Some sites are still on the legacy "hamburger dropdown"
navigation_menu setting. In this case to avoid confusion,
we want to show both the sidebar icon and the header dropdown
hamburger when visiting the admin portal. Otherwise, the
hamburger switches sides from right to left for admins
and takes on different behaviour.
The hamburger in this case _only_ shows the main panel, not
other sidebar panels like the admin one.
Recently a bug was introduced when the admin sidebar section was made bold.
When the admin sidebar is disabled, we display the original sidebar in the admin panel. In that case, an incorrect CSS rule is executed.
```CSS
.admin-area .sidebar-wrapper {
background-color: var(--d-sidebar-admin-background);
.sidebar-section-header-text {
font-weight: bold;
}
}
```
Bug in this PR: https://github.com/discourse/discourse/pull/26801
To solve it, a custom CSS class with a panel key was added which will allow granular customisations.
It used to embed the objects which could lead to duplicated objects
when the same user or category was used multiple times (user was admin,
moderator and category or category was parent for multiple categories).
This reverts commit 0f4520867b.
This has led to two problems:
1. An incompatibility with Cloudflare's "auto minify" feature. They've deprecated this feature because of incompatibility with modern JS syntax. But unfortunately it will remain enabled on existing properties until 2024-08-05.
2. Discourse fails to boot in Safari 15. This is strange, because Safari does support all the required features in our production JS bundles. Even more strangely, things start working as soon as you open the developer tools. That suggests the cause could be a Safari bug rather than a simple incompatibility.
Reverting while we work out a path forward on both those issues.
- adds a `@groupIdentifier` property which will ensure that two menus of the same group are not expanded at the same time
- adds a `@class` property which will be applied to the trigger and the content
- adds a `@triggerClass` property which will be applied to the trigger
- adds a `@contentClass` property which will be applied to the trigger
- removes `extraClassName`
It's a temporary solution while I work a better solution. The problem here is quite tricky. We are showing a modal from a modal. But if we close the previous modal, before the second one is show it means we destroy the menu holding the first modal which prevents showing the second modal.
One possible solution would be to refactor d-modal’s show function. At the moment if you await on show it will await until closed and not when the modal has been inserted to the DOM. It means we don't have a clean moment to close the d-menu.
The second issue it that even though it's possible to have multiple modals on screen, the close modal assumes only one active modal at a time.
Cases like the glimmer-site-header are complex because the swiped area is not the moved target, for now it's simpler to not apply the body scroll lock automatically.
A new property is now available on the swipe modifier: `{{swipe @lockBody=false}}`
Note I tried to have tests for this modifier in the past, but it was very inconsistent on CI causing lots of flakeys, this is why there are no tests for now. I might try to write them again using system specs.
Previously, we only updated the duration and interval values in the constructor. So whenever the initial values are updated in the form the changes are not reflected in the UI. To fix this issue we're using "get" methods in this PR.
Co-authored-by: Jarek Radosz <jradosz@gmail.com>
Fixes a bug I stumbled upon in dev env:
```
Error: Assertion Failed: You attempted to update <discourse@model:user::ember337>.status to "[object Object]", but it is being tracked by a tracking context, such as a template, computed property, or observer. In order to make sure the context updates properly, you must invalidate the property when updating it. You can mark the property as `@tracked`, or use `@ember/object#set` to do this.
```
decorator-transforms (https://github.com/ef4/decorator-transforms) is a modern replacement for babel's plugin-proposal-decorators. It provides a decorator implementation using modern browser features, without needing to enable babel's full suite of class feature transformations. This improves the developer experience and performance.
In local testing with Google's 'tachometer' tool, this reduces Discourse's 'init-to-render' time by around 3-4% (230ms -> 222ms).
It reduces our initial gzip'd JS payloads by 3.2% (2.43MB -> 2.35MB), or 7.5% (14.5MB -> 13.4MB) uncompressed.
DropdownMenu component is meant as a way to describe the content of menus.
Syntax:
```
<DropdownMenu as |dm|>
<dm.item class="test">
First
</dm.item>
<dm.divider class="foo" />
<dm.item class="bar">
Second
</dm.item>
</DropdownMenu>
```
menus and tooltips are now appended to their own portals. The service are the only responsible for managing the instances, prior to this commit, services could manage one instance, but the DMenu and DTooltip components could also take over which could cause unexpected states.
