rx-angular VS neo-observable-input

RFC: @rx-angular/state/signals - extended signal and new eventEmitter

It has been a while now since the Angular team has been exploring (way more than we can think) alternative reactivity models and looking for something that lies between the extremes of naive Zone.js (i.e. Zone.js without OnPush) and Zoneless Angular combined with special pipes & directives like those provided by RxAngular.

I would recommend exploring https://github.com/rx-angular/rx-angular. Contributors did a huge research about using Angular to get the best results.

It also includes RxAngular, a toolset focused on runtime performance and template rendering, and taiga-ui, an UI component library.

There is another reason not to use async pipes. If you provide 'noop' zone, async pipes will stop working. In my experience it's better to start without zones if you know that performance might be a problem in the future. Working without zones has more benefits https://github.com/rx-angular/rx-angular/blob/main/docs/general/zone/why-zone-less.md Here is the list.

Another cool lib which goes into the same direction as NgRx Component Store: https://github.com/rx-angular/rx-angular/blob/master/libs/state/README.md

Here are some examples: https://github.com/rx-angular/rx-angular/blob/master/libs/state/docs/usage.md

We also added listings using the same pattern. This is the first time we used *rxLet. It really makes working with all Observable in the templates easier do check it out. It allows you four states for an observable suspense, error. complete so that you can handle each one of them separately and gets rid of the *ngIf with async pipe. Like if you want to display something else while loading and error is a big disadvantage of the async pipe when used with an api.

Regarding input being observables. No, my inputs are not observables. They are declared as simple properties (aka arrays, objects, primitives - not observable) mainly for two reasons: 1) in the first version of the component they didn't need to be observables at all, so changing their type now has a very high refactoring cost and 2) from a semantic point of view, they should not be observable from the outside: you should be able to pass a simple static array\object (think about tests, for example). The observables are generated from the inputs with one of the many solutions (something like this).