If foo list screen has complete foo data, I'd cache them in memory as a

Map<ID, Foo>

, or in SQLite

@oday but from the remite foo list screen I already fetched the data, requesting from network again feels wasteful, so yea maybe as to emulate database, keep it in a hashmap as @yshrsmz said but then, when to invalidate this hashmap? and also process death, youd loose the object, so youd need to rerequest the network..

Hi, I am trying to migrate my app architecture (which is basically Clean Architecture based on RxJava) to Clean Architecture based on kotlin coroutines. There is one thing that I am still struggling is coroutines dispatchers. Previously with RxJava it was common to place main thread and background Schedulers in UseCase and then use them when creating Observable. Right now I see that in some cases developers put coroutine dispatchers in UseCase and then triggering coroutine with specific dispatchers. In other cases I saw that UseCase was made basically just a suspend function and all work related to dispatching was moved to the caller of the UseCase. Any ideas what is more correct way to do this?

is there some sane solution other than codegen?

How do you guys manage keyboard resizing your screen in a single activity architecture during screen transitions? Basically when one screen has a keyboard opened, then moving to next or previous, that given screen's layout will be resized for a brief second due to the keyboard Very annoying, and looks cheap

Now it easier to track AndroidX versions and release notes. 😊 https://developer.android.com/jetpack/androidx/versions

Do you guys think that Jetpack Compose will change current architecture patterns or introduce some new?

anyone using koin? how do you handle cleaning up? e.g. closing open realm instances on closing app?

I have a case where I'm using

BehaviorRelays

with viewModel, the problem is in case of the device rotation,

ViewModel

subscribe again which cause duplicate subscription with each rotation. I know I could save a subscription with

Disposable

and dispose it if it exists, then subscribe .. but is there any other cleaner solution.

Hey! I have some problems wrapping my head around why

by viewModels()

(part of Android KTX) requires a viewmodel factory provider as opposed to a viewmodel factory. I.e. I would prefer to be able to write

val vm by viewModels { MyViewModel(arg1, arg2) }

. Is there anything I’m missing where making the viewmodel factory as explicit as it is now is of great benefit?

Hey, I started to work with Room, first time though, and I'm used to use the Repository pattern and it comes from an usecase, the thing is that I always did an api call, so my usecase called a repository that contains my service, but now, first I want to check : if there's internet connection, if there's not, call the repository from room, and if there is internet connection call the repository from service, do I have to put both on the constructor? I mean, how would be the flow? Presenter sends find X --> I have an use case to find it --> I have a repository to get it from there So, the thing is where I have to do the check for internet connection and if it's a or b do some action? Let's say, should be : UI: User type name of the cat UI -> Presenter with the name of the cat Presenter -> Have an use case that it's GetCatUseCase that I have to put a parameter that it's the name GetCatUseCase -> On my old projects I just call the service, but now sometimes I'll have to call the repository from roomdb and other repository from service... Any ideas how to organize it?

Do I have to have 1 repository and then an interface one method to go to service and other to go to room?

R.string is an int, it doesnt change based on localization

no I dont, maybe you should look at dependency inversion

Whether or not R is part of Android framework is totally irrelevant. You have to ask yourself what you are actually trying to achieve with your architecture instead of following arbitrary rules for their own sake. Mostly we are concerned with testability since reusing ViewModels on a different platform never happens anyway. Regardless, I would say that using resources in the ViewModel always warrants a closer look whether you are mixing concerns

Hey everyone, currently I am facing an issue with the "new" Navigation Components? Whats the correct way to set different toolbar actions for different fragments while using a single activity. Currently I am using get/set methods in the activity to access the toolbar from each fragment and remove/set the correct menuitems. It works but seems wrong imo. Does anyone have any cleaner ways to solve this?

your state data would look like this (1st page list, idle status) (1st page list, started status) (2nd page list, started status) <--- this is a problem (2nd page list, success status)

that feels kind of weird; should I maybe map it? i.e.

State(val message: List<Message>, ...)

->

ExternalState(val items: List<Item>,...)

