userManager.currentUser was called too often. I was not able to prove that a bug is related to it but i think it may fix some hidden bugs.
CurrentUserProviderImpl is now used throughout the code to access the current user.
userManager.currentUser is only used from CurrentUserProviderImpl whenever the _currentUser was null (should only happen on app startup)
To avoid multiple initialization of CurrentUserProviderImpl it was changed to be a @Singleton
The handling should soon be replaced with coroutine flows. However for the v21.0.0 release it's still done with RxJava to avoid bugs.
Signed-off-by: Marcel Hibbe <dev@mhibbe.de>
- Uncommented out fetchOpenConversations
- Added fetchUsers to conversationList
- Made the messages appear first
Signed-off-by: rapterjet2004 <juliuslinus1@gmail.com>
It could happen that when sending a message it was received on server but if the servers response is not received, the retry was triggered in the talk app.
This sometimes happened when internet connection was not the best.
Best would be that messages with the same referenceId would be refused on server side, but this won't be the case for now. So messages with the same referenceId are NOT refused as this would be too much overhead as there would be additional queries on server for every received message)
For now, the automatic retry logic is just removed so duplicated messages won't be created automatically.
However it's still possible to manually trigger the retry via button. In this case it is not guaranteed that there won't be duplicates.
Signed-off-by: Marcel Hibbe <dev@mhibbe.de>
usersDao.getActiveUser() will return only one current users even if multiple exist for whatever reason.
By executing setUserAsActiveWithId(it.id) with the same user, it is made sure all other users current status is set to 0.
This change won't fix the root cause if multiple users are set to current, but it will make sure the state is fixed as soon as there is a query to get the active user.
In fact, this change might also make it harder to find the root cause because debugging may be more difficult! When searching for the root cause, always keep this in mind and maybe revert the change.
Signed-off-by: Marcel Hibbe <dev@mhibbe.de>
To make sure getAnyUserAndSetAsActive() is invoked lazily and avoid any side effects, it's explicitly wrapped it in a lambda.
Maybe.defer ensures that getAnyUserAndSetAsActive() is not invoked until switchIfEmpty decides it’s needed.
Signed-off-by: Marcel Hibbe <dev@mhibbe.de>
Before, getAnyUserAndSetAsActive did only set a random first user to current=true without to check if there is any other current user.
So it was up to the calling function of getAnyUserAndSetAsActive to check this.
I did not identify a scenario where the getAnyUserAndSetAsActive could cause to set multiple users to current, but anyway the new implementation might fix some scenario that i could not reproduce.
Signed-off-by: Marcel Hibbe <dev@mhibbe.de>
- First commit - Made options yellow
- Now possible to archive/unarchive conversations from dialog
- Now possible to archive and unarchive conversations in settings without leaving the screen
- Better UX
- Unread message bubble fix - no longer shows up when archive filter is set
Signed-off-by: rapterjet2004 <juliuslinus1@gmail.com>
when being offline and no messages are set -> hide shimmer
when messages are received -> hide shimmer and "offlineNoMessages"-info
Signed-off-by: Marcel Hibbe <dev@mhibbe.de>
Because long polling may have to be informed that the scopep was cancelled, the isActive checks are added so it can't happen that messages are added when they shouldn't.
I could not reproduce the scenario, anyway the checks should make sense in my opinion.
Signed-off-by: Marcel Hibbe <dev@mhibbe.de>
This is necessary especially to cancel the long polling when configuration change was made, e.g. screen was rotated. Otherwise multiple long polling requests would be running after configuration changes.
Because it not possible to launch a new coroutine in a scope that was canceled, it is necessary to re-initialize the scope.
Signed-off-by: Marcel Hibbe <dev@mhibbe.de>
if chatMessageList is empty after handleSystemMessages it makes no sense to call the following methods.
Also processMessagesFromTheFuture was executed which caused that the popup was shown.
A better solution for the future should be to handle(remove) the "to-hide" system messages already in the repo or viewmodel
Signed-off-by: Marcel Hibbe <dev@mhibbe.de>
The idea to show the progress bar (aka shimmer animation) was that it show not be shown at all if loading messages happens in less than a second (this was a better UX in my opinion).
The idea was to hide the progress bar when
ChatMessageStartState was triggered.
However there can be moment when adapter is still empty after ChatMessageStartState and if in this moment the DELAY_TO_SHOW_PROGRESS_BAR is reached before the adapter is actually filled, the overlay happens.
It could be a solution to move the hiding of the progress bar, however then special cases might have to be handled.
For simplicity, the logic for DELAY_TO_SHOW_PROGRESS_BAR is removed. Progress bar is always shown without a delay so it wont be triggered on a later moment and can't overlay the chat.
Signed-off-by: Marcel Hibbe <dev@mhibbe.de>
