import base64 import json from youtube import proto, util, yt_data_extract from youtube.util import concat_or_none # Here's what I know about the secret key (starting with ASJN_i) # *The secret key definitely contains the following information (or perhaps the information is stored at youtube's servers): # -Video id # -Offset # -Sort # *If the video id or sort in the ctoken contradicts the ASJN, the response is an error. The offset encoded outside the ASJN is ignored entirely. # *The ASJN is base64 encoded data, indicated by the fact that the character after "ASJN_i" is one of ("0", "1", "2", "3") # *The encoded data is not valid protobuf # *The encoded data (after the 5 or so bytes that are always the same) is indistinguishable from random data according to a battery of randomness tests # *The ASJN in the ctoken provided by a response changes in regular intervals of about a second or two. # *Old ASJN's continue to work, and start at the same comment even if new comments have been posted since # *The ASJN has no relation with any of the data in the response it came from def make_comment_ctoken(video_id, sort=0, offset=0, lc='', secret_key=''): video_id = proto.as_bytes(video_id) secret_key = proto.as_bytes(secret_key) page_info = proto.string(4,video_id) + proto.uint(6, sort) offset_information = proto.nested(4, page_info) + proto.uint(5, offset) if secret_key: offset_information = proto.string(1, secret_key) + offset_information page_params = proto.string(2, video_id) if lc: page_params += proto.string(6, proto.percent_b64encode(proto.string(15, lc))) result = proto.nested(2, page_params) + proto.uint(3,6) + proto.nested(6, offset_information) return base64.urlsafe_b64encode(result).decode('ascii') def comment_replies_ctoken(video_id, comment_id, max_results=500): params = proto.string(2, comment_id) + proto.uint(9, max_results) params = proto.nested(3, params) result = proto.nested(2, proto.string(2, video_id)) + proto.uint(3,6) + proto.nested(6, params) return base64.urlsafe_b64encode(result).decode('ascii') mobile_headers = { 'User-Agent': 'Mozilla/5.0 (iPhone; CPU iPhone OS 10_3_1 like Mac OS X) AppleWebKit/603.1.30 (KHTML, like Gecko) Version/10.0 Mobile/14E304 Safari/602.1', 'Accept': '*/*', 'Accept-Language': 'en-US,en;q=0.5', 'X-YouTube-Client-Name': '2', 'X-YouTube-Client-Version': '2.20180823', } def request_comments(ctoken, replies=False): if replies: # let's make it use different urls for no reason despite all the data being encoded base_url = "https://m.youtube.com/watch_comment?action_get_comment_replies=1&ctoken=" else: base_url = "https://m.youtube.com/watch_comment?action_get_comments=1&ctoken=" url = base_url + ctoken.replace("=", "%3D") + "&pbj=1" for i in range(0,8): # don't retry more than 8 times content = util.fetch_url(url, headers=mobile_headers, report_text="Retrieved comments", debug_name='request_comments') if content[0:4] == b")]}'": # random closing characters included at beginning of response for some reason content = content[4:] elif content[0:10] == b'\n, retrying") continue break polymer_json = json.loads(util.uppercase_escape(content.decode('utf-8'))) return polymer_json def single_comment_ctoken(video_id, comment_id): page_params = proto.string(2, video_id) + proto.string(6, proto.percent_b64encode(proto.string(15, comment_id))) result = proto.nested(2, page_params) + proto.uint(3,6) return base64.urlsafe_b64encode(result).decode('ascii') def post_process_comments_info(comments_info): for comment in comments_info['comments']: comment['author_url'] = concat_or_none( util.URL_ORIGIN, comment['author_url']) comment['author_avatar'] = concat_or_none( '/', comment['author_avatar']) comment['permalink'] = concat_or_none(util.URL_ORIGIN, '/watch?v=', comments_info['video_id'], '&lc=', comment['id']) reply_count = comment['reply_count'] if reply_count == 0: comment['replies_url'] = concat_or_none(util.URL_ORIGIN, '/post_comment?parent_id=', comment['id'], '&video_id=', comments_info['video_id']) else: comment['replies_url'] = concat_or_none(util.URL_ORIGIN, '/comments?parent_id=', comment['id'], '&video_id=', comments_info['video_id']) if reply_count == 0: comment['view_replies_text'] = 'Reply' elif reply_count == 1: comment['view_replies_text'] = '1 reply' else: comment['view_replies_text'] = str(reply_count) + ' replies' if comment['like_count'] == 1: comment['likes_text'] = '1 like' else: comment['likes_text'] = str(comment['like_count']) + ' likes' if comments_info['ctoken']: comments_info['more_comments_url'] = concat_or_none(util.URL_ORIGIN, '/comments?ctoken=', comments_info['ctoken']) comments_info['page_number'] = page_number = str(int(comments_info['offset']/20) + 1) if not comments_info['is_replies']: comments_info['sort_text'] = 'top' if comments_info['sort'] == 0 else 'newest' comments_info['video_url'] = concat_or_none(util.URL_ORIGIN, '/watch?v=', comments_info['video_id']) comments_info['video_thumbnail'] = concat_or_none('/i.ytimg.com/vi/', comments_info['video_id'], '/mqdefault.jpg') def video_comments(video_id, sort=0, offset=0, lc='', secret_key=''): comments_info = yt_data_extract.extract_comments_info(request_comments(make_comment_ctoken(video_id, sort, offset, lc, secret_key))) post_process_comments_info(comments_info) post_comment_url = util.URL_ORIGIN + "/post_comment?video_id=" + video_id other_sort_url = util.URL_ORIGIN + '/comments?ctoken=' + make_comment_ctoken(video_id, sort=1 - sort, lc=lc) other_sort_text = 'Sort by ' + ('newest' if sort == 0 else 'top') comments_info['comment_links'] = [('Post comment', post_comment_url), (other_sort_text, other_sort_url)] return comments_info return {}