[youtube:comments] Add more options for limiting number of comments extracted (#1626)

Extends `max_comments` extractor arg to support `max-parents,max-replies,max-replies-per-thread`.
Authored-by: coletdjnz
pull/2008/head
coletdjnz 3 years ago committed by GitHub
parent 0bb322b9c0
commit 46383212b3
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23

@ -1565,8 +1565,10 @@ The following extractors use this feature:
* `player_skip`: Skip some network requests that are generally needed for robust extraction. One or more of `configs` (skip client configs), `webpage` (skip initial webpage), `js` (skip js player). While these options can help reduce the number of requests needed or avoid some rate-limiting, they could cause some issues. See [#860](https://github.com/yt-dlp/yt-dlp/pull/860) for more details * `player_skip`: Skip some network requests that are generally needed for robust extraction. One or more of `configs` (skip client configs), `webpage` (skip initial webpage), `js` (skip js player). While these options can help reduce the number of requests needed or avoid some rate-limiting, they could cause some issues. See [#860](https://github.com/yt-dlp/yt-dlp/pull/860) for more details
* `include_live_dash`: Include live dash formats (These formats don't download properly) * `include_live_dash`: Include live dash formats (These formats don't download properly)
* `comment_sort`: `top` or `new` (default) - choose comment sorting mode (on YouTube's side) * `comment_sort`: `top` or `new` (default) - choose comment sorting mode (on YouTube's side)
* `max_comments`: Maximum amount of comments to download (default all) * `max_comments`: Limit the amount of comments to gather. Comma-seperated list of integers representing `max-comments,max-parents,max-replies,max-replies-per-thread`. Default is `all,all,all,all`.
* `max_comment_depth`: Maximum depth for nested comments. YouTube supports depths 1 or 2 (default) * E.g. `all,all,1000,10` will get a maximum of 1000 replies total, with up to 10 replies per thread. `1000,all,100` will get a maximum of 1000 comments, with a maximum of 100 replies total.
* `max_comment_depth` Maximum depth for nested comments. YouTube supports depths 1 or 2 (default)
* **Deprecated**: Set `max-replies` to `0` or `all` in `max_comments` instead (e.g. `max_comments=all,all,0` to get no replies)
#### youtubetab (YouTube playlists, channels, feeds, etc.) #### youtubetab (YouTube playlists, channels, feeds, etc.)
* `skip`: One or more of `webpage` (skip initial webpage download), `authcheck` (allow the download of playlists requiring authentication when no initial webpage is downloaded. This may cause unwanted behavior, see [#1122](https://github.com/yt-dlp/yt-dlp/pull/1122) for more details) * `skip`: One or more of `webpage` (skip initial webpage download), `authcheck` (allow the download of playlists requiring authentication when no initial webpage is downloaded. This may cause unwanted behavior, see [#1122](https://github.com/yt-dlp/yt-dlp/pull/1122) for more details)

@ -2,7 +2,6 @@
from __future__ import unicode_literals from __future__ import unicode_literals
import base64
import calendar import calendar
import copy import copy
import datetime import datetime
@ -13,6 +12,7 @@ import math
import os.path import os.path
import random import random
import re import re
import sys
import time import time
import traceback import traceback
@ -30,7 +30,6 @@ from ..compat import (
from ..jsinterp import JSInterpreter from ..jsinterp import JSInterpreter
from ..utils import ( from ..utils import (
bug_reports_message, bug_reports_message,
bytes_to_intlist,
clean_html, clean_html,
datetime_from_str, datetime_from_str,
dict_get, dict_get,
@ -39,7 +38,6 @@ from ..utils import (
float_or_none, float_or_none,
format_field, format_field,
int_or_none, int_or_none,
intlist_to_bytes,
is_html, is_html,
join_nonempty, join_nonempty,
mimetype2ext, mimetype2ext,
@ -2117,20 +2115,21 @@ class YoutubeIE(YoutubeBaseInfoExtractor):
'parent': parent or 'root' 'parent': parent or 'root'
} }
def _comment_entries(self, root_continuation_data, ytcfg, video_id, parent=None, comment_counts=None): def _comment_entries(self, root_continuation_data, ytcfg, video_id, parent=None, tracker=None):
get_single_config_arg = lambda c: self._configuration_arg(c, [''])[0]
def extract_header(contents): def extract_header(contents):
_continuation = None _continuation = None
for content in contents: for content in contents:
comments_header_renderer = try_get(content, lambda x: x['commentsHeaderRenderer']) comments_header_renderer = traverse_obj(content, 'commentsHeaderRenderer')
