During search indexing we "stuff" the index with additional keywords for
entities that look like domain names.
This allows searches for `cnn` to find URLs for `www.cnn.com`
The search stuffing attempted to keep indexes aligned at the correct positions
by remapping the indexed terms. However under certain edge cases a single
word can stem into 2 different lexemes. If this happened we had an off by
one which caused the entire indexing to fail.
We work around this edge case (and carry incorrect index positions) for cases
like this. It is unlikely to impact search quality at all given index position
makes almost no difference in the search algorithm.
This behavior is hard to test as it's mostly fixing a race condition: User A sends a message at the same time than User B, which as a result doesn't cause a scroll for the second message and we don't update last read unless we do a small up and down scroll.
`updateLastRead` is debounced so it has no direct consequences to call it slightly more often than what should ideally be needed.
Prior to this fix, the upload was removed from DOM when collapsed and not decorated again on expand, which was causing lightbox to not get reapplied. The fix is reverting to previous state where content was not removed from DOM.
Prior to this change `registered_bookmarkable` would return `nil` as `type` in `Bookmark.registered_bookmarkable_from_type(type)` would be `ChatMessage` and we registered a `Chat::Message` class.
This commit will now properly rely on each model `polymorphic_class_for(name)` to help us infer the proper type from a a `bookmarkable_type`.
Tests have also been added to ensure that creating/destroying chat message bookmarks is working correctly.
---
Longer explanation
Currently when you save a bookmark in the database, it's associated to another object through a polymorphic relationship, which will is represented by two columns: `bookmarkable_id` and `bookmarkable_type`. The `bookmarkable_id` contains the id of the relationship (a post ID for example) and the `bookmarkable_type` contains the type of the object as a string by default, (`"Post"` for example).
Chat plugin just started namespacing objects, as a result a model named `ChatMessage` is now named `Chat::Message`, to avoid complex and risky migrations we rely on methods provided by rails to alter the `bookmarkable_type` when we save it: we want to still save it as `"ChatMessage"` and not `"Chat::Message"`. And, to retrieve the correct model when we load the bookmark from the database: we want `"ChatMessage"` to load the `Chat::Message` model and not the `ChatMessage`model which doesn't exist anymore.
On top of this the bookmark codepath is allowing plugins to register types and will check against these types, so we alter this code path to be able to do a similar ChatMessage <-> Chat::Message dance and allow to check the type is valid. In the specific case of this commit, we were retrieving a `"ChatMessage"` bookmarkable_type from the DB and looking for it in the registered bookmarkable types which contain `Chat::Message` and not `ChatMessage`.
This commit main goal was to comply with Zeitwerk and properly rely on autoloading. To achieve this, most resources have been namespaced under the `Chat` module.
- Given all models are now namespaced with `Chat::` and would change the stored types in DB when using polymorphism or STI (single table inheritance), this commit uses various Rails methods to ensure proper class is loaded and the stored name in DB is unchanged, eg: `Chat::Message` model will be stored as `"ChatMessage"`, and `"ChatMessage"` will correctly load `Chat::Message` model.
- Jobs are now using constants only, eg: `Jobs::Chat::Foo` and should only be enqueued this way
Notes:
- This commit also used this opportunity to limit the number of registered css files in plugin.rb
- `discourse_dev` support has been removed within this commit and will be reintroduced later
<!-- 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. -->
- Install `@ember/legacy-built-in-components` and update our import statements to use it
- Remove our custom attributeBinding extensions of `TextField` and `TextArea`. Modern ember 'angle bracket syntax' allows us to apply html attributes to a component's element without needing attributeBindings
One of the problems here was coming from the ember-jquery addon. This commit skips the problematic shim from the addon and re-implements in Discourse. This hack will only be required short-term - we'll be totally dropping the ember-jquery integration as part of our upgrade to Ember 4.x.
Removing this shim means we can also remove our `discourse-ensure-deprecation-order` dummy addon which was ensuring that the ember-jquery-triggered deprecation was covered by ember-cli-deprecation-workflow.
Using `create_or_find_by!`, followed by `update_all!` requires two or three queries (two when the row doesn't already exist, three when it does). Instead, we can use postgres's native `INSERT ... ON CONFLICT ... DO UPDATE SET` feature to do the logic in a single atomic call.
Mentions are now displayed as using the non-cooked message which fixes
the problem. This is not ideal. I think we might want to rework how
these excerpts are created and rendered in the near future.
Co-authored-by: Jan Cernik <jancernik12@gmail.com>
Non-markdown tags weren't being escaped in chat excerpts. This could be
triggered by editing a chat message containing a tag (self XSS), or by
replying to a chat message with a tag (XSS).
Co-authored-by: Jan Cernik <jancernik12@gmail.com>
In production, `eager_load=true`. This sometimes leads to boot errors which are not present in dev/test environments. Running `zeitwerk:check` in CI will help us to pick up on any errors early.
This commit also introduces a `DISCOURSE_ZEITWERK_EAGER_LOAD` environment variable to make it easier to toggle the behaviour when developing locally.
This comment has nothing to do with the `eager_load` configuration. It must be left over from some historical refactoring. Removing to avoid confusion.
Previously, FETCH_HEAD would always point to tests-passed because our base docker image was configured to only fetch the tests-passed branch. Since https://github.com/discourse/discourse_docker/commit/53bbacc882, we switched to a partial clone which means that `git fetch; git checkout FETCH_HEAD` will checkout whichever remote branch is the first alphabetically. This commit makes the checkout more specific to avoid this issue.
This regressed with the commit fa543cd. Starting from that commit, we create mention records even if a user shouldn't be notified. So when sending emails, we should be making sure if a notification was actually created for a mention. This is essentially the whole fix that we need here. Tests will be provided in a following PR.
# Context
https://meta.discourse.org/t/timeline-timestamp-not-updating/256447/1
During the upgrade of the topic-timeline to glimmer the "latest post" timestamp was not updating on the timeline in relation to the relative age of the post. It was only updating on a hard refresh.
# Fix
Use the `age-with-tooltip` helper to update the created_at date automatically as time passes.
# Additional
Add the ability to pass params to `age-with-tooltip` so that we can include options like `addAgo` and `defaultFormat`
Tested with Chrome <100, Firefox <100 and Samsung Internet 20, using
percentage-based height here works better for these browsers.
This was especially a problem for the Samsung Internet browser, because
it previously was hiding the "Dismiss" button on the user profile menu.
