2018-02-07 01:37:23 +08:00
|
|
|
# frozen_string_literal: true
|
|
|
|
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
# This class is used to mirror unread and new status back to end users
|
|
|
|
# in JavaScript there is a mirror class that is kept in-sync using MessageBus
|
2013-05-21 14:39:51 +08:00
|
|
|
# the allows end users to always know which topics have unread posts in them
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
# and which topics are new. This is used in various places in the UI, such as
|
|
|
|
# counters, indicators, and messages at the top of topic lists, so the user
|
|
|
|
# knows there is something worth reading at a glance.
|
|
|
|
#
|
|
|
|
# The TopicTrackingState.report data is preloaded in ApplicationController
|
|
|
|
# for the current user under the topicTrackingStates key, and the existing
|
|
|
|
# state is loaded into memory on page load. From there the MessageBus is
|
|
|
|
# used to keep topic state up to date, as well as syncing with topics from
|
|
|
|
# corresponding lists fetched from the server (e.g. the /new, /latest,
|
|
|
|
# /unread topic lists).
|
|
|
|
#
|
|
|
|
# See discourse/app/models/topic-tracking-state.js
|
2013-05-29 16:11:04 +08:00
|
|
|
class TopicTrackingState
|
2013-05-24 18:58:26 +08:00
|
|
|
include ActiveModel::SerializerSupport
|
2021-09-09 09:16:53 +08:00
|
|
|
include TopicTrackingStatePublishable
|
2013-05-24 18:58:26 +08:00
|
|
|
|
2020-04-30 14:48:34 +08:00
|
|
|
UNREAD_MESSAGE_TYPE = "unread"
|
|
|
|
LATEST_MESSAGE_TYPE = "latest"
|
|
|
|
MUTED_MESSAGE_TYPE = "muted"
|
2020-12-10 13:49:05 +08:00
|
|
|
UNMUTED_MESSAGE_TYPE = "unmuted"
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
NEW_TOPIC_MESSAGE_TYPE = "new_topic"
|
|
|
|
RECOVER_MESSAGE_TYPE = "recover"
|
|
|
|
DELETE_MESSAGE_TYPE = "delete"
|
|
|
|
DESTROY_MESSAGE_TYPE = "destroy"
|
|
|
|
READ_MESSAGE_TYPE = "read"
|
|
|
|
DISMISS_NEW_MESSAGE_TYPE = "dismiss_new"
|
2023-07-13 18:05:56 +08:00
|
|
|
DISMISS_NEW_POSTS_MESSAGE_TYPE = "dismiss_new_posts"
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
MAX_TOPICS = 5000
|
2013-05-21 14:39:51 +08:00
|
|
|
|
2023-02-01 07:18:45 +08:00
|
|
|
NEW_MESSAGE_BUS_CHANNEL = "/new"
|
|
|
|
LATEST_MESSAGE_BUS_CHANNEL = "/latest"
|
|
|
|
UNREAD_MESSAGE_BUS_CHANNEL = "/unread"
|
|
|
|
RECOVER_MESSAGE_BUS_CHANNEL = "/recover"
|
|
|
|
DELETE_MESSAGE_BUS_CHANNEL = "/delete"
|
|
|
|
DESTROY_MESSAGE_BUS_CHANNEL = "/destroy"
|
|
|
|
|
2013-05-29 16:11:04 +08:00
|
|
|
def self.publish_new(topic)
|
2018-03-05 16:18:23 +08:00
|
|
|
return unless topic.regular?
|
2013-05-29 16:11:04 +08:00
|
|
|
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
tag_ids, tags = nil
|
|
|
|
tag_ids, tags = topic.tags.pluck(:id, :name).transpose if SiteSetting.tagging_enabled
|
2020-06-12 14:20:07 +08:00
|
|
|
|
|
|
|
payload = {
|
|
|
|
last_read_post_number: nil,
|
|
|
|
highest_post_number: 1,
|
|
|
|
created_at: topic.created_at,
|
|
|
|
category_id: topic.category_id,
|
|
|
|
archetype: topic.archetype,
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
created_in_new_period: true,
|
2020-06-12 14:20:07 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
if tags
|
|
|
|
payload[:tags] = tags
|
|
|
|
payload[:topic_tag_ids] = tag_ids
|
|
|
|
end
|
|
|
|
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
message = { topic_id: topic.id, message_type: NEW_TOPIC_MESSAGE_TYPE, payload: payload }
|
2013-05-29 16:11:04 +08:00
|
|
|
|
2023-01-11 06:15:52 +08:00
|
|
|
group_ids = secure_category_group_ids(topic)
|
2013-05-29 16:11:04 +08:00
|
|
|
|
2023-02-01 07:18:45 +08:00
|
|
|
MessageBus.publish(NEW_MESSAGE_BUS_CHANNEL, message.as_json, group_ids: group_ids)
|
2021-09-08 11:55:12 +08:00
|
|
|
publish_read(topic.id, 1, topic.user)
|
2013-05-29 16:11:04 +08:00
|
|
|
end
|
|
|
|
|
2022-06-30 08:18:12 +08:00
|
|
|
def self.publish_latest(topic, whisper = false)
|
2018-03-05 16:18:23 +08:00
|
|
|
return unless topic.regular?
|
2014-08-05 11:27:34 +08:00
|
|
|
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
tag_ids, tags = nil
|
|
|
|
tag_ids, tags = topic.tags.pluck(:id, :name).transpose if SiteSetting.tagging_enabled
|
|
|
|
|
2014-08-05 11:27:34 +08:00
|
|
|
message = {
|
|
|
|
topic_id: topic.id,
|
2018-03-05 16:18:23 +08:00
|
|
|
message_type: LATEST_MESSAGE_TYPE,
|
2014-08-05 11:27:34 +08:00
|
|
|
payload: {
|
|
|
|
bumped_at: topic.bumped_at,
|
2016-03-30 08:17:52 +08:00
|
|
|
category_id: topic.category_id,
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
archetype: topic.archetype,
|
2014-08-05 11:27:34 +08:00
|
|
|
},
|
|
|
|
}
|
|
|
|
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
if tags
|
|
|
|
message[:payload][:tags] = tags
|
|
|
|
message[:payload][:topic_tag_ids] = tag_ids
|
|
|
|
end
|
|
|
|
|
2016-12-02 14:03:31 +08:00
|
|
|
group_ids =
|
2022-06-30 08:18:12 +08:00
|
|
|
if whisper
|
|
|
|
[Group::AUTO_GROUPS[:staff], *SiteSetting.whispers_allowed_group_ids]
|
2016-12-02 14:03:31 +08:00
|
|
|
else
|
2023-01-11 06:15:52 +08:00
|
|
|
secure_category_group_ids(topic)
|
2016-12-02 14:03:31 +08:00
|
|
|
end
|
2023-02-01 07:18:45 +08:00
|
|
|
MessageBus.publish(LATEST_MESSAGE_BUS_CHANNEL, message.as_json, group_ids: group_ids)
|
2014-08-05 11:27:34 +08:00
|
|
|
end
|
|
|
|
|
2018-03-05 16:18:23 +08:00
|
|
|
def self.unread_channel_key(user_id)
|
|
|
|
"/unread/#{user_id}"
|
|
|
|
end
|
|
|
|
|
2020-05-01 06:33:57 +08:00
|
|
|
def self.publish_muted(topic)
|
2023-01-12 08:22:28 +08:00
|
|
|
return unless topic.regular?
|
|
|
|
|
2020-05-01 06:33:57 +08:00
|
|
|
user_ids =
|
|
|
|
topic
|
|
|
|
.topic_users
|
2020-04-23 12:57:35 +08:00
|
|
|
.where(notification_level: NotificationLevels.all[:muted])
|
|
|
|
.joins(:user)
|
|
|
|
.where("users.last_seen_at > ?", 7.days.ago)
|
|
|
|
.order("users.last_seen_at DESC")
|
|
|
|
.limit(100)
|
|
|
|
.pluck(:user_id)
|
|
|
|
return if user_ids.blank?
|
2023-02-01 07:18:45 +08:00
|
|
|
|
2020-04-23 12:57:35 +08:00
|
|
|
message = { topic_id: topic.id, message_type: MUTED_MESSAGE_TYPE }
|
2023-02-01 07:18:45 +08:00
|
|
|
|
|
|
|
MessageBus.publish(LATEST_MESSAGE_BUS_CHANNEL, message.as_json, user_ids: user_ids)
|
2020-04-23 12:57:35 +08:00
|
|
|
end
|
|
|
|
|
2020-12-10 13:49:05 +08:00
|
|
|
def self.publish_unmuted(topic)
|
2023-01-12 08:22:28 +08:00
|
|
|
return unless topic.regular?
|
|
|
|
|
2022-10-06 08:10:43 +08:00
|
|
|
user_ids =
|
|
|
|
User
|
|
|
|
.watching_topic(topic)
|
2020-12-10 13:49:05 +08:00
|
|
|
.where("users.last_seen_at > ?", 7.days.ago)
|
|
|
|
.order("users.last_seen_at DESC")
|
|
|
|
.limit(100)
|
|
|
|
.pluck(:id)
|
|
|
|
return if user_ids.blank?
|
2023-02-01 07:18:45 +08:00
|
|
|
|
2020-12-10 13:49:05 +08:00
|
|
|
message = { topic_id: topic.id, message_type: UNMUTED_MESSAGE_TYPE }
|
2023-02-01 07:18:45 +08:00
|
|
|
|
|
|
|
MessageBus.publish(LATEST_MESSAGE_BUS_CHANNEL, message.as_json, user_ids: user_ids)
|
2020-12-10 13:49:05 +08:00
|
|
|
end
|
|
|
|
|
2013-05-29 16:11:04 +08:00
|
|
|
def self.publish_unread(post)
|
2018-03-05 16:18:23 +08:00
|
|
|
return unless post.topic.regular?
|
2013-05-29 16:11:04 +08:00
|
|
|
# TODO at high scale we are going to have to defer this,
|
|
|
|
# perhaps cut down to users that are around in the last 7 days as well
|
2021-08-11 11:01:13 +08:00
|
|
|
tags = nil
|
|
|
|
tag_ids = nil
|
|
|
|
tag_ids, tags = post.topic.tags.pluck(:id, :name).transpose if include_tags_in_report?
|
|
|
|
|
FIX: Issues with incorrect unread and private message topic tracking state (#16474)
This commit fixes two issues at play. The first was introduced
in f6c852b (or maybe not introduced
but rather revealed). When a user posted a new message in a topic,
they received the unread topic tracking state MessageBus message,
and the Unread (X) indicator was incremented by one, because with the
aforementioned perf commit we "guess" the correct last read post
for the user, because we no longer calculate individual users' read
status there. This meant that every time a user posted in a topic
they tracked, the unread indicator was incremented. To get around
this, we can just exclude the user who created the post from the
target users of the unread state message.
