fix: fix shorts-only playlists by using on-demand ctoken generator

this will allow for retrieving playlist videos as playlistVideoRenderer
which we already have a parser for. another benefit is being able to
list videos in a playlist which have been deleted/made private.
also fixes livestream parsing

requirements.txt

@@ -14,3 +14,4 @@ requests>=2.32.3
 yt_dlp
 brotli>=1.1.0
 dateparser>=1.2.2
+bbpb>=1.4.2

ythdd_extractor.py

@@ -1,6 +1,7 @@
 #!/usr/bin/python3
 import brotli, yt_dlp, requests, json, time
 from ythdd_globals import safeTraverse
+import ythdd_proto
 import ythdd_globals
 
 ytdl_opts = {
@@ -518,12 +519,15 @@ def WEBgetVideoComments(ctoken: str) -> tuple:
 
 	return actual_comments, new_continuation
 
-def WEBextractPlaylist(plid: str = "", ctoken: str = "", prefix: str = "VL"):
+def WEBextractPlaylist(plid: str = "", ctoken: str = ""):
 
-	additional_context = {'browseId': prefix + plid}
+	# if ctoken has been provided, use it
 	if ctoken:
 		# playlist id can be omitted if ctoken is provided
 		additional_context = {'continuation': ctoken}
+	else:
+		# try to create ctoken which will allow for accessing the full playlist, including delisted/deleted videos
+		additional_context = {'continuation': ythdd_proto.producePlaylistContinuation(plid, offset=0)}
 
 	context = makeWebContext(additional_context)
 
@@ -540,52 +544,15 @@ def WEBextractPlaylist(plid: str = "", ctoken: str = "", prefix: str = "VL"):
 	if not ctoken:
 		metadata = {
 			"header": safeTraverse(resp_json, ["header"]),
-			#"microformat": safeTraverse(resp_json, ["microformat"]),
 			"sidebar": safeTraverse(resp_json, ["sidebar"])
 		}
 
-	# TODO (after python protodec implementation/wrapper is done):
-	# # try to extract ctoken containing the full playlist, including delisted/deleted videos
-	# full_playlist_ctoken = urllib.parse.quote(protodec.from_json({
-	# 	"80226972:embedded": {
-	# 		"2:string": prefix + plid,
-	# 		"3:base64": {
-	# 			"1:varint": request_count, # todo: increment by 200 with an external index
-	# 			"15:string": "PT:" + urllib.parse.quote(protodec.from_json({"1:varint": index})),
-	# 			"104:embedded": {"1:0:varint": 0}
-	# 		},
-	# 		"35:string": plid
-	# 	}
-	# }))
-	# # if ctoken creation succeeded
-	# if full_playlist_ctoken:
-	# 	# make another request
-	# 	response = requests.post(
-	# 		'https://www.youtube.com/youtubei/v1/browse?prettyPrint=false',
-	# 		headers = ythdd_globals.getHeaders(),
-	# 		json    = makeWebContext({'continuation': full_playlist_ctoken})
-	# 	)
-	# 	resp_json = json.loads(response.text)
-	# else:
-	# 	print("error(WEBextractPlaylist): full playlist metadata extraction failed. Delisted/deleted videos will be missing.")
-
 	# extract continuation
-	new_continuation = None
-	if ctoken:
-		# subsequent request
 	new_continuation = safeTraverse(resp_json, ["onResponseReceivedActions", 0, "appendContinuationItemsAction", "continuationItems", -1, "continuationItemRenderer", "continuationEndpoint", "continuationCommand", "token"])
-	else:
-		# first-time request
-		new_continuation = safeTraverse(resp_json, ["contents", "twoColumnBrowseResultsRenderer", "tabs", 0, "tabRenderer", "content", "sectionListRenderer", "contents", 0, "itemSectionRenderer", "contents", 0, "playlistVideoListRenderer", "contents", -1, "continuationItemRenderer", "continuationEndpoint", "commandExecutorCommand", "commands", -1, "continuationCommand", "token"])
 
 	# "best-effort" playlist's videos extraction
 	# "best-effort" because None's (unsuccessful video extraction = None) are passed as they are
-	# warning! todo: iterate over this, as shorts cannot be currently extracted (they use richGridRenderer, not playlistVideoListRenderer)
-	videos = None
-	if ctoken: # or full_playlist_ctoken:
 	videos = safeTraverse(resp_json, ["onResponseReceivedActions", 0, "appendContinuationItemsAction", "continuationItems"]) # includes continuation as last element of list, which will be ignored
-	else:
-		videos = safeTraverse(resp_json, ["contents", "twoColumnBrowseResultsRenderer", "tabs", 0, "tabRenderer", "content", "sectionListRenderer", "contents", 0, "itemSectionRenderer", "contents", 0, "playlistVideoListRenderer", "contents"])
 