This change also allows nested menus/tooltips.
Other notable changes:
- few months ago core copied the CloseOnClickOutside modifier of float-kit without removing the float-kit one, this commit now only use the core one.
- the close function is now trully async
- the close function accepts an instance or an identifier as parameter
This commit also:
uses the swipe modifier in the glimmer-site-header component
changes closing condition for d-modal and toast from distance to velocity
cancels toast auto close on touch
Previously, if you supplied your own content to DButton it would still add the character:
```hbs
<DButton>my text</DButton>
```
```html
<button>​ my text</button>
```
In this PR we started redirecting to the guide page after the wizard - https://github.com/discourse/discourse/pull/26696
The guide will require rebrand and until it is ready, we should redirect to `/latest`
Followup 2d2329095c
Previous to the above commit, in PMs the bookmark button
was icon-only and did not show a label. This restores the
same functionality.
- Rename `discourse-booted` to 'discourse-init' (because 'booted' makes it sound like boot was finished. When in fact, it was just starting)
- Introduce `discourse-paint`, which is fired after the Ember application has been painted to the screen by the browser. This happens slightly after DOMContentLoaded
- Add a `performance.measure` call to link those two marks, so they're easily visible in performance traces
Also removes an ember boot-order workaround which is no longer required.
- Use 'cheap-source-map' webpack config on low-memory machines
This results in worse quality sourcemaps in browser dev tools, but it significantly reduces memory use in our webpack build. In approximate local testing it drops from 1100mb to 590mb. This should make the rebuild process on low-memory machines much faster and less likely to trigger OOM errors.
In development, and on higher-memory machines, the higher-quality 'source-map' option is maintained.
- Disable Webpack's built-in `minimize` feature. Embroider already applies Terser after the webpack build is complete. There is no need to double-minimize the output.
- Update ember-cli-progress-ci to print to stderr instead of stdout. For some reason, pups (used by discourse_docker) buffers the stdout of commands and only prints when they are finished. stderr does not have this same limitation, so switching will mean sysadmins can see the progress of the ember build in real-time.
Given the number of variables it's hard to promise exact numbers. But, in my tests on a DO droplet with 1GB RAM (+2GB swap), this reduced the `ember build` portion of a `./launcher rebuild app` from ~50 minutes to ~15 minutes.
* Simplify config nav link generation to always inject the Settings
tab
* Auto-redirect to the first non-settings config link (if there is one)
when the user lands on /admin/plugins/:plugin_id
* Add `extras` to admin plugin serializer so plugins can add more
data on first load
* Add PikadayCalendar page object for system specs, extracted from the
CalendarDateTimePicker to make it more generic.
Those were all low hanging fruits - all were already glimmer components, so this was mostly merging js and hbs files and adding imports.
(occasionally also adds/fixes class names)
In this PR we introduced an admin sidebar for moderators - https://github.com/discourse/discourse/pull/26795
`What's new` and `all reports` links were missing as moderators have access to those pages.
At the moment, there is no way to create a group of related watched words together. If a user needed a set of words to be created together, they'll have to create them individually one at a time.
This change attempts to allow related watched words to be created as a group. The idea here is to have a list of words be tied together via a common `WatchedWordGroup` record. Given a list of words, a `WatchedWordGroup` record is created and assigned to each `WatchedWord` record. The existing WatchedWord creation behaviour remains largely unchanged.
Co-authored-by: Selase Krakani <skrakani@gmail.com>
Co-authored-by: Martin Brennan <martin@discourse.org>
This commit introduces a few changes as a result of
customer issues with finding why a topic was relisted.
In one case, if a user edited the OP of a topic that was
unlisted and hidden because of too many flags, the topic
would get relisted by directly changing topic.visible,
instead of going via TopicStatusUpdater.
To improve tracking we:
* Introduce a visibility_reason_id to topic which functions
in a similar way to hidden_reason_id on post, this column is
set from the various places we change topic visibility
* Fix Post#unhide! which was directly modifying topic.visible,
instead we use TopicStatusUpdater which sets visibility_reason_id
and also makes a small action post
* Show the reason topic visibility changed when hovering the
unlisted icon in topic status on topic titles
A change in relative picker was causing a serie of events which ultimately would cause the whole list of time options to be reset and re-rendered which would cause a new instance of the picker to be created, causing a reset.