The second issue was related to the private message topic tracking
state, and was somewhat similar. Whenever a user created a new private
message, the New (X) indicator was incremented, and could not be
cleared until the page was refreshed. To solve this, we just don't
update the topic state for the user when the new_topic tracking state
message comes through if the user who created the topic is the
same as the current user.
cf. https://meta.discourse.org/t/bottom-of-topic-shows-there-is-1-unread-remaining-when-there-are-actually-0-unread-topics-remaining/220817
2022-04-19 09:37:01 +08:00
|
|
|
# We don't need to publish unread to the person who just made the post,
|
|
|
|
# this is why they are excluded from the initial scope.
|
2021-08-11 11:01:13 +08:00
|
|
|
scope =
|
FIX: Issues with incorrect unread and private message topic tracking state (#16474)
This commit fixes two issues at play. The first was introduced
in f6c852b (or maybe not introduced
but rather revealed). When a user posted a new message in a topic,
they received the unread topic tracking state MessageBus message,
and the Unread (X) indicator was incremented by one, because with the
aforementioned perf commit we "guess" the correct last read post
for the user, because we no longer calculate individual users' read
status there. This meant that every time a user posted in a topic
they tracked, the unread indicator was incremented. To get around
this, we can just exclude the user who created the post from the
target users of the unread state message.
The second issue was related to the private message topic tracking
state, and was somewhat similar. Whenever a user created a new private
message, the New (X) indicator was incremented, and could not be
cleared until the page was refreshed. To solve this, we just don't
update the topic state for the user when the new_topic tracking state
message comes through if the user who created the topic is the
same as the current user.
cf. https://meta.discourse.org/t/bottom-of-topic-shows-there-is-1-unread-remaining-when-there-are-actually-0-unread-topics-remaining/220817
2022-04-19 09:37:01 +08:00
|
|
|
TopicUser.tracking(post.topic_id).includes(user: :user_stat).where.not(user_id: post.user_id)
|
2016-12-02 14:03:31 +08:00
|
|
|
|
|
|
|
group_ids =
|
|
|
|
if post.post_type == Post.types[:whisper]
|
2022-06-30 08:18:12 +08:00
|
|
|
[Group::AUTO_GROUPS[:staff], *SiteSetting.whispers_allowed_group_ids]
|
2016-12-02 14:03:31 +08:00
|
|
|
else
|
|
|
|
post.topic.category && post.topic.category.secure_group_ids
|
|
|
|
end
|
2013-05-29 16:11:04 +08:00
|
|
|
|
2021-08-11 11:01:13 +08:00
|
|
|
if group_ids.present?
|
|
|
|
scope =
|
|
|
|
scope.joins("INNER JOIN group_users gu ON gu.user_id = topic_users.user_id").where(
|
|
|
|
"gu.group_id IN (?)",
|
|
|
|
group_ids,
|
|
|
|
)
|
2020-06-12 14:20:07 +08:00
|
|
|
end
|
|
|
|
|
2022-02-22 23:27:46 +08:00
|
|
|
user_ids = scope.pluck(:user_id)
|
|
|
|
return if user_ids.empty?
|
2020-06-12 14:20:07 +08:00
|
|
|
|
2022-02-22 23:27:46 +08:00
|
|
|
payload = {
|
|
|
|
highest_post_number: post.post_number,
|
|
|
|
updated_at: post.topic.updated_at,
|
|
|
|
created_at: post.created_at,
|
|
|
|
category_id: post.topic.category_id,
|
|
|
|
archetype: post.topic.archetype,
|
|
|
|
}
|
2013-05-29 16:11:04 +08:00
|
|
|
|
2022-02-22 23:27:46 +08:00
|
|
|
if tags
|
|
|
|
payload[:tags] = tags
|
|
|
|
payload[:topic_tag_ids] = tag_ids
|
2013-05-21 14:39:51 +08:00
|
|
|
end
|
2013-11-25 14:37:51 +08:00
|
|
|
|
2022-02-22 23:27:46 +08:00
|
|
|
message = { topic_id: post.topic_id, message_type: UNREAD_MESSAGE_TYPE, payload: payload }
|
|
|
|
|
2023-02-01 07:18:45 +08:00
|
|
|
MessageBus.publish(UNREAD_MESSAGE_BUS_CHANNEL, message.as_json, user_ids: user_ids)
|
2013-05-21 14:39:51 +08:00
|
|
|
end
|
|
|
|
|
2016-03-30 08:17:52 +08:00
|
|
|
def self.publish_recover(topic)
|
2023-01-12 08:22:28 +08:00
|
|
|
return unless topic.regular?
|
|
|
|
|
2023-01-11 06:15:52 +08:00
|
|
|
group_ids = secure_category_group_ids(topic)
|
2016-03-30 08:17:52 +08:00
|
|
|
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
message = { topic_id: topic.id, message_type: RECOVER_MESSAGE_TYPE }
|
2016-03-30 08:17:52 +08:00
|
|
|
|
2023-02-01 07:18:45 +08:00
|
|
|
MessageBus.publish(RECOVER_MESSAGE_BUS_CHANNEL, message.as_json, group_ids: group_ids)
|
2016-03-30 08:17:52 +08:00
|
|
|
end
|
|
|
|
|
|
|
|
def self.publish_delete(topic)
|
2023-01-12 08:22:28 +08:00
|
|
|
return unless topic.regular?
|
|
|
|
|
2023-01-11 06:15:52 +08:00
|
|
|
group_ids = secure_category_group_ids(topic)
|
2016-03-30 08:17:52 +08:00
|
|
|
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
message = { topic_id: topic.id, message_type: DELETE_MESSAGE_TYPE }
|
2016-03-30 08:17:52 +08:00
|
|
|
|
|
|
|
MessageBus.publish("/delete", message.as_json, group_ids: group_ids)
|
|
|
|
end
|
|
|
|
|
2020-11-10 12:40:48 +08:00
|
|
|
def self.publish_destroy(topic)
|
2023-01-12 08:22:28 +08:00
|
|
|
return unless topic.regular?
|
|
|
|
|
2023-01-11 06:15:52 +08:00
|
|
|
group_ids = secure_category_group_ids(topic)
|
2020-11-10 12:40:48 +08:00
|
|
|
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
message = { topic_id: topic.id, message_type: DESTROY_MESSAGE_TYPE }
|
2020-11-10 12:40:48 +08:00
|
|
|
|
2023-02-01 07:18:45 +08:00
|
|
|
MessageBus.publish(DESTROY_MESSAGE_BUS_CHANNEL, message.as_json, group_ids: group_ids)
|
2020-11-10 12:40:48 +08:00
|
|
|
end
|
|
|
|
|
2021-09-08 11:55:12 +08:00
|
|
|
def self.publish_read(topic_id, last_read_post_number, user, notification_level = nil)
|
2021-09-09 09:16:53 +08:00
|
|
|
self.publish_read_message(
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
message_type: READ_MESSAGE_TYPE,
|
2021-09-09 09:16:53 +08:00
|
|
|
channel_name: self.unread_channel_key(user.id),
|
|
|
|
topic_id: topic_id,
|
|
|
|
user: user,
|
|
|
|
last_read_post_number: last_read_post_number,
|
|
|
|
notification_level: notification_level,
|
|
|
|
)
|
2013-05-21 14:39:51 +08:00
|
|
|
end
|
|
|
|
|
2021-02-15 05:50:33 +08:00
|
|
|
def self.publish_dismiss_new(user_id, topic_ids: [])
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
message = { message_type: DISMISS_NEW_MESSAGE_TYPE, payload: { topic_ids: topic_ids } }
|
2019-11-25 03:17:31 +08:00
|
|
|
MessageBus.publish(self.unread_channel_key(user_id), message.as_json, user_ids: [user_id])
|
|
|
|
end
|
|
|
|
|
2023-07-13 18:05:56 +08:00
|
|
|
def self.publish_dismiss_new_posts(user_id, topic_ids: [])
|
|
|
|
message = { message_type: DISMISS_NEW_POSTS_MESSAGE_TYPE, payload: { topic_ids: topic_ids } }
|
|
|
|
MessageBus.publish(self.unread_channel_key(user_id), message.as_json, user_ids: [user_id])
|
|
|
|
end
|
|
|
|
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
def self.new_filter_sql
|
|
|
|
TopicQuery
|
|
|
|
.new_filter(Topic, treat_as_new_topic_clause_sql: treat_as_new_topic_clause)
|
|
|
|
.where_clause
|
|
|
|
.ast
|
|
|
|
.to_sql + " AND topics.created_at > :min_new_topic_date" +
|
|
|
|
" AND dismissed_topic_users.id IS NULL"
|
|
|
|
end
|
|
|
|
|
2022-06-30 08:18:12 +08:00
|
|
|
def self.unread_filter_sql(whisperer: false)
|
|
|
|
TopicQuery.unread_filter(Topic, whisperer: whisperer).where_clause.ast.to_sql
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
end
|
|
|
|
|
2013-05-23 13:21:07 +08:00
|
|
|
def self.treat_as_new_topic_clause
|
2014-03-04 05:11:59 +08:00
|
|
|
User
|
|
|
|
.where(
|
|
|
|
"GREATEST(CASE
|
2016-02-18 13:57:22 +08:00
|
|
|
WHEN COALESCE(uo.new_topic_duration_minutes, :default_duration) = :always THEN u.created_at
|
|
|
|
WHEN COALESCE(uo.new_topic_duration_minutes, :default_duration) = :last_visit THEN COALESCE(u.previous_visit_at,u.created_at)
|
|
|
|
ELSE (:now::timestamp - INTERVAL '1 MINUTE' * COALESCE(uo.new_topic_duration_minutes, :default_duration))
|
2021-02-15 05:50:33 +08:00
|
|
|
END, u.created_at, :min_date)",
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
treat_as_new_topic_params,
|
2020-12-22 07:38:59 +08:00
|
|
|
)
|
|
|
|
.where_clause
|
|
|
|
.ast
|
|
|
|
.to_sql
|
2013-05-21 14:39:51 +08:00
|
|
|
end
|
|
|
|
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
def self.treat_as_new_topic_params
|
|
|
|
{
|
|
|
|
now: DateTime.now,
|
|
|
|
last_visit: User::NewTopicDuration::LAST_VISIT,
|
|
|
|
always: User::NewTopicDuration::ALWAYS,
|
|
|
|
default_duration: SiteSetting.default_other_new_topic_duration_minutes,
|
|
|
|
min_date: Time.at(SiteSetting.min_new_topics_time).to_datetime,
|
|
|
|
}
|
|
|
|
end
|
|
|
|
|
2020-05-29 10:57:46 +08:00
|
|
|
def self.include_tags_in_report?