expected_comment_count = parse_count(self._get_text( expected_comment_count = parse_count(self._get_text(
comments_header_renderer, 'countText', 'commentsCount', max_runs=1)) comments_header_renderer, 'countText', 'commentsCount', max_runs=1))
if expected_comment_count: if expected_comment_count:
comment_counts[1] = expected_comment_count tracker['est_total'] = expected_comment_count
self.to_screen('Downloading ~%d comments' % expected_comment_count) self.to_screen(f'Downloading ~{expected_comment_count} comments')
sort_mode_str = self._configuration_arg('comment_sort', [''])[0] comment_sort_index = int(get_single_config_arg('comment_sort') != 'top') # 1 = new, 0 = top
comment_sort_index = int(sort_mode_str != 'top') # 1 = new, 0 = top
sort_menu_item = try_get( sort_menu_item = try_get(
comments_header_renderer, comments_header_renderer,
@ -2141,76 +2140,84 @@ class YoutubeIE(YoutubeBaseInfoExtractor):
if not _continuation: if not _continuation:
continue continue
sort_text = sort_menu_item.get('title') sort_text = str_or_none(sort_menu_item.get('title'))
if isinstance(sort_text, compat_str): if not sort_text:
sort_text = sort_text.lower()
else:
sort_text = 'top comments' if comment_sort_index == 0 else 'newest first' sort_text = 'top comments' if comment_sort_index == 0 else 'newest first'
self.to_screen('Sorting comments by %s' % sort_text) self.to_screen('Sorting comments by %s' % sort_text.lower())
break break
return _continuation return _continuation
def extract_thread(contents): def extract_thread(contents):
if not parent: if not parent:
comment_counts[2] = 0 tracker['current_page_thread'] = 0
for content in contents: for content in contents:
if not parent and tracker['total_parent_comments'] >= max_parents:
yield
comment_thread_renderer = try_get(content, lambda x: x['commentThreadRenderer']) comment_thread_renderer = try_get(content, lambda x: x['commentThreadRenderer'])
comment_renderer = try_get( comment_renderer = get_first(
comment_thread_renderer, (lambda x: x['comment']['commentRenderer'], dict)) or try_get( (comment_thread_renderer, content), [['commentRenderer', ('comment', 'commentRenderer')]],
content, (lambda x: x['commentRenderer'], dict)) expected_type=dict, default={})
if not comment_renderer:
continue
comment = self._extract_comment(comment_renderer, parent) comment = self._extract_comment(comment_renderer, parent)
if not comment: if not comment:
continue continue
comment_counts[0] += 1
tracker['running_total'] += 1
tracker['total_reply_comments' if parent else 'total_parent_comments'] += 1
yield comment yield comment
# Attempt to get the replies # Attempt to get the replies
comment_replies_renderer = try_get( comment_replies_renderer = try_get(
comment_thread_renderer, lambda x: x['replies']['commentRepliesRenderer'], dict) comment_thread_renderer, lambda x: x['replies']['commentRepliesRenderer'], dict)
if comment_replies_renderer: if comment_replies_renderer:
comment_counts[2] += 1 tracker['current_page_thread'] += 1
comment_entries_iter = self._comment_entries( comment_entries_iter = self._comment_entries(
comment_replies_renderer, ytcfg, video_id, comment_replies_renderer, ytcfg, video_id,
parent=comment.get('id'), comment_counts=comment_counts) parent=comment.get('id'), tracker=tracker)
for reply_comment in itertools.islice(comment_entries_iter, min(max_replies_per_thread, max(0, max_replies - tracker['total_reply_comments']))):
for reply_comment in comment_entries_iter:
yield reply_comment yield reply_comment
# Keeps track of counts across recursive calls
if not tracker:
tracker = dict(
running_total=0,
est_total=0,
current_page_thread=0,
total_parent_comments=0,
total_reply_comments=0)
# TODO: Deprecated
# YouTube comments have a max depth of 2 # YouTube comments have a max depth of 2
max_depth = int_or_none(self._configuration_arg('max_comment_depth', [''])[0]) or float('inf') max_depth = int_or_none(get_single_config_arg('max_comment_depth'))
if max_depth:
self._downloader.deprecation_warning(
'[youtube] max_comment_depth extractor argument is deprecated. Set max replies in the max-comments extractor argument instead.')
if max_depth == 1 and parent: if max_depth == 1 and parent:
return return
if not comment_counts:
# comment so far, est. total comments, current comment thread #
comment_counts = [0, 0, 0]
continuation = self._extract_continuation(root_continuation_data) max_comments, max_parents, max_replies, max_replies_per_thread, *_ = map(
if continuation and len(continuation['continuation']) < 27: lambda p: int_or_none(p, default=sys.maxsize), self._configuration_arg('max_comments', ) + [''] * 4)
self.write_debug('Detected old API continuation token. Generating new API compatible token.')
