Viewless, Action Driven Architecture

Getting

You can install this module using npm with:

npm install vada

User's Guide

Routing

To create routes (or more specifically, RouteId instances), just declare them as follows:

export const main = new vada.RouteId<{}>("main");
export const search = new vada.RouteId<{term: string}>("search");

Background

While both migrating existing applications and creating new ones, incorporating redux and react into my applications has pushed me toward an approach that takes a hard line in the separation of view from application logic.

My main motivation was around making testing as easy as possible. My goal was to create applications where the view was completely external to the application (hence the term "viewless"). Following this approach, the flow of the application is entirely described via a redux store and actions dispatched against that store (hence the term "action driven").

Another motivation behind this module, which is probably not of much interest to the larger Javascript development community, was to provide type constraints around many of these patterns. So if you use TypeScript, you can benefit from my efforts to include as many type constraints as possible. But the module should still be useful for ordinary Javascript developers as well.

This repository implements several patterns that I found useful in trying to achieve these goals. There are three main pieces of functionality that this library provides...

Routing

The first challenge I ran into was how to provide routing without involving a view. Libraries like UI-router and react-router are widely used and well engineered approaches to routing. However, they both integrate with the view to provide their functionality. So I wanted to create an alternative approach. There are several viewless routing tools and I've chosen to build my approach on top of crossroads. But the contribution of this library is to connect crossroads with a set of actions that can be used to manipulate application state contained in a redux store so that the entire collection of routes associated with an application can be completely external to the view.

My initial goal with this approach was to migrate code from angular to react by externalizing all the routing specific code from any particular view framework. But just as important, this view agnostic approach means that applications can be written independent of a particular view framework. This means that the same application code could be used across desktop, web and mobile applications. In other words, by keeping as much logic as possible out of the view, it makes it easier to reuse application code between web and mobile versions.

It should be noted that if you are working with TypeScript, these implementations also provide some type constraints that help to ensure that route parameters can be statically checked at compile time.

Reactors

In trying to improve application testing, I wanted to see how far I could push logic out of my view. My goal was to be able to go back to doing testing with simple tools like mocha instead of having to build up tests that used tools like PhantomJS, etc. to actually traverse my UI. I find the UI design is too fluid to build testing on. But the underlying application structure can be relatively invariant.

The obstacle here is that it is so common and relatively convenient to embed logic in a view. But I think this is actually unwise for many reasons. Not only does it mix view and application logic in such a way that the view must become part of the testing. It also makes it difficult to modularize the application.

So I wanted to get the application logic out of my view. But the question was...where to put it? What form should this logic take and how can I incorporate it into applications? I was inspired by this article on Actors. But I found the implementation to be problematic. So I came up with a similar approach that I called reactors. The re prefix applies on several levels. First, this pattern provides a way to react to changes in the application state unilaterally. Second, the functionality itself is not associated with the redux store, but rather with the reducer. Finally, I find that it really helps keep the actions simple and reusable while providing a way to extend the application with additional functionality.

Operations

I think redux is a great way to incorporate state monads into Javascript applications. I think @gaeron has done a great job of keeping the API clean, simple and composable. What I've tried to do with the concept of operations is to build on those basic abstractions the ability to create both actions and reducers in a way that is more natural to me. In a typical redux application you'll find both a reducers.js file and actions.js file. One defines the identities of the actions and other provides the effect of the actions.

I find it a bit confusing to keep these two things segregated. So I created a reducer implementation that allows me to combine the actions and effects into what I call an operation. This allows me to define everything in one place. I've also incorporated a scheme for generating unique action names so as to avoid action naming conflicts. Finally, if you are working with TypeScript, of providing some nice type constraints for type checking action payloads.

Additional Functionality

This module represents the core functionality for creating an application where logic is decoupled from the view. As such, to connect a vada application to a particular view, some additional functionality is required. The following modules build upon the functionality defined by vada.

• vada-react - Functionality to help syncronize react based applications to vada application state.

• vada-browser - This module is for browser based application and it ties the browser's current window.location to routes defined using vada.

TODO

More tests, more docs, more feedback.