|
2022-12-08 09:44:29 +08:00
|
|
|
SiteSetting.tagging_enabled && (@include_tags_in_report || !SiteSetting.legacy_navigation_menu?)
|
2020-05-29 10:57:46 +08:00
|
|
|
end
|
|
|
|
|
|
|
|
def self.include_tags_in_report=(v)
|
|
|
|
@include_tags_in_report = v
|
|
|
|
end
|
|
|
|
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
# Sam: this is a hairy report, in particular I need custom joins and fancy conditions
|
|
|
|
# Dropping to sql_builder so I can make sense of it.
|
|
|
|
#
|
|
|
|
# Keep in mind, we need to be able to filter on a GROUP of users, and zero in on topic
|
|
|
|
# all our existing scope work does not do this
|
|
|
|
#
|
|
|
|
# This code needs to be VERY efficient as it is triggered via the message bus and may steal
|
|
|
|
# cycles from usual requests
|
2016-12-02 14:03:31 +08:00
|
|
|
def self.report(user, topic_id = nil)
|
2020-05-13 11:09:40 +08:00
|
|
|
tag_ids = muted_tag_ids(user)
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
sql = new_and_unread_sql(topic_id, user, tag_ids)
|
|
|
|
sql = tags_included_wrapped_sql(sql)
|
|
|
|
|
|
|
|
report =
|
|
|
|
DB.query(
|
|
|
|
sql + "\n\n LIMIT :max_topics",
|
|
|
|
{
|
|
|
|
user_id: user.id,
|
|
|
|
topic_id: topic_id,
|
|
|
|
min_new_topic_date: Time.at(SiteSetting.min_new_topics_time).to_datetime,
|
|
|
|
max_topics: TopicTrackingState::MAX_TOPICS,
|
PERF: Optimise `TopicTrackingState.report` query to speed up query (#22871)
In the query generated by `TopicTrackingState.report`, there are two
subqueies being executed. The first subquery fetches all the topics
that are new for a given user while the second subquery fetches all the topics with
unread posts for a given user. For the second subquery, there is a
filter `topics.updated_at >= user_stats.first_unread_at` which is used
as a performance optimisation to reduce the number of rows that PG has
to scan through the `topics` table.
However, we started to notice in production that the PG planner doesn't
always execute the filter first to reduce the number of rows that it has
to scan through. Running the following query in one of our production
instance,
```
EXPLAIN ANALYZE
SELECT
DISTINCT topics.id as topic_id,
u.id as user_id,
topics.created_at,
topics.updated_at,
topics.highest_staff_post_number AS highest_post_number,
last_read_post_number,
c.id as category_id,
c.topic_id AS category_topic_id,
tu.notification_level,
us.first_unread_at,
GREATEST(
CASE
WHEN COALESCE(uo.new_topic_duration_minutes, 2880) = -1 THEN u.created_at
WHEN COALESCE(uo.new_topic_duration_minutes, 2880) = -2 THEN COALESCE(
u.previous_visit_at,u.created_at
)
ELSE ('2023-07-31 03:29:45.737630'::timestamp - INTERVAL '1 MINUTE' * COALESCE(uo.new_topic_duration_minutes, 2880))
END, u.created_at, '2023-07-25 15:06:44'
) AS treat_as_new_topic_start_date
FROM topics
JOIN users u on u.id = 13455
JOIN user_stats AS us ON us.user_id = u.id
JOIN user_options AS uo ON uo.user_id = u.id
JOIN categories c ON c.id = topics.category_id
LEFT JOIN topic_users tu ON tu.topic_id = topics.id AND tu.user_id = u.id
WHERE u.id = 13455 AND
topics.updated_at >= us.first_unread_at AND
topics.archetype <> 'private_message' AND
(("topics"."deleted_at" IS NULL AND (tu.last_read_post_number < topics.highest_staff_post_number) AND (COALESCE(tu.notification_level, 1) >= 2)) OR (1=0)) AND
NOT (
COALESCE((select array_agg(tag_id) from topic_tags where topic_tags.topic_id = topics.id), ARRAY[]::int[]) && ARRAY[451,452,453]
) AND
topics.deleted_at IS NULL AND
NOT (
last_read_post_number IS NULL AND
(
topics.category_id IN (SELECT "categories"."id" FROM "categories" LEFT JOIN categories categories2 ON categories2.id = categories.parent_category_id LEFT JOIN category_users ON category_users.category_id = categories.id AND category_users.user_id = 13455 LEFT JOIN category_users category_users2 ON category_users2.category_id = categories2.id AND category_users2.user_id = 13455 WHERE ((category_users.id IS NULL AND COALESCE(category_users2.notification_level, 1) = 0) OR COALESCE(category_users.notification_level, 1) = 0))
AND tu.notification_level <= 1
)
)
```
we get the following
```
QUERY PLAN
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Unique (cost=201606.06..201608.15 rows=76 width=60) (actual time=91.279..91.294 rows=14 loops=1)
-> Sort (cost=201606.06..201606.25 rows=76 width=60) (actual time=91.278..91.284 rows=14 loops=1)
Sort Key: topics.id, topics.created_at, topics.updated_at, topics.highest_staff_post_number, tu.last_read_post_number, c.id, c.topic_id, tu.notification_level, us.first_unread_at, (GREATEST(CASE WHEN (COALESCE(uo.new_topic_duration_minutes, 2880) = '-1'::integer) THEN u.created_at WHEN (COALESCE(uo.new_topic_duration_minutes, 2880) = '-2'::integer) THEN COALESCE(u.previous_visit_at, u.created_at) ELSE ('2023-07-31 03:29:45.73763'::timestamp without time zone - ('00:01:00'::interval * (COALESCE(uo.new_topic_duration_minutes, 2880))::double precision)) END, u.created_at, '2023-07-25 15:06:44'::timestamp without time zone))
Sort Method: quicksort Memory: 26kB
-> Hash Join (cost=97519.51..201603.69 rows=76 width=60) (actual time=87.662..91.268 rows=14 loops=1)
Hash Cond: (topics.id = tu.topic_id)
Join Filter: ((tu.last_read_post_number < topics.highest_staff_post_number) AND ((tu.last_read_post_number IS NOT NULL) OR (NOT (hashed SubPlan 2)) OR (tu.notification_level > 1)))
Rows Removed by Join Filter: 10
-> Nested Loop (cost=1.54..104075.36 rows=3511 width=68) (actual time=0.055..3.609 rows=548 loops=1)
-> Nested Loop (cost=1.13..25.20 rows=1 width=32) (actual time=0.027..0.033 rows=1 loops=1)
-> Nested Loop (cost=0.71..16.76 rows=1 width=28) (actual time=0.020..0.023 rows=1 loops=1)
-> Index Scan using users_pkey on users u (cost=0.42..8.44 rows=1 width=20) (actual time=0.010..0.012 rows=1 loops=1)
Index Cond: (id = 13455)
-> Index Scan using user_stats_pkey on user_stats us (cost=0.29..8.31 rows=1 width=12) (actual time=0.008..0.010 rows=1 loops=1)
Index Cond: (user_id = 13455)
-> Index Scan using index_user_options_on_user_id_and_default_calendar on user_options uo (cost=0.42..8.44 rows=1 width=8) (actual time=0.007..0.008 rows=1 loops=1)
Index Cond: (user_id = 13455)
-> Nested Loop (cost=0.41..104015.12 rows=3504 width=36) (actual time=0.026..3.503 rows=548 loops=1)
-> Seq Scan on categories c (cost=0.00..13.73 rows=73 width=8) (actual time=0.003..0.039 rows=73 loops=1)
-> Index Only Scan using index_topics_on_updated_at_public on topics (cost=0.41..1424.20 rows=48 width=28) (actual time=0.012..0.046 rows=8 loops=73)
Index Cond: ((updated_at >= us.first_unread_at) AND (category_id = c.id))
Filter: (NOT (COALESCE((SubPlan 1), '{}'::integer[]) && '{451,452,453}'::integer[]))
Heap Fetches: 553
SubPlan 1
-> Aggregate (cost=4.31..4.32 rows=1 width=32) (actual time=0.002..0.002 rows=1 loops=548)
-> Index Only Scan using index_topic_tags_on_topic_id_and_tag_id on topic_tags (cost=0.29..4.31 rows=1 width=4) (actual time=0.002..0.002 rows=1 loops=548)
Index Cond: (topic_id = topics.id)
Heap Fetches: 178
-> Hash (cost=97222.14..97222.14 rows=19914 width=16) (actual time=87.545..87.546 rows=42884 loops=1)
Buckets: 65536 (originally 32768) Batches: 1 (originally 1) Memory Usage: 2387kB
-> Bitmap Heap Scan on topic_users tu (cost=1217.47..97222.14 rows=19914 width=16) (actual time=14.419..78.286 rows=42884 loops=1)
Recheck Cond: (user_id = 13455)
Filter: (COALESCE(notification_level, 1) >= 2)
Rows Removed by Filter: 15839
Heap Blocks: exact=45285
-> Bitmap Index Scan on index_topic_users_on_user_id_and_topic_id (cost=0.00..1212.49 rows=59741 width=0) (actual time=6.448..6.448 rows=58723 loops=1)
Index Cond: (user_id = 13455)
SubPlan 2
-> Nested Loop Left Join (cost=0.74..46.90 rows=1 width=4) (never executed)
Join Filter: (category_users2.category_id = categories2.id)
Filter: (((category_users.id IS NULL) AND (COALESCE(category_users2.notification_level, 1) = 0)) OR (COALESCE(category_users.notification_level, 1) = 0))
-> Nested Loop Left Join (cost=0.45..32.31 rows=73 width=16) (never executed)
Join Filter: (category_users.category_id = categories.id)
-> Nested Loop Left Join (cost=0.15..18.45 rows=73 width=8) (never executed)
-> Seq Scan on categories (cost=0.00..13.73 rows=73 width=8) (never executed)
-> Memoize (cost=0.15..0.28 rows=1 width=4) (never executed)
Cache Key: categories.parent_category_id
Cache Mode: logical
-> Index Only Scan using categories_pkey on categories categories2 (cost=0.14..0.27 rows=1 width=4) (never executed)
Index Cond: (id = categories.parent_category_id)
Heap Fetches: 0
-> Materialize (cost=0.29..11.69 rows=2 width=12) (never executed)
-> Index Scan using idx_category_users_user_id_category_id on category_users (cost=0.29..11.68 rows=2 width=12) (never executed)
Index Cond: (user_id = 13455)
-> Materialize (cost=0.29..11.69 rows=2 width=8) (never executed)
-> Index Scan using idx_category_users_user_id_category_id on category_users category_users2 (cost=0.29..11.68 rows=2 width=8) (never executed)
Index Cond: (user_id = 13455)
Planning Time: 1.740 ms
Execution Time: 91.414 ms
(59 rows)
```
From the execution plan, we can see the most of the time is spent
joining about 42888 rows in the `topics` table to the `topic_users` table.