continuation_token = self._generate_comment_continuation(video_id)
continuation = self._build_api_continuation_query(continuation_token, None)
continuation = self._extract_continuation(root_continuation_data)
message = self._get_text(root_continuation_data, ('contents', ..., 'messageRenderer', 'text'), max_runs=1) message = self._get_text(root_continuation_data, ('contents', ..., 'messageRenderer', 'text'), max_runs=1)
if message and not parent: if message and not parent:
self.report_warning(message, video_id=video_id) self.report_warning(message, video_id=video_id)
visitor_data = None response = None
is_first_continuation = parent is None is_first_continuation = parent is None
for page_num in itertools.count(0): for page_num in itertools.count(0):
if not continuation: if not continuation:
break break
headers = self.generate_api_headers(ytcfg=ytcfg, visitor_data=visitor_data) headers = self.generate_api_headers(ytcfg=ytcfg, visitor_data=self._extract_visitor_data(response))
comment_prog_str = '(%d/%d)' % (comment_counts[0], comment_counts[1]) comment_prog_str = f"({tracker['running_total']}/{tracker['est_total']})"
if page_num == 0: if page_num == 0:
if is_first_continuation: if is_first_continuation:
note_prefix = 'Downloading comment section API JSON' note_prefix = 'Downloading comment section API JSON'
else: else:
note_prefix = ' Downloading comment API JSON reply thread %d %s' % ( note_prefix = ' Downloading comment API JSON reply thread %d %s' % (
comment_counts[2], comment_prog_str) tracker['current_page_thread'], comment_prog_str)
else: else:
note_prefix = '%sDownloading comment%s API JSON page %d %s' % ( note_prefix = '%sDownloading comment%s API JSON page %d %s' % (
' ' if parent else '', ' replies' if parent else '', ' ' if parent else '', ' replies' if parent else '',
@ -2219,83 +2226,32 @@ class YoutubeIE(YoutubeBaseInfoExtractor):
response = self._extract_response( response = self._extract_response(
item_id=None, query=continuation, item_id=None, query=continuation,
ep='next', ytcfg=ytcfg, headers=headers, note=note_prefix, ep='next', ytcfg=ytcfg, headers=headers, note=note_prefix,
check_get_keys=('onResponseReceivedEndpoints', 'continuationContents')) check_get_keys='onResponseReceivedEndpoints')
if not response:
break
visitor_data = try_get(
response,
lambda x: x['responseContext']['webResponseContextExtensionData']['ytConfigData']['visitorData'],
compat_str) or visitor_data
continuation_contents = dict_get(response, ('onResponseReceivedEndpoints', 'continuationContents')) continuation_contents = traverse_obj(
response, 'onResponseReceivedEndpoints', expected_type=list, default=[])
continuation = None continuation = None
if isinstance(continuation_contents, list): for continuation_section in continuation_contents:
for continuation_section in continuation_contents: continuation_items = traverse_obj(
if not isinstance(continuation_section, dict): continuation_section,
continue (('reloadContinuationItemsCommand', 'appendContinuationItemsAction'), 'continuationItems'),
continuation_items = try_get( get_all=False, expected_type=list) or []
continuation_section, if is_first_continuation:
(lambda x: x['reloadContinuationItemsCommand']['continuationItems'], continuation = extract_header(continuation_items)
lambda x: x['appendContinuationItemsAction']['continuationItems']), is_first_continuation = False
list) or []
if is_first_continuation:
continuation = extract_header(continuation_items)
is_first_continuation = False
if continuation:
break
continue
count = 0
for count, entry in enumerate(extract_thread(continuation_items)):
yield entry
continuation = self._extract_continuation({'contents': continuation_items})
if continuation: if continuation:
# Sometimes YouTube provides a continuation without any comments
# In most cases we end up just downloading these with very little comments to come.
if count == 0:
if not parent:
self.report_warning('No comments received - assuming end of comments')
continuation = None
break break
continue
# Deprecated response structure for entry in extract_thread(continuation_items):
elif isinstance(continuation_contents, dict): if not entry:
known_continuation_renderers = ('itemSectionContinuation', 'commentRepliesContinuation') return
for key, continuation_renderer in continuation_contents.items(): yield entry
if key not in known_continuation_renderers: continuation = self._extract_continuation({'contents': continuation_items})
continue if continuation:
if not isinstance(continuation_renderer, dict):
continue
if is_first_continuation:
header_continuation_items = [continuation_renderer.get('header') or {}]
continuation = extract_header(header_continuation_items)
is_first_continuation = False
if continuation:
break
# Sometimes YouTube provides a continuation without any comments
# In most cases we end up just downloading these with very little comments to come.
count = 0
for count, entry in enumerate(extract_thread(continuation_renderer.get('contents') or {})):
yield entry
continuation = self._extract_continuation(continuation_renderer)
if count == 0:
if not parent:
self.report_warning('No comments received - assuming end of comments')
continuation = None
break break
@staticmethod
def _generate_comment_continuation(video_id):
"""
Generates initial comment section continuation token from given video id
"""
b64_vid_id = base64.b64encode(bytes(video_id.encode('utf-8')))
parts = ('Eg0SCw==', b64_vid_id, 'GAYyJyIRIgs=', b64_vid_id, 'MAB4AjAAQhBjb21tZW50cy1zZWN0aW9u')
new_continuation_intlist = list(itertools.chain.from_iterable(
[bytes_to_intlist(base64.b64decode(part)) for part in parts]))
return base64.b64encode(intlist_to_bytes(new_continuation_intlist)).decode('utf-8')
def _get_comments(self, ytcfg, video_id, contents, webpage): def _get_comments(self, ytcfg, video_id, contents, webpage):
"""Entry for comment extraction""" """Entry for comment extraction"""
def _real_comment_extract(contents): def _real_comment_extract(contents):

Loading…
Cancel
Save