The fix is using id in the each loop to help ember identify that it doesn’t have to re-render a specific component.
Selecting the +subcategories option does not work sometimes when "lazy
load categories" is enabled because the subcategories may not be
fetched. This ensures that subcategories are loaded by requesting them
before being used.
Our 'page_view_crawler' / 'page_view_anon' metrics are based purely on the User Agent sent by clients. This means that 'badly behaved' bots which are imitating real user agents are counted towards 'anon' page views.
This commit introduces a new method of tracking visitors. When an initial HTML request is made, we assume it is a 'non-browser' request (i.e. a bot). Then, once the JS application has booted, we notify the server to count it as a 'browser' request. This reliance on a JavaScript-capable browser matches up more closely to dedicated analytics systems like Google Analytics.
Existing data collection and graphs are unchanged. Data collected via the new technique is available in a new 'experimental' report.
Fixes two issues:
- frontend was reloading the page when clicking-to-remove avatar
- backend wasn't allowing resetting the setting by deleting all avatars
The `secondFactorMethod` property is defined as a @discourseComputed` which means it can't be overridden. Yet, we do override it in `app/assets/javascripts/discourse/app/components/security-key-form.js` and `app/assets/javascripts/discourse/app/components/second-factor-form.js` by doing `this.set("secondFactorMethod", ...)`.
This commit sets a default property `secondFactorMethod` on the `email-login` controller after the model has been loaded. Given this property is no longer computed, it can be set again at other places.
Followups:
- Ideally we would follow DDAU pattern but this is quite a significant refactor.
- The test I added is very limited, ideally we should start writing system specs for this, but it means having to deal with the email, it's a significant work.
This service-worker caching functionality was disabled by default in 1c58395bca, and the setting to re-enable was marked as experimental. Now we are dropping all the related logic.
When the user sees no results in their admin sidebar query,
we are adding two additional links:
* "Search site settings" - Navigates to the site settings page
with the filter prefilled in the search
* "Admin user list" - Navigates to the user list with the filter
prefilled in the username search
This will bridge the gap until we have a full admin-wide search.
Also make admin site setting search param refresh on filter changes
---------
Co-authored-by: Jarek Radosz <jradosz@gmail.com>
Followup to 67a8080e33
This commit makes it so the topic footer button for bookmarks
uses the new BookmarkMenu component, and makes some tweaks to
that component to allow for a label and CSS class options.
Also introduces a TopicBookmarkManager to manage the saving/editing/
deleting of the topic level bookmarks and the reactivity that happens
in the topic UI afterward.
Next commit should rip out old bookmark associated code in the
topic controller as it will no longer be needed.
---------
Co-authored-by: Joffrey JAFFEUX <j.jaffeux@gmail.com>
This commit ensures that additional keywords for admin
sidebar links (which are also stored in the admin sidebar state
manager) are translated with I18n, which was discussed
in https://meta.discourse.org/t/introducing-experimental-admin-sidebar-navigation/289281/58?u=martin
This also changes the admin sidebar state manager keywords to
not be a TrackedObject -- this is not necessary as keywords are
only set once, and it was causing rendering issues because
the keywords were being set at the same time they were read.
Finally this adds a "theme" keyword to the "Components" link
because we often refer to components as Theme Components
Co-authored-by: Jarek Radosz <jradosz@gmail.com>
* DEV: replace postAttrs dependencies in topic-map component by passing in topicDetails and postStream to topic-map to ensure state changes are passed properly down to child components
Automatically add `moderators` and `admins` auto groups to specific site settings.