However, we know that we only have to scan through a
subset of the `topics` table because the user's last unread at is '2023-07-20 11:33:05'.
If we filter the `topics` table with `topics.updated_at >= '2023-07-20 11:33:05'`, this would only
return about 1500 rows.
From our testing in production, the PG planner is able to execute a
better query plan when we avoid the unnecessary joins on `user_stats` just to be
able to get the user's `UserStat#first_unread_at`. Instead, we can just
pass the value of `UserStat#first_unread_at` directly as a query
parameter.
```
EXPLAIN ANALYZE
SELECT
DISTINCT topics.id as topic_id,
u.id as user_id,
topics.created_at,
topics.updated_at,
topics.highest_staff_post_number AS highest_post_number,
last_read_post_number,
c.id as category_id,
c.topic_id AS category_topic_id,
tu.notification_level,
GREATEST(
CASE
WHEN COALESCE(uo.new_topic_duration_minutes, 2880) = -1 THEN u.created_at
WHEN COALESCE(uo.new_topic_duration_minutes, 2880) = -2 THEN COALESCE(
u.previous_visit_at,u.created_at
)
ELSE ('2023-07-31 03:29:45.737630'::timestamp - INTERVAL '1 MINUTE' * COALESCE(uo.new_topic_duration_minutes, 2880))
END, u.created_at, '2023-07-25 15:06:44'
) AS treat_as_new_topic_start_date
FROM topics
JOIN users u on u.id = 13455
JOIN user_options AS uo ON uo.user_id = u.id
JOIN categories c ON c.id = topics.category_id
LEFT JOIN topic_users tu ON tu.topic_id = topics.id AND tu.user_id = u.id
WHERE u.id = 13455 AND
topics.updated_at >= '2023-07-20 11:33:05' AND
topics.archetype <> 'private_message' AND
(("topics"."deleted_at" IS NULL AND (tu.last_read_post_number < topics.highest_staff_post_number) AND (COALESCE(tu.notification_level, 1) >= 2)) OR (1=0)) AND
NOT (
COALESCE((select array_agg(tag_id) from topic_tags where topic_tags.topic_id = topics.id), ARRAY[]::int[]) && ARRAY[451,452,453]
) AND
topics.deleted_at IS NULL AND
NOT (
last_read_post_number IS NULL AND
(
topics.category_id IN (SELECT "categories"."id" FROM "categories" LEFT JOIN categories categories2 ON categories2.id = categories.parent_category_id LEFT JOIN category_users ON category_users.category_id = categories.id AND category_users.user_id = 13455 LEFT JOIN category_users category_users2 ON category_users2.category_id = categories2.id AND category_users2.user_id = 13455 WHERE ((category_users.id IS NULL AND COALESCE(category_users2.notification_level, 1) = 0) OR COALESCE(category_users.notification_level, 1) = 0))
AND tu.notification_level <= 1
)
);
```
Note how the filter is now `topics.updated_at >= '2023-07-20 11:33:05'`
instead of `topics.updated_at >= us.first_unread_at`. The modified query
above generates the following execution plan.
```
QUERY PLAN
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Unique (cost=5189.86..5189.88 rows=1 width=52) (actual time=4.991..5.002 rows=14 loops=1)
-> Sort (cost=5189.86..5189.86 rows=1 width=52) (actual time=4.990..4.994 rows=14 loops=1)
Sort Key: topics.id, topics.created_at, topics.updated_at, topics.highest_staff_post_number, tu.last_read_post_number, c.id, c.topic_id, tu.notification_level, (GREATEST(CASE WHEN (COALESCE(uo.new_topic_duration_minutes, 2880) = '-1'::integer) THEN u.created_at WHEN (COALESCE(uo.new_topic_duration_minutes, 2880) = '-2'::integer) THEN COALESCE(u.previous_visit_at, u.created_at) ELSE ('2023-07-31 03:29:45.73763'::timestamp without time zone - ('00:01:00'::interval * (COALESCE(uo.new_topic_duration_minutes, 2880))::double precision)) END, u.created_at, '2023-07-25 15:06:44'::timestamp without time zone))
Sort Method: quicksort Memory: 26kB
-> Nested Loop (cost=52.11..5189.85 rows=1 width=52) (actual time=0.093..4.974 rows=14 loops=1)
-> Nested Loop (cost=51.70..5181.39 rows=1 width=60) (actual time=0.084..4.931 rows=14 loops=1)
-> Nested Loop (cost=51.28..5172.94 rows=1 width=44) (actual time=0.076..4.887 rows=14 loops=1)
-> Nested Loop (cost=0.41..1698.46 rows=59 width=36) (actual time=0.029..3.537 rows=548 loops=1)
-> Seq Scan on categories c (cost=0.00..13.73 rows=73 width=8) (actual time=0.005..0.039 rows=73 loops=1)
-> Index Only Scan using index_topics_on_updated_at_public on topics (cost=0.41..23.07 rows=1 width=28) (actual time=0.012..0.047 rows=8 loops=73)
Index Cond: ((updated_at >= '2023-07-20 11:33:05'::timestamp without time zone) AND (category_id = c.id))
Filter: (NOT (COALESCE((SubPlan 1), '{}'::integer[]) && '{451,452,453}'::integer[]))
Heap Fetches: 552
SubPlan 1
-> Aggregate (cost=4.31..4.32 rows=1 width=32) (actual time=0.002..0.002 rows=1 loops=548)
-> Index Only Scan using index_topic_tags_on_topic_id_and_tag_id on topic_tags (cost=0.29..4.31 rows=1 width=4) (actual time=0.002..0.002 rows=1 loops=548)
Index Cond: (topic_id = topics.id)
Heap Fetches: 178
-> Index Scan using index_topic_users_on_user_id_and_topic_id on topic_users tu (cost=50.86..58.88 rows=1 width=16) (actual time=0.002..0.002 rows=0 loops=548)
Index Cond: ((user_id = 13455) AND (topic_id = topics.id))
Filter: ((COALESCE(notification_level, 1) >= 2) AND (last_read_post_number < topics.highest_staff_post_number) AND ((last_read_post_number IS NOT NULL) OR (NOT (hashed SubPlan 2)) OR (notification_level > 1)))
Rows Removed by Filter: 0
SubPlan 2
-> Nested Loop Left Join (cost=0.74..50.43 rows=1 width=4) (never executed)
Join Filter: (category_users2.category_id = categories2.id)
Filter: (((category_users.id IS NULL) AND (COALESCE(category_users2.notification_level, 1) = 0)) OR (COALESCE(category_users.notification_level, 1) = 0))
-> Nested Loop Left Join (cost=0.45..35.84 rows=73 width=16) (never executed)
Join Filter: (category_users.category_id = categories.id)
-> Nested Loop Left Join (cost=0.15..21.97 rows=73 width=8) (never executed)
-> Seq Scan on categories (cost=0.00..13.73 rows=73 width=8) (never executed)
-> Memoize (cost=0.15..0.61 rows=1 width=4) (never executed)
Cache Key: categories.parent_category_id
Cache Mode: logical
-> Index Only Scan using categories_pkey on categories categories2 (cost=0.14..0.60 rows=1 width=4) (never executed)
Index Cond: (id = categories.parent_category_id)
Heap Fetches: 0
-> Materialize (cost=0.29..11.69 rows=2 width=12) (never executed)
-> Index Scan using idx_category_users_user_id_category_id on category_users (cost=0.29..11.68 rows=2 width=12) (never executed)
Index Cond: (user_id = 13455)
-> Materialize (cost=0.29..11.69 rows=2 width=8) (never executed)
-> Index Scan using idx_category_users_user_id_category_id on category_users category_users2 (cost=0.29..11.68 rows=2 width=8) (never executed)
Index Cond: (user_id = 13455)
-> Index Scan using users_pkey on users u (cost=0.42..8.44 rows=1 width=20) (actual time=0.003..0.003 rows=1 loops=14)
Index Cond: (id = 13455)
-> Index Scan using index_user_options_on_user_id_and_default_calendar on user_options uo (cost=0.42..8.44 rows=1 width=8) (actual time=0.002..0.002 rows=1 loops=14)
Index Cond: (user_id = 13455)
Planning Time: 1.281 ms
Execution Time: 5.092 ms
(48 rows)
```
With the new query, PG first does an index scan using the `index_topics_on_updated_at_public` index to filter away most of the topics making the subsequent joins much cheaper. Total query time has been reduced from ~90ms to ~5ms.
