What is a decent strategy to refresh a FutureProvider when disposing another one?

[matuella]

Let's say that we have two providers: providerA and providerB. There are some scenarios where after disposing of providerB, I wanted to call ref.refresh(providerA). If I'm not mistaken, there is only one way to know when a provider was disposed, and it's on the create function, onDispose. The problem is that when I call the ref.refresh on the dispose method, that particular ref is already disposed of, meaning it'll throw an error.

Thus, my question: how can I listen to some provider being disposed, from another reference?

[rrousselGit]

There are some scenarios where after disposing of providerB, I wanted to call ref.refresh(providerA)

That sounds fishy. What exactly is this scenario? You most likely don't want to do that.

[matuella]

Knowing when a provider is (or not) alive is also something that would be valuable in this situation, but it's also not supported.

[rrousselGit]

Now, when disposing of providerB, I wanted to refresh the providerA because changes on this second screen impacts on the previous one.

Why does this needs to be done "when providerB is disposed"? Isn't there a better place to do that? Such as when the second screen is poped

[matuella]

Sure, but we could argue that, let's say this second screen is popped in multiple possible ways, not necessarily in just a single "point", I would have to always go there and imperatively call the providerB to reload itself. This also opens problems for things like: "when a new "pop" method is added, I'd have to remember to also call all depending providers to refresh", which is a design flaw for me.

Having the ability to observe events like "disposals" and "being/not being alive" are useful for side-effects, but I'm unable to achieve a reactive approach with how things are structured.

I'm not saying that all things SHOULD be observable, it's just that, IMHO, it makes sense to be able to react to all of the provider lifecycle.

[matuella]

Sorry to bring up this discussion again, but what do you think about the current state of riverpod and the observability of a provider lifecycle?

[rrousselGit]

I don't get the question.

But for your request, that's not something allowed.

Providers are voluntarily restricted in various ways. This is a requirement for making them maintainable.

Refreshing a provider when another one is disposed is the wrong way to architecture things.
Rather, have the former watch the later.