In the new group-based permissions systems, we just want to check the user’s groups since it more accurately reflects reality
Affected settings:
- tag_topic_allowed_groups
- create_tag_allowed_groups
- send_email_messages_allowed_groups
- personal_message_enabled_groups
- here_mention_allowed_groups
- approve_unless_allowed_groups
- approve_new_topics_unless_allowed_groups
- skip_review_media_groups
- email_in_allowed_groups
- create_topic_allowed_groups
- edit_wiki_post_allowed_groups
- edit_post_allowed_groups
- self_wiki_allowed_groups
- flag_post_allowed_groups
- post_links_allowed_groups
- embedded_media_post_allowed_groups
- profile_background_allowed_groups
- user_card_background_allowed_groups
- invite_allowed_groups
- ignore_allowed_groups
- user_api_key_allowed_groups
This doesn't change anything in 99.9% of cases - by default d-header has only one child element `.wrap`.
There are only two themes that I'm aware of that add another child:
1. discourse-categories-navbar - which I'm currently updating to use the new header APIs, and where the column layout makes more sense
2. a private theme, which was recently updated to use those APIs and that works around the current layout by applying `flex-wrap: wrap` to d-header (I'll remove that override as it conflicts with this change)
To add a components link to the sidebar refactoring was required to create unique URLs for themes and components. Before the query param was used. After changes, we have two URLs `/admin/customize/themes` and `/admin/customize/components`.
Prior to this fix we were manually expanding the composer but not setting the correct height. This commit adds a new `unshrink` function on the composer service to correctly set the state and the height on the composer model.
In discourse-assign the assign menu in the modal is using this `expandedOnInsert` option and was sometimes not opening correctly resulting in a broken state until you click two times on it. This should prevent this issue.
This commit will now change two behaviors:
- If composer is already opened on a specific post and we click on edit again for the same post, we will do nothing and not show the discard draft modal
- if composer is shrinked and we click on edit for the same currently edited post, we will just open the composer and not show the discard draft modal
I needed to run `rake javascript:update_constants` for my previous
commit: 72ac675e4e
The change in this commit though was completely unrelated, so I saved it
for it's own commit.
This fixes a timing issue where, if a user (or the CI) was
on a slow network connection, clicking one of the bookmark
menu options would cause an error because we hadn't yet received
the response from the server after creating the bookmark.
It should be very smooth most of the times because (paraphrasing j.jaffeux):
a) Most likely when user clicks it’s already saved
b) If it’s not saved when user clicks, it should already be almost done so
the perceived wait when click the reminder option should be rather short
The breadcrumbs were updated everytime there were changes to the
categories which was not efficient and caused unnecessary rerendering
of the CategoryDrop elements when "lazy load categories" is enabled.
This commit also ensures that all category fields are serialized for
ancestors too for the categories#search endpoint.
The bulk actions menu for topics has multiple options to work
with tags on topics (append, replace, remove). Our tagging system
along with categories allows for some complicated tag restrictions
to be applied via tag groups. This was a problem for the topic bulk
actions because you couldn't append restricted tags to topics.
This commit allows restricted tags to be used in bulk tagging actions
as long as all selected topics are for a sole category. The category
information will be shown in the modal, and the category ID is used
for the tag search.
Include categories when fetching admin/web_hooks and make
'extras' more useful. 'extras' is the mechanism we use to provide
context for rest objects.
However, previously:
* When you fetched many objects, extras was only set on the ResultSet,
not on each object,
* If you need derived data from extras, there wasn't a sensible place to
put this code. Now, you can create an 'ExtrasClass' static field on
your rest model and this class will be used for your extras data,
This commit changes the identifier displayed in the navigation tree of
the theme objects editor from the generic "category 1" to "<category
name>, <category 2 name>" when a property of typed categories is set as
the identifier.
Example:
For the following theme objects schema:
```
some_setting:
type: objects
default: []
schema:
name: <some scheme name>
identifier: list_of_categories
properties:
list_of_categories:
type: categories
```
If the `list_of_categories` property's value has been set to `category
1` and `category 2`, the navigation tree will display `category 1,
category 2` as the text to represent the object in the navigation tree.
When lazy load categories is enabled, categories should be loaded with
user activity items and drafts because the categories may not be
preloaded on the client side.
This will automatically enable the glimmer header when all installed themes/plugins are ready. This replaces the old group-based site setting.
In 'auto' mode, we check for calls to deprecated APIs (e.g. decorateWidget) which affect the old header. If any are present, we stick to the old header implementation and print a message to the console alongside the normal deprecation messages.
To override this automatic behavior, a new `glimmer_header_mode` site setting can be set to 'disabled' or 'enabled'.