This optimisation will mostly affect users with very few/recent unread topics since a large `UserStat#firsts_unread_at` value will still mean scanning through a large portion of the `topics` table.
2023-07-31 12:21:41 +08:00
|
|
|
user_first_unread_at: user.user_stat.first_unread_at,
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
}.merge(treat_as_new_topic_params),
|
|
|
|
)
|
|
|
|
|
|
|
|
report
|
|
|
|
end
|
2020-05-13 11:09:40 +08:00
|
|
|
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
def self.new_and_unread_sql(topic_id, user, tag_ids)
|
|
|
|
sql =
|
|
|
|
report_raw_sql(
|
2019-04-05 12:25:19 +08:00
|
|
|
topic_id: topic_id,
|
|
|
|
skip_unread: true,
|
|
|
|
skip_order: true,
|
|
|
|
staff: user.staff?,
|
2019-11-14 13:11:34 +08:00
|
|
|
admin: user.admin?,
|
2022-06-30 08:18:12 +08:00
|
|
|
whisperer: user.whisperer?,
|
2019-12-10 06:50:05 +08:00
|
|
|
user: user,
|
2020-05-13 11:09:40 +08:00
|
|
|
muted_tag_ids: tag_ids,
|
2019-04-05 12:25:19 +08:00
|
|
|
)
|
|
|
|
|
2015-09-29 09:55:09 +08:00
|
|
|
sql << "\nUNION ALL\n\n"
|
2019-04-05 12:25:19 +08:00
|
|
|
|
|
|
|
sql << report_raw_sql(
|
|
|
|
topic_id: topic_id,
|
|
|
|
skip_new: true,
|
|
|
|
skip_order: true,
|
|
|
|
staff: user.staff?,
|
|
|
|
filter_old_unread: true,
|
2019-11-14 13:11:34 +08:00
|
|
|
admin: user.admin?,
|
2022-06-30 08:18:12 +08:00
|
|
|
whisperer: user.whisperer?,
|
2019-12-10 06:50:05 +08:00
|
|
|
user: user,
|
2020-05-13 11:09:40 +08:00
|
|
|
muted_tag_ids: tag_ids,
|
2019-04-05 12:25:19 +08:00
|
|
|
)
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
end
|
2015-09-07 09:57:50 +08:00
|
|
|
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
def self.tags_included_wrapped_sql(sql)
|
2020-05-29 10:57:46 +08:00
|
|
|
return <<~SQL if SiteSetting.tagging_enabled && TopicTrackingState.include_tags_in_report?
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
WITH tags_included_cte AS (
|
|
|
|
#{sql}
|
|
|
|
)
|
2020-05-29 10:57:46 +08:00
|
|
|
SELECT *, (
|
|
|
|
SELECT ARRAY_AGG(name) from topic_tags
|
|
|
|
JOIN tags on tags.id = topic_tags.tag_id
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
WHERE topic_id = tags_included_cte.topic_id
|
2020-05-29 10:57:46 +08:00
|
|
|
) tags
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
FROM tags_included_cte
|
2020-05-29 10:57:46 +08:00
|
|
|
SQL
|
|
|
|
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
sql
|
2015-09-07 09:57:50 +08:00
|
|
|
end
|
|
|
|
|
2019-12-10 06:50:05 +08:00
|
|
|
def self.muted_tag_ids(user)
|
|
|
|
TagUser.lookup(user, :muted).pluck(:tag_id)
|
|
|
|
end
|
|
|
|
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
def self.report_raw_sql(
|
|
|
|
user:,
|
|
|
|
muted_tag_ids:,
|
|
|
|
topic_id: nil,
|
|
|
|
filter_old_unread: false,
|
|
|
|
skip_new: false,
|
|
|
|
skip_unread: false,
|
|
|
|
skip_order: false,
|
|
|
|
staff: false,
|
|
|
|
admin: false,
|
2022-06-30 08:18:12 +08:00
|
|
|
whisperer: false,
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
select: nil,
|
2021-06-03 09:21:33 +08:00
|
|
|
custom_state_filter: nil,
|
|
|
|
additional_join_sql: nil
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
)
|
2015-09-07 09:57:50 +08:00
|
|
|
unread =
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
if skip_unread
|
2015-09-07 09:57:50 +08:00
|
|
|
"1=0"
|
|
|
|
else
|
2022-06-30 08:18:12 +08:00
|
|
|
unread_filter_sql(whisperer: whisperer)
|
2015-09-07 09:57:50 +08:00
|
|
|
end
|
|
|
|
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
filter_old_unread_sql =
|
|
|
|
if filter_old_unread
|
PERF: Optimise `TopicTrackingState.report` query to speed up query (#22871)
In the query generated by `TopicTrackingState.report`, there are two
subqueies being executed. The first subquery fetches all the topics
that are new for a given user while the second subquery fetches all the topics with
unread posts for a given user. For the second subquery, there is a
filter `topics.updated_at >= user_stats.first_unread_at` which is used
as a performance optimisation to reduce the number of rows that PG has
to scan through the `topics` table.
However, we started to notice in production that the PG planner doesn't
always execute the filter first to reduce the number of rows that it has
to scan through. Running the following query in one of our production
instance,
```
EXPLAIN ANALYZE
SELECT
DISTINCT topics.id as topic_id,
u.id as user_id,
topics.created_at,
topics.updated_at,
topics.highest_staff_post_number AS highest_post_number,
last_read_post_number,
c.id as category_id,
c.topic_id AS category_topic_id,
tu.notification_level,
us.first_unread_at,
GREATEST(
CASE
WHEN COALESCE(uo.new_topic_duration_minutes, 2880) = -1 THEN u.created_at
WHEN COALESCE(uo.new_topic_duration_minutes, 2880) = -2 THEN COALESCE(
u.previous_visit_at,u.created_at
)
ELSE ('2023-07-31 03:29:45.737630'::timestamp - INTERVAL '1 MINUTE' * COALESCE(uo.new_topic_duration_minutes, 2880))
END, u.created_at, '2023-07-25 15:06:44'
) AS treat_as_new_topic_start_date
FROM topics
JOIN users u on u.id = 13455
JOIN user_stats AS us ON us.user_id = u.id
JOIN user_options AS uo ON uo.user_id = u.id
JOIN categories c ON c.id = topics.category_id
LEFT JOIN topic_users tu ON tu.topic_id = topics.id AND tu.user_id = u.id
WHERE u.id = 13455 AND
topics.updated_at >= us.first_unread_at AND
topics.archetype <> 'private_message' AND
(("topics"."deleted_at" IS NULL AND (tu.last_read_post_number < topics.highest_staff_post_number) AND (COALESCE(tu.notification_level, 1) >= 2)) OR (1=0)) AND
NOT (
COALESCE((select array_agg(tag_id) from topic_tags where topic_tags.topic_id = topics.id), ARRAY[]::int[]) && ARRAY[451,452,453]
) AND
topics.deleted_at IS NULL AND
NOT (
last_read_post_number IS NULL AND
(
topics.category_id IN (SELECT "categories"."id" FROM "categories" LEFT JOIN categories categories2 ON categories2.id = categories.parent_category_id LEFT JOIN category_users ON category_users.category_id = categories.id AND category_users.user_id = 13455 LEFT JOIN category_users category_users2 ON category_users2.category_id = categories2.id AND category_users2.user_id = 13455 WHERE ((category_users.id IS NULL AND COALESCE(category_users2.notification_level, 1) = 0) OR COALESCE(category_users.notification_level, 1) = 0))
AND tu.notification_level <= 1
)
)
```
we get the following
```
QUERY PLAN
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Unique (cost=201606.06..201608.15 rows=76 width=60) (actual time=91.279..91.294 rows=14 loops=1)
-> Sort (cost=201606.06..201606.25 rows=76 width=60) (actual time=91.278..91.284 rows=14 loops=1)
Sort Key: topics.id, topics.created_at, topics.updated_at, topics.highest_staff_post_number, tu.last_read_post_number, c.id, c.topic_id, tu.notification_level, us.first_unread_at, (GREATEST(CASE WHEN (COALESCE(uo.new_topic_duration_minutes, 2880) = '-1'::integer) THEN u.created_at WHEN (COALESCE(uo.new_topic_duration_minutes, 2880) = '-2'::integer) THEN COALESCE(u.previous_visit_at, u.created_at) ELSE ('2023-07-31 03:29:45.73763'::timestamp without time zone - ('00:01:00'::interval * (COALESCE(uo.new_topic_duration_minutes, 2880))::double precision)) END, u.created_at, '2023-07-25 15:06:44'::timestamp without time zone))