Should I make a repository as an

Object

or

Class

Hi, I have a margin item decoration which overrides getItemOffsets and for simplicity lets say adds bottom margin X to every 3rd item, and some smaller X/2 margin to others. So my issue is that without invalidating the item decoration, views keep these margins, and when animation recyclerview via notifyItemInserted and friends (via DiffUtil), it causes wrong margin (not big margin every 3rd item) So I google'd, I need to invalidate like this

adapter.items = diff.list
diff.result.dispatchUpdatesTo(adapter)
recyclerView.invalidateItemDecorations()

Which does make the margins correct, however this doesn't get calculated into the animations*, and view jump to some incorrect intial position and animate to their end position* Has anyone encountered this? Its very annoying. Do I need to write my own ItemAnimator?

Some folks from iOS has been talking for some time about the popularity of MVVM-Coordinator or just Coordinators Architecture in iOS. Has anyone tried a similar approach to tackle navigation logic in Android?

Hey all ! I’ve recently open sourced an example modularized app - using some of most common libraries, tried to implement cases like structured DI, clean architecture, navigation between features etc. ( you can see modularization graph & DI graph on README ). I’ll be writing about many things I’ve came across and learned during development when I have enough time ! I’m trying to improve it with more cases to cover, not going too fast, but still it’s something. If you have any suggestions, it’d be great hear ! https://github.com/melihaksoy/Android-Kotlin-Modulerized-CleanArchitecture

Does anyone know if flow extensions for

DataSource.Factory

already exist as a package? If not, is there somewhere that makes sense to contribute them?

I'm happy to share the live sample app for the Android Unidirectional Data Flow (UDF) pattern, featuring Coinverse. Code in the wild is always better to learn from than slide samples. Coinverse’s Open App showcases a live UDF pattern which makes the app more readable, easier to develop for, and debug. Next step - JUnit testing the ViewModels! 🧪🔬🥼 https://buff.ly/2T9Yyw3

We are trying to understand how viewmodels and recyclerviews can work together. In our case we have a custom view, used in an activity, with its own viewmodel. Now this custom view need to be part of the layout of items in a recyclerview. That means each ViewHolder will have a ViewModel attached to it. We think it’s not a good solution but we’re wondering what would it be a way to structure this type of situation. ViewModel attached to the adapter (we think it’s bad)? use DataBinding to connect VH and VM? we don’t have databinding set up in the project though. What do you guys think?

What’s the best way to retrieve LiveData value in a click listener? Is it ok to just perform a null check to synchronously get the livedata value?

button.setOnClickListener {
     if (viewmodel.getMyLiveData().value != null) {
         // do something
     }
}

Hey all ! I was wondering how do you structure your project to apply dependency inversion between modules - it seems to end up in low modules depending on high ones for sake of inversion - but I was wondering if this is correct approach in general, or something else ?

Hi guys! I have a question regarding MVVM. Let’s say I have a ViewPager that has 3 pages and has automatic scrolling feature. (this viewpager goes back to page 1 when reaching page 3). The auto scrolling feature is PAUSED everytime the viewpager receives a touch action. Now, should the logic of setting view pager’s current item to 1 if it reaches 3, and the pausing of auto scroll when view pager receives touch event RESIDE in ViewModel? In my understanding. Purely UI logic can stay in activity/fragment. Logic that can reside in ViewModel can include enabling/disabling buttons (depending on the input), validating inputs, etc. Any inputs are greatly appreciated! 🙂

Hi, has anyone tried to automate generating DiffUtil.Callback methods? Mostly `areContentsTheSame`and

getChangePayload

) Say I have a model, the usual case fo areItemsEqual is to use equals(), but what if I dont want to compare all the fields? So, I could write a reflection util / codegen for this, but it requires to have some sort of a @Exclude annotation on a given field *which in turn means that our data/domain layer Foo model knows about diff util stuff so not sure if this is the way to go* should it maybe be manual code and keep reflection (traversing all the fields) unit test to check if new fields are added) (edited) -- or, create a UI layer level model from it, and use just the plain equals from it (or not, doesnt matter, since now diffutil annotations make sense)