Need help with remx?
Click the “chat” button below for chat support from the developer who created it, or find similar developers for support.

About the developer

152 Stars 16 Forks MIT License 238 Commits 9 Opened issues


Opinionated mobx

Services available


Need anything else?

Contributors list

remx Build Status

Opinionated Mobx


Writing a react / react-native project with or without TDD, we have multiple well known and battle tested choices when it comes to state management:

  • local component state
    • Pros: Easy to implement, zero boilerplate, works out of the box
    • Cons: In our opinion not very good in terms of scale, component becomes a God Object, very hard to test.
  • redux
    • Pros: is a design pattern, scales very well, individual parts can be tested.
    • Cons: lots of boilerplate, hard to test-drive (TDD), tests tend to be tightly coupled to the redux library, state has low cohesion
  • mobx
    • Pros: Easy to implement, almost no boilerplate, performant, high cohesion
    • Cons: Can be used in lots of different ways, In our opinion complex API due to large amount of features.

When we look at those battle-proven solutions, and especially if we try to build a large scale project that can be worked on by lots of different people from different teams, we see that we have a need for some combintaion of the above. Let's try to break down our actual requirements and see if we can build a system that will answer those.

What do we really need from a state management library?

Well for starters, starting to write our business logic by doing TDD does not require any state management at all. In fact, I would argue that it can only hinder. Try that for example: start by writing test-first all features of a JS(node) application with unit tests without any mention of redux. It is really difficult: if all we want to do is, for example, write a unit of logic that does some login flow, with loading state and authentication and error flows, we see that we have little use of any of these libraries.

We can just test-drive some

with regular functions/objects, call those functions from
and assert on some exported state. That's it. No frameworks needed.

redux itself is really just an implementation of flux. Which itself is a close relative of MVC. We can take this design pattern and use it to build a scalable architecture, we just don't need redux itself here yet as it doesn't provide anything new and just introduces coupling and boilerplate.

So one thing we do want to enforce is some data-flow architecture. Still, no frameworks required.

Well, then where does state management come into play?

When we want to connect this

into a react view. We want this view to listen to changes and re-render itself according to some presentation logic (which will be test-driven as well).

For our example, we can think about some

that displays the login username and password, any sucess/failure message, and maybe a loading indicator.

All we need to do is somehow connect those 2 together.

We can't really use redux here because it will require us to rewrite everything, and will make our business logic and tests tightly coupled to redux itself.

We can however use mobx. If we figure out a way to wrap all of our business logic with mobx's

, and our component as
, we will achieve our requiremnents.

But doing that is not as simple as wrapping everything with mobx. One major thing mobx does is impact all objects that it touches, in order to observe on their state. We want to avoid that, where possible, and es6 proxies allows us to do just that.

Furthermore, we don't need the large amount of features mobx supports. We need a small subset of the api, and we need to be very explicit about what goes where, to enforce our chosen redux-like data-flow architecture.

These decisions are what gave birth to remx. A mixed child of mobx implementation of observable state (backed by es6 proxies) with flux/redux design pattern enforcement. An opinionated mobx.

But what about dispatch? actions? middlewere?

In general, redux's middlewere is something we want to avoid. Previous experience taught us that adding just 1 middlewere to a redux store can cause serious performance hit, so we recommend not to do that and just call a function explicitly. Otherwise, if we really want, we can create our own global function that wraps any other function with some logic, for example a logging function. This can be done easily without any framework.

Actions is where we put our imperative business logic, and so we call them by simply invoking them (with arguments if needed). Actions can be asynchronous, and shouldn't return anything (to enforce uni-directional data flow). We don't need any dispatching function because our stores are just plain old JS objects (that we test separately). So although not a part of the api (there's really nothing special about

anyway), we encourage the separation of actions and stores for low coupling, and to put action files next to the same use-case store files, for high cohesion.

Take a look at the example project to see how remx is intended to be used.