Sort Method: quicksort Memory: 26kB
-> Hash Join (cost=97519.51..201603.69 rows=76 width=60) (actual time=87.662..91.268 rows=14 loops=1)
Hash Cond: (topics.id = tu.topic_id)
Join Filter: ((tu.last_read_post_number < topics.highest_staff_post_number) AND ((tu.last_read_post_number IS NOT NULL) OR (NOT (hashed SubPlan 2)) OR (tu.notification_level > 1)))
Rows Removed by Join Filter: 10
-> Nested Loop (cost=1.54..104075.36 rows=3511 width=68) (actual time=0.055..3.609 rows=548 loops=1)
-> Nested Loop (cost=1.13..25.20 rows=1 width=32) (actual time=0.027..0.033 rows=1 loops=1)
-> Nested Loop (cost=0.71..16.76 rows=1 width=28) (actual time=0.020..0.023 rows=1 loops=1)
-> Index Scan using users_pkey on users u (cost=0.42..8.44 rows=1 width=20) (actual time=0.010..0.012 rows=1 loops=1)
Index Cond: (id = 13455)
-> Index Scan using user_stats_pkey on user_stats us (cost=0.29..8.31 rows=1 width=12) (actual time=0.008..0.010 rows=1 loops=1)
Index Cond: (user_id = 13455)
-> Index Scan using index_user_options_on_user_id_and_default_calendar on user_options uo (cost=0.42..8.44 rows=1 width=8) (actual time=0.007..0.008 rows=1 loops=1)
Index Cond: (user_id = 13455)
-> Nested Loop (cost=0.41..104015.12 rows=3504 width=36) (actual time=0.026..3.503 rows=548 loops=1)
-> Seq Scan on categories c (cost=0.00..13.73 rows=73 width=8) (actual time=0.003..0.039 rows=73 loops=1)
-> Index Only Scan using index_topics_on_updated_at_public on topics (cost=0.41..1424.20 rows=48 width=28) (actual time=0.012..0.046 rows=8 loops=73)
Index Cond: ((updated_at >= us.first_unread_at) AND (category_id = c.id))
Filter: (NOT (COALESCE((SubPlan 1), '{}'::integer[]) && '{451,452,453}'::integer[]))
Heap Fetches: 553
SubPlan 1
-> Aggregate (cost=4.31..4.32 rows=1 width=32) (actual time=0.002..0.002 rows=1 loops=548)
-> Index Only Scan using index_topic_tags_on_topic_id_and_tag_id on topic_tags (cost=0.29..4.31 rows=1 width=4) (actual time=0.002..0.002 rows=1 loops=548)
Index Cond: (topic_id = topics.id)
Heap Fetches: 178
-> Hash (cost=97222.14..97222.14 rows=19914 width=16) (actual time=87.545..87.546 rows=42884 loops=1)
Buckets: 65536 (originally 32768) Batches: 1 (originally 1) Memory Usage: 2387kB
-> Bitmap Heap Scan on topic_users tu (cost=1217.47..97222.14 rows=19914 width=16) (actual time=14.419..78.286 rows=42884 loops=1)
Recheck Cond: (user_id = 13455)
Filter: (COALESCE(notification_level, 1) >= 2)
Rows Removed by Filter: 15839
Heap Blocks: exact=45285
-> Bitmap Index Scan on index_topic_users_on_user_id_and_topic_id (cost=0.00..1212.49 rows=59741 width=0) (actual time=6.448..6.448 rows=58723 loops=1)
Index Cond: (user_id = 13455)
SubPlan 2
-> Nested Loop Left Join (cost=0.74..46.90 rows=1 width=4) (never executed)
Join Filter: (category_users2.category_id = categories2.id)
Filter: (((category_users.id IS NULL) AND (COALESCE(category_users2.notification_level, 1) = 0)) OR (COALESCE(category_users.notification_level, 1) = 0))
-> Nested Loop Left Join (cost=0.45..32.31 rows=73 width=16) (never executed)
Join Filter: (category_users.category_id = categories.id)
-> Nested Loop Left Join (cost=0.15..18.45 rows=73 width=8) (never executed)
-> Seq Scan on categories (cost=0.00..13.73 rows=73 width=8) (never executed)
-> Memoize (cost=0.15..0.28 rows=1 width=4) (never executed)
Cache Key: categories.parent_category_id
Cache Mode: logical
-> Index Only Scan using categories_pkey on categories categories2 (cost=0.14..0.27 rows=1 width=4) (never executed)
Index Cond: (id = categories.parent_category_id)
Heap Fetches: 0
-> Materialize (cost=0.29..11.69 rows=2 width=12) (never executed)
-> Index Scan using idx_category_users_user_id_category_id on category_users (cost=0.29..11.68 rows=2 width=12) (never executed)
Index Cond: (user_id = 13455)
-> Materialize (cost=0.29..11.69 rows=2 width=8) (never executed)
-> Index Scan using idx_category_users_user_id_category_id on category_users category_users2 (cost=0.29..11.68 rows=2 width=8) (never executed)
Index Cond: (user_id = 13455)
Planning Time: 1.740 ms
Execution Time: 91.414 ms
(59 rows)
```
From the execution plan, we can see the most of the time is spent
joining about 42888 rows in the `topics` table to the `topic_users` table.
However, we know that we only have to scan through a
subset of the `topics` table because the user's last unread at is '2023-07-20 11:33:05'.
If we filter the `topics` table with `topics.updated_at >= '2023-07-20 11:33:05'`, this would only
return about 1500 rows.
From our testing in production, the PG planner is able to execute a
better query plan when we avoid the unnecessary joins on `user_stats` just to be
able to get the user's `UserStat#first_unread_at`. Instead, we can just
pass the value of `UserStat#first_unread_at` directly as a query
parameter.
```
EXPLAIN ANALYZE
SELECT
DISTINCT topics.id as topic_id,
u.id as user_id,
topics.created_at,
topics.updated_at,
topics.highest_staff_post_number AS highest_post_number,
last_read_post_number,
c.id as category_id,
c.topic_id AS category_topic_id,
tu.notification_level,
GREATEST(
CASE
WHEN COALESCE(uo.new_topic_duration_minutes, 2880) = -1 THEN u.created_at
WHEN COALESCE(uo.new_topic_duration_minutes, 2880) = -2 THEN COALESCE(
u.previous_visit_at,u.created_at
)
ELSE ('2023-07-31 03:29:45.737630'::timestamp - INTERVAL '1 MINUTE' * COALESCE(uo.new_topic_duration_minutes, 2880))
END, u.created_at, '2023-07-25 15:06:44'
) AS treat_as_new_topic_start_date
FROM topics
JOIN users u on u.id = 13455
JOIN user_options AS uo ON uo.user_id = u.id
JOIN categories c ON c.id = topics.category_id
LEFT JOIN topic_users tu ON tu.topic_id = topics.id AND tu.user_id = u.id
WHERE u.id = 13455 AND
topics.updated_at >= '2023-07-20 11:33:05' AND
topics.archetype <> 'private_message' AND
(("topics"."deleted_at" IS NULL AND (tu.last_read_post_number < topics.highest_staff_post_number) AND (COALESCE(tu.notification_level, 1) >= 2)) OR (1=0)) AND
NOT (
COALESCE((select array_agg(tag_id) from topic_tags where topic_tags.topic_id = topics.id), ARRAY[]::int[]) && ARRAY[451,452,453]
) AND
topics.deleted_at IS NULL AND
NOT (
last_read_post_number IS NULL AND
(
topics.category_id IN (SELECT "categories"."id" FROM "categories" LEFT JOIN categories categories2 ON categories2.id = categories.parent_category_id LEFT JOIN category_users ON category_users.category_id = categories.id AND category_users.user_id = 13455 LEFT JOIN category_users category_users2 ON category_users2.category_id = categories2.id AND category_users2.user_id = 13455 WHERE ((category_users.id IS NULL AND COALESCE(category_users2.notification_level, 1) = 0) OR COALESCE(category_users.notification_level, 1) = 0))
AND tu.notification_level <= 1
)
);
```
Note how the filter is now `topics.updated_at >= '2023-07-20 11:33:05'`
instead of `topics.updated_at >= us.first_unread_at`. The modified query
above generates the following execution plan.
