,
|
||||
'is_live': True,
|
||||
'is_live': stream_type == 'live',
|
||||
}
|
||||
|
||||
|
||||
|
|
Loading…
Reference in a new issue