A modern programming language that makes developers happier.

kotlinlang

Hey guys, i was wondering why we can't inject the Viewmodel directly, but instead we've to inject a factory. Which ate the benefits came from this practice?

Why can't you inject view model directly? I do it all the time

The benefits, in theory, is decoupling the ViewModel creation to the view model usage

Because if you ever want to change how the viewmodel is created, you only need to change the factory, and not the classes that create a view model

Suppose you have 20 classes that inject `new ViewModel()`

If you change that to `new ViewModel(foo)`, you'll have to change 20 classes

If you use the factory, you'll have to change only the factory

But this isn't a rule, you aren't obligated to comply with every single bit of clean architecture/clean code

If you inject ViewModel then it will be recreated on every configuration change

point of the factory is that it can create new VM, but it doesn't have to (it can reuse old one)

Hmm i get it. But we had a problem few days ago with it. We were testing a Video feature and the tester rotated the screen, this action recreated the VM.
We tested binding a scope in it, but same problem

I don't think it's recreated necessarily. It depends on how you inject the vm

For instance, with `Koin` it won't be recreated

Koin better than Dagger hahah. I'm joking guys, please do not kill me :disappointed:
We were using Dagger by the way

we inject the factory via dagger and then ViewModelProviders does the rest

factory is created by dagger so there is no extra manual work

Hmm thanks <@UJBC14H8A> and <@UB9K6R4JH>. One last doubt, do you have an example how do you implement this factory? The way that we're doing the Factory is being recreated every time that he is injected.
We're using this tutorial as guide: <https://android.jlelse.eu/android-less-dagger-boilerplate-code-for-viewmodel-creation-9fec91a27bf3>

If you are using this tutorial, then factory is not recreated every time since it's `@Singleton`

Hmm, we're implementing the second way. I'm going to discuss the first way with my team and consider the tradeoff. Thanks <@UJBC14H8A>

We are actually doing it the second way too

it is much simpler and allows for parameters

I don't think one object created per activity is that big of a deal

Hmm, we're considering to use Lazy initialization to solve this kind of problem.

Hmm one point to consider. We were thinking in fragment case, but we can retrieve using by viewModels. Howsoever we are going to use as described in tutorial.

I think you are prematurely optimizing this. It's one allocation per activity creation, it's not big enough deal to warrant such hacks.