```
QUERY PLAN
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Unique (cost=5189.86..5189.88 rows=1 width=52) (actual time=4.991..5.002 rows=14 loops=1)
-> Sort (cost=5189.86..5189.86 rows=1 width=52) (actual time=4.990..4.994 rows=14 loops=1)
Sort Key: topics.id, topics.created_at, topics.updated_at, topics.highest_staff_post_number, tu.last_read_post_number, c.id, c.topic_id, tu.notification_level, (GREATEST(CASE WHEN (COALESCE(uo.new_topic_duration_minutes, 2880) = '-1'::integer) THEN u.created_at WHEN (COALESCE(uo.new_topic_duration_minutes, 2880) = '-2'::integer) THEN COALESCE(u.previous_visit_at, u.created_at) ELSE ('2023-07-31 03:29:45.73763'::timestamp without time zone - ('00:01:00'::interval * (COALESCE(uo.new_topic_duration_minutes, 2880))::double precision)) END, u.created_at, '2023-07-25 15:06:44'::timestamp without time zone))
Sort Method: quicksort Memory: 26kB
-> Nested Loop (cost=52.11..5189.85 rows=1 width=52) (actual time=0.093..4.974 rows=14 loops=1)
-> Nested Loop (cost=51.70..5181.39 rows=1 width=60) (actual time=0.084..4.931 rows=14 loops=1)
-> Nested Loop (cost=51.28..5172.94 rows=1 width=44) (actual time=0.076..4.887 rows=14 loops=1)
-> Nested Loop (cost=0.41..1698.46 rows=59 width=36) (actual time=0.029..3.537 rows=548 loops=1)
-> Seq Scan on categories c (cost=0.00..13.73 rows=73 width=8) (actual time=0.005..0.039 rows=73 loops=1)
-> Index Only Scan using index_topics_on_updated_at_public on topics (cost=0.41..23.07 rows=1 width=28) (actual time=0.012..0.047 rows=8 loops=73)
Index Cond: ((updated_at >= '2023-07-20 11:33:05'::timestamp without time zone) AND (category_id = c.id))
Filter: (NOT (COALESCE((SubPlan 1), '{}'::integer[]) && '{451,452,453}'::integer[]))
Heap Fetches: 552
SubPlan 1
-> Aggregate (cost=4.31..4.32 rows=1 width=32) (actual time=0.002..0.002 rows=1 loops=548)
-> Index Only Scan using index_topic_tags_on_topic_id_and_tag_id on topic_tags (cost=0.29..4.31 rows=1 width=4) (actual time=0.002..0.002 rows=1 loops=548)
Index Cond: (topic_id = topics.id)
Heap Fetches: 178
-> Index Scan using index_topic_users_on_user_id_and_topic_id on topic_users tu (cost=50.86..58.88 rows=1 width=16) (actual time=0.002..0.002 rows=0 loops=548)
Index Cond: ((user_id = 13455) AND (topic_id = topics.id))
Filter: ((COALESCE(notification_level, 1) >= 2) AND (last_read_post_number < topics.highest_staff_post_number) AND ((last_read_post_number IS NOT NULL) OR (NOT (hashed SubPlan 2)) OR (notification_level > 1)))
Rows Removed by Filter: 0
SubPlan 2
-> Nested Loop Left Join (cost=0.74..50.43 rows=1 width=4) (never executed)
Join Filter: (category_users2.category_id = categories2.id)
Filter: (((category_users.id IS NULL) AND (COALESCE(category_users2.notification_level, 1) = 0)) OR (COALESCE(category_users.notification_level, 1) = 0))
-> Nested Loop Left Join (cost=0.45..35.84 rows=73 width=16) (never executed)
Join Filter: (category_users.category_id = categories.id)
-> Nested Loop Left Join (cost=0.15..21.97 rows=73 width=8) (never executed)
-> Seq Scan on categories (cost=0.00..13.73 rows=73 width=8) (never executed)
-> Memoize (cost=0.15..0.61 rows=1 width=4) (never executed)
Cache Key: categories.parent_category_id
Cache Mode: logical
-> Index Only Scan using categories_pkey on categories categories2 (cost=0.14..0.60 rows=1 width=4) (never executed)
Index Cond: (id = categories.parent_category_id)
Heap Fetches: 0
-> Materialize (cost=0.29..11.69 rows=2 width=12) (never executed)
-> Index Scan using idx_category_users_user_id_category_id on category_users (cost=0.29..11.68 rows=2 width=12) (never executed)
Index Cond: (user_id = 13455)
-> Materialize (cost=0.29..11.69 rows=2 width=8) (never executed)
-> Index Scan using idx_category_users_user_id_category_id on category_users category_users2 (cost=0.29..11.68 rows=2 width=8) (never executed)
Index Cond: (user_id = 13455)
-> Index Scan using users_pkey on users u (cost=0.42..8.44 rows=1 width=20) (actual time=0.003..0.003 rows=1 loops=14)
Index Cond: (id = 13455)
-> Index Scan using index_user_options_on_user_id_and_default_calendar on user_options uo (cost=0.42..8.44 rows=1 width=8) (actual time=0.002..0.002 rows=1 loops=14)
Index Cond: (user_id = 13455)
Planning Time: 1.281 ms
Execution Time: 5.092 ms
(48 rows)
```
With the new query, PG first does an index scan using the `index_topics_on_updated_at_public` index to filter away most of the topics making the subsequent joins much cheaper. Total query time has been reduced from ~90ms to ~5ms.
This optimisation will mostly affect users with very few/recent unread topics since a large `UserStat#firsts_unread_at` value will still mean scanning through a large portion of the `topics` table.
2023-07-31 12:21:41 +08:00
|
|
|
" topics.updated_at >= :user_first_unread_at AND "
|
2019-04-05 09:44:36 +08:00
|
|
|
else
|
|
|
|
""
|
|
|
|
end
|
|
|
|
|
2015-09-07 09:57:50 +08:00
|
|
|
new =
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
if skip_new
|
2015-09-07 09:57:50 +08:00
|
|
|
"1=0"
|
|
|
|
else
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
new_filter_sql
|
2015-09-07 09:57:50 +08:00
|
|
|
end
|
|
|
|
|
2023-03-06 10:13:10 +08:00
|
|
|
category_topic_id_column_select =
|
|
|
|
if SiteSetting.show_category_definitions_in_topic_lists
|
|
|
|
""
|
|
|
|
else
|
|
|
|
"c.topic_id AS category_topic_id,"
|
|
|
|
end
|
|
|
|
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
select_sql =
|
|
|
|
select ||
|
|
|
|
"
|
2021-06-03 09:21:33 +08:00
|
|
|
DISTINCT topics.id as topic_id,
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
u.id as user_id,
|
2014-02-27 04:37:42 +08:00
|
|
|
topics.created_at,
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
topics.updated_at,
|
2022-06-30 08:18:12 +08:00
|
|
|
#{highest_post_number_column_select(whisperer)},
|
2014-02-27 04:37:42 +08:00
|
|
|
last_read_post_number,
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
c.id as category_id,
|
2023-03-06 10:13:10 +08:00
|
|
|
#{category_topic_id_column_select}
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
tu.notification_level,
|
|
|
|
GREATEST(
|
|
|
|
CASE
|
|
|
|
WHEN COALESCE(uo.new_topic_duration_minutes, :default_duration) = :always THEN u.created_at
|
|
|
|
WHEN COALESCE(uo.new_topic_duration_minutes, :default_duration) = :last_visit THEN COALESCE(
|
|
|
|
u.previous_visit_at,u.created_at
|
|
|
|
)
|
|
|
|
ELSE (:now::timestamp - INTERVAL '1 MINUTE' * COALESCE(uo.new_topic_duration_minutes, :default_duration))
|
|
|
|
END, u.created_at, :min_date
|
|
|
|
) AS treat_as_new_topic_start_date"
|
2015-09-07 09:57:50 +08:00
|
|
|
|
2019-04-05 12:25:19 +08:00
|
|
|
category_filter =
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
if admin
|
2019-04-05 12:25:19 +08:00
|
|
|
""
|
|
|
|
else
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
append = "OR u.admin" if !admin
|
2019-04-05 12:25:19 +08:00
|
|
|
<<~SQL
|
|
|
|
(
|
2019-11-14 13:11:34 +08:00
|
|
|
NOT c.read_restricted #{append} OR c.id IN (
|
2019-04-05 12:25:19 +08:00
|
|
|
SELECT c2.id FROM categories c2
|
|
|
|
JOIN category_groups cg ON cg.category_id = c2.id
|
|
|
|
JOIN group_users gu ON gu.user_id = :user_id AND cg.group_id = gu.group_id
|
|
|
|
WHERE c2.read_restricted )
|
|
|
|
) AND
|
|
|
|
SQL
|
|
|
|
end
|
|
|
|
|
|
|
|
visibility_filter =
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
if staff
|
2019-04-05 12:25:19 +08:00
|
|
|
""
|
|
|
|
else
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
append = "OR u.admin OR u.moderator" if !staff
|
2019-04-05 12:25:19 +08:00
|
|
|
"(topics.visible #{append}) AND"
|
|
|
|
end
|
|
|
|
|
2020-05-29 10:59:34 +08:00
|
|
|
tags_filter = ""
|
|
|
|
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
if muted_tag_ids.present? &&
|
|
|
|
%w[always only_muted].include?(SiteSetting.remove_muted_tags_from_latest)
|
2020-05-29 10:59:34 +08:00
|
|
|
existing_tags_sql =
|
|
|
|
"(select array_agg(tag_id) from topic_tags where topic_tags.topic_id = topics.id)"
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
muted_tags_array_sql = "ARRAY[#{muted_tag_ids.join(",")}]"
|
2020-05-29 10:59:34 +08:00
|
|
|
|
|
|
|
if SiteSetting.remove_muted_tags_from_latest == "always"
|
|
|
|
tags_filter = <<~SQL
|
|
|
|
NOT (
|
|
|
|
COALESCE(#{existing_tags_sql}, ARRAY[]::int[]) && #{muted_tags_array_sql}
|
|
|
|
) AND
|
2019-12-10 06:50:05 +08:00
|
|
|
SQL
|
2020-05-29 10:59:34 +08:00
|
|
|
else # only muted
|
|
|
|
tags_filter = <<~SQL
|
|
|
|
NOT (
|
|
|
|
COALESCE(#{existing_tags_sql}, ARRAY[-999]) <@ #{muted_tags_array_sql}
|
|
|
|
) AND
|
2019-12-10 06:50:05 +08:00
|
|
|
SQL
|
|
|
|
end
|
2020-05-29 10:59:34 +08:00
|
|
|
end
|
2019-12-10 06:50:05 +08:00
|
|
|
|
2019-04-05 12:25:19 +08:00
|
|
|
sql = +<<~SQL
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
SELECT #{select_sql}
|
|
|
|
FROM topics
|
|
|
|
JOIN users u on u.id = :user_id
|
|
|
|
JOIN user_options AS uo ON uo.user_id = u.id
|
|
|
|
JOIN categories c ON c.id = topics.category_id
|
|
|
|
LEFT JOIN topic_users tu ON tu.topic_id = topics.id AND tu.user_id = u.id
|
2021-08-16 12:18:26 +08:00
|
|
|
#{skip_new ? "" : "LEFT JOIN dismissed_topic_users ON dismissed_topic_users.topic_id = topics.id AND dismissed_topic_users.user_id = :user_id"}
|
2021-06-03 09:21:33 +08:00
|
|
|
#{additional_join_sql}
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
WHERE u.id = :user_id AND
|
|
|
|
#{filter_old_unread_sql}
|
|
|
|
topics.archetype <> 'private_message' AND
|
|
|
|
#{custom_state_filter ? custom_state_filter : "((#{unread}) OR (#{new})) AND"}
|
|
|
|
#{visibility_filter}
|
|
|
|
#{tags_filter}
|
|
|
|
topics.deleted_at IS NULL AND
|
|
|
|
#{category_filter}
|
|
|
|
NOT (
|
|
|
|
#{(skip_new && skip_unread) ? "" : "last_read_post_number IS NULL AND"}
|
|
|
|
(
|
2022-03-02 12:02:09 +08:00
|
|
|
topics.category_id IN (#{CategoryUser.muted_category_ids_query(user, include_direct: true).select("categories.id").to_sql})
|
DEV: Topic tracking state improvements (#13218)
I merged this PR in yesterday, finally thinking this was done https://github.com/discourse/discourse/pull/12958 but then a wild performance regression occurred. These are the problem methods:
https://github.com/discourse/discourse/blob/1aa20bd681e634f7fff22953ed62d90c2573b331/app/serializers/topic_tracking_state_serializer.rb#L13-L21
Turns out date comparison is super expensive on the backend _as well as_ the frontend.