This change also means that our test suite is running with the glimmer header. This unveiled a couple of small issues (e.g. some incorrect `aria-*` and `alt` text) which are now fixed. A number of selectors had to be updated to ensure the tests were clicking the actual `<button>` elements rather than the surrounding `<li>` elements.
This started as a way to prevent "previewUpdated" from doing the same work twice when morphing.
Ended up refactoring "previewUpdated" and extracted into 5 distinct methods for clearer understanding and more consistent debouncing (using the "@debounce" decorator instead of the "discourseDebounce" method).
No "feature" was changed, other than not doing the "decorateCookedElement" when morphing is enabled, since we already did it _before_ morphing.
A lot of plugins are using "api.decorateCookedElement" to decorate the cooked preview.
Some of those plugins, like the "image-grid" ([grid]...[/grid]), changes the DOM. We have to call them _before_ morphing otherwise, there's no real point in morphing the preview.
The missing piece was triggering the "decorate-non-stream-cooked-element" application event.
Chrome's default is already Lax, so this change is a no-op there.
Firefox will soon be follow them, and has started warning about cookies with no samesite attribute. That's the motivation for this commit.
When choosing the "Custom..." option in the new bookmark
menu and then choosing a date + time in the modal for the
reminder, the bookmark icon on the post was not updating to
show the one with the clock to indicate the reminder.
This was just a data syncing issue between BookmarkFormData
and what the modal sets. Ideally all this would be refactored
because the data flow is messy...but hard to find time for
that right now.
Followup 67a8080e33
This commit changes enum typed theme objects property to be optional.
Previously, an enum typed property is always required but we have found
that this might not be ideal so we want to change it.
Display additional confirmation when:
- The public section is going to be updated;
- The public section is going to be deleted;
- The public section is going to be marked as private.
The complexity of the situation is that we don't want to load faker into production by default but fabricators and styleguide are available on production.
This is made possible through app/assets/javascripts/discourse/app/lib/load-faker.js which contains a function to ensure faker is loaded asynchronously (loadFaker) and another function to access the loaded faker (getLoadedFaker).
Note 1: this commit also refactors fabricators to have access to context and use faker where possible
Note 2: this commit moves automation to admin bundle
---------
Co-authored-by: David Taylor <david@taylorhq.com>
It was originally named bootstrap-json because it contacted a 'bootstrap' API in rails to generate the ember-cli html response. However, it has since been overhauled to remove that 'bootstrap' system. Now it is a much simpler proxy server which transforms the HTML sent by Rails.
The bug was due to the fact that the `<DModal />` is displayed inside a if block, when the condition was false to close the menu, the modal was just hidden without calling callbacks. The fix ensures we are correctly calling `modal.close()` before in this case.
This commit adds a new option `@modalForMobile` for `<DMenu />` which allows to display a `<DModal />` when expanding a menu on mobile.
This commit also adds a `@views` options to toasts which is an array accepting `['mobile', 'desktop']` and will control if the toast is show on desktop and/or mobile.
Finally this commit allows to hide the progressBar even if the toast is set to `@autoClose=true`. This is controlled through the `@showProgressBar` option.
This commmit removes the unused `/u/:username/preferences/categories`
route which was merged into the `/u/:username/preferences/tracking`
route in 2fc2d7d828.
Available as a normal synchronous module in tests
Available as an async import in core, or via the `loadFaker` helper in themes/plugins (which cannot use async import directly)
This change adds a progress bar to toast notifications when autoClose is enabled (true by default).
The progress bar allows users to visually see how long is left before the notification disappears.
When hovered on desktop, the progress and autoclose timer will be paused, it will resume again once the mouse is moved away from the toast notification.
Why this change?
For a `typed: objects` theme setting with an enum property, we are
adding a `default` key for `type: enum` fields which will be used
as the default value on the client side.
```
some_objects_setting:
type: objects
schema:
name: field
properties:
enum_field:
type: enum
default: awesome
choices:
- nice
- cool
- awesome
```
Adds the new quick menu for bookmarking. When you bookmark
a post (chat message behaviour will come later) we show this new quick
menu and bookmark the item straight away.
