F# 9 nullability migration

[xperiandri]

High-level description
I've migrated my code to F# 9 compiler and nullability support

Expected or desired behavior
You can pick my changes if you want the same migration
Ecierge/Elmish.Uno#1

Additional Information
The other question is should we restrict a model and a message on a view model to be non-null?

type DynamicViewModel<'model, 'msg when 'model : not null and 'msg : not null>
      ( args: ViewModelArgs<'model, 'msg>,
        bindings: Binding<'model, 'msg> list)
type ViewModelBase<'model, 'msg when 'model : not null and 'msg : not null>(args: ViewModelArgs<'model, 'msg>)

https://github.com/Ecierge/Elmish.Uno/blob/be64c50b6b7cca1ac774704145e09fd17ee4cf3f/src/Elmish.Uno/ViewModels.fs#L143
https://github.com/Ecierge/Elmish.Uno/blob/be64c50b6b7cca1ac774704145e09fd17ee4cf3f/src/Elmish.Uno/ViewModels.fs#L359

[xperiandri]

@TysonMN @marner2 @YkTru your thoughts?

[TysonMN]

I expect most things should be not null. Maybe everything.

Do you want to contribute a PR with this change?

[YkTru]

Just to clarify, when you ask:

The other question is should we restrict a model and a message on a view model to be non-null?

Is the primary benefit of F# 9 in this context limited to getting compiler warnings and clearer communication of the API intent? Are there additional advantages beyond these?

I think opening a PR would be relevant too

[xperiandri]

The main feature is nullable reference types same as in C# that allows to propagate that from .NET SDK and other libraries to F# code

[YkTru]

Yes, I understand that. However, I'm wondering what additional benefits this approach provides compared to Option.ofObj or box f = null, aside from improved syntax, clearer intent, and compiler warnings.

Personnally, the main nullability issue I face with static VM is with SubModelSelectedItem, especially in recursive data structures. It often fails to find the member collection's name due to the stringly-typed parameter and the lack of compiler support. Additionally, it must reference a ViewModel instance that may not exist at the time of access.

Do you think using a Nullable type could help with these core issues? I feel like SubModelSelectedItem helper should be completely redesigned from the ground up, if possible.

[YkTru]

@xperiandri, could you open a PR, please? It's been a while since the library was last updated, and this one seems relevant. Thanks

[BentTranberg]

I have TreatWarningsAsErrors true in all my F# source, and then only turn off a handful that would otherwise be extremely annoying. I mention this because "getting compiler warnings" was mentioned. So for me, it's almost all errors, which is phenomenal IMHO. I greatly appreciate all the safety and good guidance I can get from the compiler, and this recent addition of nullability support has been an enormous help. There are unfortunately some libraries from NuGet that haven't implemented it, but I am surprised by how many there are that have.