To conclude:

  • remx takes the redux (flux) architecture and enforces it through a short, simple, clean and strict API:
    • state
    • setters
    • getters
    • connect
  • almost zero boilerpate
  • zero impact on tests
    • can be added/removed as a plugin
    • does not impact any design decisions
  • implemented with mobx, thus benefits from all the performance of
    • memoization
    • avoids unnecessary re-renders
  • uses es6 Proxies (where possible)
    • avoids mobx's Observable wrappers which can cause weird bugs and behaviours



The state function takes a plain object and makes it observable. The state should be defined inside the store, and should not be exported. All the interactions with the state should be done through exported getters and setters. Any change to the state will trigger a re-render of any connected react component that should be effected from the change. If for example you have a state with two props, A and B, and you have a connected component that is using only prop A, only changes to prop A will triger re-render of the component.


: ```javascript import * as remx from 'remx';

const initialState = { loading: true, posts: {}, selectedPosts: [], };

const state = remx.state(initialState); ```


All the functions that are going to return parts of the state should be wrapped within the Getters function. The wrapped getters functions should be defined inside the same store file and should be exported.


import * as remx from 'remx';

const getters = remx.getters({

isLoading() { return state.loading; },

getPostsByIndex(index) { return state.posts[index]; }


export const store = { ...getters };


All the functions that are going to change parts of the state should be wrapped within the Setters function. The wrapped setters functions should be defined inside the store and should be exported.


import * as remx from 'remx';

const setters = remx.setters({

setLoading(isLoading) { state.loading = isLoading; },

addPost(post) { state.posts.push(post); }


export const store = { ...setters };

Makes component re-render when store data used during previous render changes.

import { connect } from 'remx';

class SomeComponent extends React.Component { render() { return (

); } }

export default connect(SomeComponent);

Also works with functional components:

import { connect } from 'remx';

export default connect(props => (



Connects a react component to the state. This function can optionally take a mapStateToProps function, for mapping the state into props. in

import React, { PureComponent } from 'react';
import { connect } from 'remx';
import { store } from './someStore';

class SomeComponent extends PureComponent { render() { return (

); } }

function mapStateToProps(ownProps) { return { selectedPostTitle: store.getPostById(ownProps.selectedPostId); }; }

export default connect(mapStateToProps)(SomeComponent);

remx.useConnect(fn, arguments)

Hook-style alternative to It makes sure, the component is re-rendered on observable values change. Second argument (optional) is array of arguments that will be passed to the provided function.

import React, { PureComponent } from 'react';
import { useConnect } from 'remx';
import { store } from './someStore';

const SomeComponent = (props) => { const {selectedPostTitle} = useSomeComponentConnect(props);

return (

); }

const useSomeComponentConnect = (props) => useConnect(() => ({ selectedPostTitle: store.getPostById(props.selectedPostId); }));

export default SomeComponent;

Alternative style:

const SomeComponent = (props) => {
  const selectedPostTitle = useConnect(store.getPostById, [props.selectedPostId]);

return (

); }

Note that accessing props outside of mapStateToProps won't be tracked and may cause issues with components not being updated.

// Bad (product.price accessing is not tracked):
const ProductPriceComponent = (props) => {
  const {product} = useConnect(() => ({
    product: store.getters.getProduct(),

return (

Price: {product.price} USD
); }

// Good: const ProductPriceComponent = (props) => { const {price} = useConnect(() => ({ price: store.getters.getProduct().price }));

return (

Price: {price} USD
); }


Enables writing in console.error (in DEV mode) on accessing state in React components which are not wrapped in

, so they may not re-render automatically when needed.


Takes a logger and call it on the following actions:

import {registerLoggerForDebug} from 'remx'
registerLoggerForDebug(console.log); //will log all remx actions:
//on setter call: {action: "setter", name: "someSetterName", args: ["arg1", "arg2"])}
//on getter call: {action: "getter", name: "someGetterName", args: ["arg1", "arg2"])}
//on mapsStateToProps: 
      "triggeredEvents":[] // an arry of actions that have been triggered during the run of mapStateToProps

//on component re-rendered: {action: "componentRender", name: "SomeComponent"}

We use cookies. If you continue to browse the site, you agree to the use of cookies. For more information on our use of cookies please see our Privacy Policy.