atlas-munchlax
Reactive variable and autorun library in 15 lines of code -- inspired by Meteor.
Munchlax gives you reactive variables, computed variables, observers, and autoruns:
const val comp obs = ; // reactive variables with initial valuesconst firstName = ;const lastName = ; // reactive variable with no initial valueconst nickName = ; // setting values (send a new value) // getting values (ask for current value)const currentValue = // derived/computed reactive variableconst initials = ; // reactive "autorun" that depends on `nickName` and `initials` // scoped reactivity via nested computations // reactively "observe" render functions (like in Mobx)// shares all the same properties as reactive computations// * can nest comps and children can be observed, too// * scoped reactivity, no redundant renders, etc.const App =
If you read the source code, you might notice that this library is (relatively) trivial compared to MobX or Meteor.Tracker. That's because Relax is doing all of the heavy lifting -- we've reduced the task of state management to picking a pattern we'd like to use! All of these patterns can be implemented on top of Relax's lego-like primitives. The key here is that the same engine for VDOM/graph diffing and reconcilation can easily power both UI frameworks and state management frameworks.
concepts
Everything in Munchlax is either a reactive variable or a reactive computation (an "observed" function). If a reactive computation returns a value, then that value is a "derived" value, and can be depended on by other reactive computations as if it were a reactive variable. Reactive computations are just functions which track ("observe") the values they depend on. If an underlying value changes, the computation is rerun efficiently and atomically.
getting/setting reactive variables in other frameworks
- Meteor: Meteor has a very explicit syntax for getting/setting reactive variables. First, you create the variable with
const name = new ReactiveVar("atlas")
, then you can get it withname.get()
and set it withname.set("jai")
. This is a bit verbose. - MobX: MobX is on the other end of the spectrum. You initialize instance values with
@observable name = "atlas"
, then you can get them by callingthis.name
and set them withthis.name = "jai"
. This is too implicit and obfuscates reactivity intent -- there are too many gotchyas for beginners. - S: S introduced an API that is half-way between the above two. To create values, you use
const name = S.data("atlas")
, to get them you doname()
and to set them, you send the variable a new value withname("jai")
.
Munchlax keeps things explicit and terse by using S's syntax.
just relax!
Munchlax is just a taste of what you can build with Relax in a few lines of code. The idea behind Relax is that the same code for a view framework can power a full-blown state management framework, since they use the same structures and algorithms under-the-hood.
Don't think of Munchlax as a state management library. Relax is the state management library here. Munchlax is just a set of tiny helper functions for Relax that expose a terse API you might be more familiar with.
advanced
With Munchlax, you can batch updates (powered by Relax) and you can provide equality functions to reactive values that tell them when not to change. The unit tests will show you how to use these features.
install
npm install --save atlas-relax # peer dependency
npm install --save atlas-munchlax
obs
vs. comp
A comp
is like MobX's autorun
and @computed
APIs in one function. It is similar to Meteor's Tracker.autorun
as well.
obs
is basically the same as comp
, except it lets you "wrap" render functions and automatically turn them into reactive autoruns. The obs
implementation is 3 lines of code. It's so short because it's an API wrapper for something you can already do with Relax.
Relax is a state management library, but it is not opinionated. It provides the necessary primitives for you to form your own opinion. Munchlax is merely an opinion, and it's for the people who love Meteor and MobX. If you aren't one of those people, don't worry -- Relax makes it easy for you to use a different opinion. Hopefully you will feel at home with val
, comp
, and obs
:
import val obs from "atlas-munchlax";import t from "atlas-relax-jsx-pragma";/**@jsx t */ // global "sideways" dataconst isDarkMode = // every instance is automatically reactiveconst MyDashboard =
Instead of importing an entire reactive framework like MobX, we get most of the power of MobX with a 3 line helper function.
obs
has a few caveats, but comp
solves them all:
-
obs
can't wrap classes (or constructors) yet:obs(class MyApp extends Frame {...})
. I'll implement this if people start using this, or feel free to open a PR. The basic idea is to rewrite the render prototype method for the class with a wrapped method. -
obs
hasn't been defined to take non-render functions. This is by design.obs
is specifically designed to decorate render functions (i.e. functional components). -
obs
returns a functional component. Just as you are never supposed to instantiate components directly (const dashboard = new MyDashboardComponent()
), you are also not supposed to manually call the return function fromobs
, since it's a decorated component. Again, this is by design.
If you find yourself needing these things, then you should use comp
, since it was designed for these use-cases.
credits
GhostyDoesStuff made the "Munchlax in Sunglasses" image.