You can then choose a reminder quick option, or choose Custom... to open
the old modal. If you click on an existing bookmark, we show the same quick menu
but with Edit and Delete options.
A later PR will introduce a new bookmark modal, but for now we
are using the old modal for Edit and Custom... options.
When opening the user menu, we display old cached data, and then replace it with fresh data immediately afterwards. The vast majority of the time the data is unchanged, and so there is no visible change. When rendering HTML elements directly, Ember realizes that there is no change, and does not make any changes to the DOM. Great!
However, our `avatar` helper returns a blob of HTML. With raw HTML, Ember does not make any attempt to 'diff' the existing DOM. Instead, it replaces the old string with the new string. That can be a little wasteful, but normally it's not a big deal. But, when it comes to `<img lazy="lazy"`, re-rendering the `img` element causes a visible flicker in Safari.
To work around that, this commit replaces the `{{avatar}}` helper with an ember-rendered `<img` element. Now that Ember is responsible for rendering, it can detect there is no real change to the attributes and skip it, thereby avoiding the flicker.
If we find ourselves doing this more frequently, we may want to consider creating an `<Avatar` component. But for now, I think it's simple enough to justify building the `<img` manually in this case.
The "new topic" route can open the composer with a category preselected.
This commit ensures that the category is loaded before the composer is
opened.
Why this change?
Prior to this change, the input fields were not displaying when adding
an object to a objects typed theme setting which has a default value of
`[]`. This is because the `fields` getter was not being recomputed.
- Add a "Skip tips" button to first notification tip
- Add a "Skip tips" button to the admin guide tip
- Fixes the timeline tip showing when no timeline was present
- Fixes post menu tip showing when no "..." button is present
- Adds system tests
- Marks each tip as seen as soon as it is displayed so that refreshing,
clicking outside, etc. won't show it again
- Change just above means we no longer need a MessageBus track
Co-authored-by: Bianca Nenciu <nbianca@users.noreply.github.com>
Some of the properties, like 'categoriesById', 'parentCategory' and
'subcategories', were updated manually when categories were loaded.
This was not ideal because it required a lot of code to keep the
objects in sync and some of the properties were not updated correctly.
Why this change?
Instead of dealing with a generic object for `Theme#settings`, we want
to always be dealing with `ThemeSettings` objects.` Previously, we
converted the generic objects to `ThemeSettings` objects in the theme's
adapter `afterFindAll` function. This is not correct because updating
or saving the theme individual reverted the `Theme#settings` back to an
array of generic object.
To fix this problem, the proper way with our REST models is to overwrite
the static `munge` method and create `ThemeSettings` instances there.
When a user is manually deactivated, they should not be deleted by our background job that purges inactive users.
In addition, site settings keywords should accept an array of keywords.
Originally we planned to do this rename after dropping the old widget implementation. However, as we continue rolling out the update, there is a risk that people will start depending on the component names (e.g. for modifyClass) so it seems best to make the rename now to reduce risk later.
In this PR, all references in the UI to the word "`upgrade`" are changed to "`update`". This is to differentiate the update process in self-hosted sites from the plan "upgrade" process in hosted sites.
Follow-up to the PR: https://github.com/discourse/docker_manager/pull/208
Prior to this fix the `swipe` modifier could not be disabled and we were not using the `this.dimissable` property to apply/not apply it.
This commit adds a new `enabled` param to the `swipe` modifier, which is used in modals with the value of `this.dismissable`.
Note this commit also adds tests for this modifier.
Why this change?
Before this change, the validation error message shown to the user when
saving a theme objects setting is very cryptic. This commit changes the
validation error messages to be displayed on top of the editor instead.
Note that I don't think this way of displaying is the ideal state we
want to get to but given the time we have this will do for now.
Why this change?
In the categories, groups and tags selectors, we were showing a
validation error message when a property that is not required but
has a min validation is empty. In this case, we should not be displaying
the min validation error message because the property is allowed to be
empty.
Why this change?
When adding a new object, we want to switch to the input fields of the
new object instead of just appending the new object to the list of
objects as we believe this is a better UX flow.
Why this change?
The field components to select categories, groups and tags had quite a
bit of logic duplicated between them. This commit refactors the logic
to remove most of the duplication so that we can introduce changes
without having to make the changes in multiple places.