The fix was to just move the `treat_as_new_topic_start_date` into the SQL query rather than using the slower `UserOption#treat_as_new_topic_start_date` method in ruby. After this change, 1% of the total time is spent with the `created_in_new_period` comparison instead of ~20%.
----
History:
Original PR which had to be reverted **https://github.com/discourse/discourse/pull/12555**. See the description there for what this PR is achieving, plus below.
The issue with the original PR is addressed in https://github.com/discourse/discourse/pull/12958/commits/92ef54f4020111ffacb0f2a27da5d5c2855f9d5d
If you went to the `x unread` link for a tag Chrome would freeze up and possibly crash, or eventually unfreeze after nearly 10 mins. Other routes for unread/new were similarly slow. From profiling the issue was the `sync` function of `topic-tracking-state.js`, which calls down to `isNew` which in turn calls `moment`, a change I had made in the PR above. The time it takes locally with ~1400 topics in the tracking state is 2.3 seconds.
To solve this issue, I have moved these calculations for "created in new period" and "unread not too old" into the tracking state serializer.
When I was looking at the profiler I also noticed this issue which was just compounding the problem. Every time we modify topic tracking state we recalculate the sidebar tracking/everything/tag counts. However this calls `forEachTracked` and `countTags` which can be quite expensive as they go through the whole tracking state (and were also calling the removed moment functions).
I added some logs and this was being called 30 times when navigating to a new /unread route because `sync` is being called from `build-topic-route` (one for each topic loaded due to pagination). So I just added a debounce here and it makes things even faster.
Finally, I changed topic tracking state to use a Map so our counts of the state keys is faster (Maps have .size whereas objects you have to do Object.keys(obj) which is O(n).)
<!-- NOTE: All pull requests should have tests (rspec in Ruby, qunit in JavaScript). If your code does not include test coverage, please include an explanation of why it was omitted. -->
2021-06-02 07:06:29 +08:00
|
|
|
AND tu.notification_level <= #{TopicUser.notification_levels[:regular]}
|
|
|
|
)
|
|
|
|
)
|
|
|
|
SQL
|
|
|
|
|
2015-07-21 19:53:54 +08:00
|
|
|
sql << " AND topics.id = :topic_id" if topic_id
|
2014-09-10 20:19:24 +08:00
|
|
|
|
2015-09-29 09:55:09 +08:00
|
|
|
sql << " ORDER BY topics.bumped_at DESC" unless skip_order
|
|
|
|
|
|
|
|
sql
|
2013-05-21 14:39:51 +08:00
|
|
|
end
|
|
|
|
|
2022-06-30 08:18:12 +08:00
|
|
|
def self.highest_post_number_column_select(whisperer)
|
|
|
|
"#{whisperer ? "topics.highest_staff_post_number AS highest_post_number" : "topics.highest_post_number"}"
|
2021-06-03 09:21:33 +08:00
|
|
|
end
|
|
|
|
|
2019-08-27 20:09:00 +08:00
|
|
|
def self.publish_read_indicator_on_write(topic_id, last_read_post_number, user_id)
|
|
|
|
topic =
|
|
|
|
Topic
|
|
|
|
.includes(:allowed_groups)
|
|
|
|
.select(:highest_post_number, :archetype, :id)
|
|
|
|
.find_by(id: topic_id)
|
|
|
|
|
2019-08-30 00:38:23 +08:00
|
|
|
if topic&.private_message?
|
2019-08-27 20:09:00 +08:00
|
|
|
groups = read_allowed_groups_of(topic)
|
2019-08-29 23:03:43 +08:00
|
|
|
update_topic_list_read_indicator(topic, groups, topic.highest_post_number, user_id, true)
|
2019-08-27 20:09:00 +08:00
|
|
|
end
|
|
|
|
end
|
|
|
|
|
|
|
|
def self.publish_read_indicator_on_read(topic_id, last_read_post_number, user_id)
|
|
|
|
topic =
|
|
|
|
Topic
|
|
|
|
.includes(:allowed_groups)
|
|
|
|
.select(:highest_post_number, :archetype, :id)
|
|
|
|
.find_by(id: topic_id)
|
|
|
|
|
2019-08-30 00:38:23 +08:00
|
|
|
if topic&.private_message?
|
2019-08-27 20:09:00 +08:00
|
|
|
groups = read_allowed_groups_of(topic)
|
|
|
|
post = Post.find_by(topic_id: topic.id, post_number: last_read_post_number)
|
2019-09-09 09:29:15 +08:00
|
|
|
trigger_post_read_count_update(post, groups, last_read_post_number, user_id)
|
2019-08-29 23:03:43 +08:00
|
|
|
update_topic_list_read_indicator(topic, groups, last_read_post_number, user_id, false)
|
2019-08-27 20:09:00 +08:00
|
|
|
end
|
|
|
|
end
|
|
|
|
|
|
|
|
def self.read_allowed_groups_of(topic)
|
|
|
|
topic
|
|
|
|
.allowed_groups
|
|
|
|
.joins(:group_users)
|
|
|
|
.where(publish_read_state: true)
|
|
|
|
.select("ARRAY_AGG(group_users.user_id) AS members", :name, :id)
|
|
|
|
.group("groups.id")
|
|
|
|
end
|
|
|
|
|
2019-08-29 23:03:43 +08:00
|
|
|
def self.update_topic_list_read_indicator(
|
|
|
|
topic,
|
|
|
|
groups,
|
|
|
|
last_read_post_number,
|
|
|
|
user_id,
|
|
|
|
write_event
|
|
|
|
)
|
2019-08-27 20:09:00 +08:00
|
|
|
return unless last_read_post_number == topic.highest_post_number
|
2019-08-29 23:03:43 +08:00
|
|
|
message = { topic_id: topic.id, show_indicator: write_event }.as_json
|
2019-08-27 20:09:00 +08:00
|
|
|
groups_to_update = []
|
|
|
|
|
|
|
|
groups.each do |group|
|
|
|
|
member = group.members.include?(user_id)
|
|
|
|
|
2019-08-29 23:03:43 +08:00
|
|
|
member_writing = (write_event && member)
|
|
|
|
non_member_reading = (!write_event && !member)
|
2019-08-27 20:09:00 +08:00
|
|
|
next if non_member_reading || member_writing
|
|
|
|
|
|
|
|
groups_to_update << group
|
|
|
|
end
|
|
|
|
|
|
|
|
return if groups_to_update.empty?
|
2019-08-29 23:03:43 +08:00
|
|
|
MessageBus.publish(
|
|
|
|
"/private-messages/unread-indicator/#{topic.id}",
|
|
|
|
message,
|
|
|
|
user_ids: groups_to_update.flat_map(&:members),
|
|
|
|
)
|
2019-08-27 20:09:00 +08:00
|
|
|
end
|
|
|
|
|
2019-09-09 09:29:15 +08:00
|
|
|
def self.trigger_post_read_count_update(post, groups, last_read_post_number, user_id)
|
2019-10-02 08:57:34 +08:00
|
|
|
return if !post
|
2019-08-27 20:09:00 +08:00
|
|
|
return if groups.empty?
|
2019-09-09 09:29:15 +08:00
|
|
|
opts = { readers_count: post.readers_count, reader_id: user_id }
|
|
|
|
post.publish_change_to_clients!(:read, opts)
|
2019-08-27 20:09:00 +08:00
|
|
|
end
|
2023-01-11 06:15:52 +08:00
|
|
|
|
|
|
|
def self.secure_category_group_ids(topic)
|
2023-01-12 08:22:28 +08:00
|
|
|
category = topic.category
|
2023-01-11 06:15:52 +08:00
|
|
|
|
2023-04-24 22:28:10 +08:00
|
|
|
return [Group::AUTO_GROUPS[:admins]] if category.nil?
|
|
|
|
|
2023-01-12 08:22:28 +08:00
|
|
|
if category.read_restricted
|
|
|
|
ids = [Group::AUTO_GROUPS[:admins]]
|
|
|
|
ids.push(*category.secure_group_ids)
|
|
|
|
ids.uniq
|
2023-01-11 06:15:52 +08:00
|
|
|
else
|
2023-01-12 08:22:28 +08:00
|
|
|
nil
|
2023-01-11 06:15:52 +08:00
|
|
|
end
|
|
|
|
end
|
|
|
|
private_class_method :secure_category_group_ids
|
2013-05-21 14:39:51 +08:00
|
|
|
end
|