 	return metadata, new_continuation, videos
 

ythdd_proto.py

@@ -0,0 +1,85 @@
+from ythdd_globals import safeTraverse
+import base64
+import blackboxprotobuf as bbpb
+import json
+import urllib.parse
+import ythdd_globals
+
+def bbpbToB64(msg_and_typedef: tuple, urlsafe: bool = False, padding: bool = False) -> str:
+	encoded_protobuf     = bbpb.encode_message(*msg_and_typedef)
+	if urlsafe:
+		b64_protobuf     = base64.urlsafe_b64encode(encoded_protobuf)
+	else:
+		b64_protobuf     = base64.b64encode(encoded_protobuf)
+	if padding:
+		url_encoded_b64  = urllib.parse.quote(b64_protobuf.decode())
+	else:
+		url_encoded_b64  = b64_protobuf.decode().rstrip('=')
+	return url_encoded_b64
+
+def fdictToBbpb(msg: dict) -> tuple:
+	# Requires Python 3.7+ or CPython 3.6+,
+	# as these versions preserve dictionary insertion order.
+	# Structural matching (match, case) requires Python 3.10+.
+	clean_msg = {}
+	clean_type = {}
+	for key in msg:
+		num, type = key.split(":")
+
+		match type:
+			case "message":
+				# if the type is an embedded message
+				internal_msg, internal_type = fdictToBbpb(msg[key])
+				# msg can just be appended as usual
+				clean_msg[num]  = internal_msg
+				# type contains more fields than normally
+				clean_type[num] = {
+					'field_order': list(internal_msg.keys()),
+					'message_typedef': internal_type,
+					'type': type
+				}
+
+			case "base64" | "base64u" | "base64p" | "base64up":
+				# if the type is a base64-embedded message
+				internal_msg, internal_type = fdictToBbpb(msg[key])
+				match type.removeprefix("base64"):
+					case "":
+						b64_encoded_msg = bbpbToB64((internal_msg, internal_type))
+					case "u":
+						b64_encoded_msg = bbpbToB64((internal_msg, internal_type), urlsafe=True)
+					case "p":
+						b64_encoded_msg = bbpbToB64((internal_msg, internal_type), padding=True)
+					case "up":
+						b64_encoded_msg = bbpbToB64((internal_msg, internal_type), urlsafe=True, padding=True)
+				clean_msg[num]  = b64_encoded_msg
+				clean_type[num] = {'type': 'string'}
+
+			case "int" | "string":
+				clean_msg[num]  = msg[key]
+				clean_type[num] = {'type': type}
+
+			case _:
+				raise KeyError(f'error(fmsgToBBPBTuple): invalid key "{type}"')
+
+
+	return (clean_msg, clean_type)
+
+def producePlaylistContinuation(plid: str, offset: int = 0) -> str:
+	msge = {
+		'80226972:message': {
+			'2:string': f'VL{plid}',
+			'3:base64': {
+				'1:int': int(offset / 100),
+				'15:string': f'PT:{bbpbToB64(fdictToBbpb({"1:int": offset}))}',
+				'104:message': {
+					'1:int': 0
+				}
+			},
+			'35:string': plid
+		}
+	}
+
+	bbpb_dicts = fdictToBbpb(msge)
+	b64_ctoken = bbpbToB64(bbpb_dicts, urlsafe=True, padding=True)
+
+	return b64_ctoken

ythdd_struct_parser.py

@@ -313,6 +313,7 @@ def parseRenderers(entry: dict, context: dict = {}) -> dict:
 			}
 
 		case "playlistVideoRenderer":
+			# used by all content inside of playlists which have at least one non-shorts video/livestream
 
 			video_id       = safeTraverse(entry, ["playlistVideoRenderer", "videoId"], default="UnknownVideoId")
 			title          = safeTraverse(entry, ["playlistVideoRenderer", "title", "runs", 0, "text"], default="Unknown video title")
@@ -321,7 +322,11 @@ def parseRenderers(entry: dict, context: dict = {}) -> dict:
 			video_index    = int(safeTraverse(entry, ["playlistVideoRenderer", "index", "simpleText"], default="1")) - 1
 			length         = parseLengthFromTimeBadge(safeTraverse(entry, ["playlistVideoRenderer", "lengthText", "simpleText"], default="0:0"))
 			published_date = safeTraverse(entry, ["playlistVideoRenderer", "videoInfo", "runs", -1, "text"], default="2000-01-01")
-			published_date = published_date.removeprefix("Streamed ")
+			published_date = published_date.removeprefix("Streamed ").removeprefix(" watching")
+
+			# handle livestreams
+			if not published_date:
+				published_date = "now"
 
 			return {
 				"type": "video",