Home

Awesome

version Tests

React Recollect

What?

Recollect is a state management library for React, an alternative to Redux.

Why?

Recollect aims to solve two problems with the traditional React/Redux approach:

  1. Immutability logic is verbose, complicated, and prone to bugs.
  2. Developers must write code to define which parts of the store a component plans to use. Even then, components can be re-rendered when they don't use the data that has just changed.

How?

  1. The Recollect store is immutable, but the implementation is hidden. So, you can interact with the store as though it were a plain JavaScript object.
  2. Recollect records access to the store during the render cycle of a component. When a property in your store changes, only components that use that property are re-rendered.

The result is simpler code and a faster app. Take it for a spin in this CodeSandbox.

Caution: there is no support for any version of IE, Opera mini, or Android browser 4.4 (because Recollect uses the Proxy object). Check out the latest usage stats for proxies at caniuse.com.

Quick start

npm i react-recollect

The store object and the collect function are all you need to know to get started.

The store is where your data goes; you can treat it just like you'd treat any JavaScript object. You can import, read from, and write to the store in any file.

Here's some code doing normal things with the normal-looking store:

import { store } from 'react-recollect';

store.tasks = ['one', 'two', 'three']; // Fine

store.tasks.push('four'); // Good

store.site = { title: 'Page one' }; // Acceptable

Object.assign(store.site, { title: 'Page two' }); // Neato

store.site.title += '!'; // Exciting!

delete store.site; // Seems extreme, but works a treat

store = 'foo'; // Nope! (can't reassign a constant)

Play with this code in a CodeSandbox

These operations behave just like you'd expect them to, except none of them mutate the store contents. In fact, it's impossible to mutate the data in a Recollect store.

Next up: the collect function. This wraps a React component, allowing Recollect to take care of it. This will provide the store as a prop, and update the component when it needs updating.

Here's collect and store working together:

import { collect } from 'react-recollect';

const TaskList = ({ store }) => (
  <div>
    {store.tasks.map((task) => (
      <div>{task.name}</div>
    ))}

    <button
      onClick={() => {
        store.tasks.push({
          name: 'A new task',
          done: false,
        });
      }}
    >
      Add a task
    </button>
  </div>
);

export default collect(TaskList);

Congratulations my friend, you've finished learning Recollect. I am very proud of you.

Go have a play, and when you're ready for more readme, come back to read on.

If you've got a question, make sure to read the FAQ to see if your Q is FA. Otherwise, open a GitHub issue.

<!-- START doctoc generated TOC please keep comment here to allow auto update --> <!-- DON'T EDIT THIS SECTION, INSTEAD RE-RUN doctoc TO UPDATE --> <!-- END doctoc generated TOC please keep comment here to allow auto update -->

Installation

NPM

Install with npm:

npm install react-recollect

Or Yarn:

yarn add react-recollect

You can then import it in the usual ways:

import { collect, store } from 'react-recollect';

// or
const { collect, store } = require('react-recollect');

CDN

You can also load Recollect from the unpkg CDN.

<script src="https://unpkg.com/react-recollect"></script>

This will create a global ReactRecollect object. See demo/public/browser.html for a working example with React and Babel.

It's a good idea to reference an exact version in the URL, so that it can be cached. Click here to get the full URL.

API

store

<small>Added in 1.0.0</small>

The store object that Recollect exposes is designed to behave like a plain old JavaScript object. But it's a bit different because it's immutable. You can write code as though you were mutating it, but internally it will clone the parts of itself that it needs to clone to apply your changes, without mutating anything.

When the store is then passed to a component, React can do its clever shallow comparisons to know whether something has changed and update efficiently.

collect(ReactComponent)

<small>Added in 1.0.0</small>

When you wrap a component in collect, Recollect will:

Internally, Recollect 'subscribes' components to property 'paths'. For example, this component would be subscribed to the store.page.title path and re-rendered when that property changes.

import { collect } from 'react-recollect';

const Header = ({ store }) => (
  <header>
    <h1>{store.page.title}</h1>
  </header>
);

export default collect(Header);

afterChange(callback)

<small>Added in 1.0.0, the below applies to 4.0.0 and up</small>

afterChange will call the provided callback whenever the store updates.

The callback receives an event object with these properties:

For example, if you want to save the current page to local storage when a particular value in the store changes, you could do the following (anywhere in your app).

import { afterChange } from 'react-recollect';

afterChange((e) => {
  if (e.changedProps.includes('currentPage')) {
    localStorage.currentPage = e.store.currentPage;
  }
});

initStore(data)

<small>Added in 2.4.0</small>

The initStore function will replace the contents of the store with the object you pass in.

data is optional — if you don't pass anything, the store will be emptied (useful in tests).

If you're only using Recollect in the browser, you don't need to use this, but it's handy to set the default state of your store. You can also use Object.assign(store, { foo: 'bar' }) if you want to shallow-merge new data into the store.

When you render on the server though, you do need to initialize the store, because unlike a browser, a server is shared between many users and state needs to be fresh for each request.

On the server

Here's a minimal implementation of server-side rendering with Express and Recollect.

// Create an express app instance
const app = express();

// Read the HTML template on start up (this is the create-react-app output)
const htmlTemplate = fs.readFileSync(
  path.resolve(__dirname, '../../build/index.html'),
  'utf8'
);

// We'll serve our page to requests at '/'
app.get('/', async (req, res) => {
  // Fetch some data
  const tasks = await fetchTasksForUser(req.query.userId);

  // Populate the Recollect store (discarding any previous state)
  initStore({ tasks });

  // Render the app. Components will read from the Recollect store as usual
  const appMarkup = ReactDOMServer.renderToString(<App />);

  // Serialize the store (replacing left tags for security)
  const safeStoreString = JSON.stringify(store).replace(/</g, '\\u003c');

  // Insert the markup and the data into the template
  const htmlWithBody = htmlTemplate.replace(
    '<div id="root"></div>',
    `<div id="root">${appMarkup}</div>
    <script>window.__PRELOADED_STATE__ = ${safeStoreString};</script>`
  );

  // Return the rendered page to the user
  res.send(htmlWithBody);
});

It's important that you populate the store using initStore, and do so before rendering your app with ReactDOMServer.renderToString().

This is because your Node server might receive several requests from several users at the same time. All of these requests share the same global state, including the store object.

So, you must make sure that for each request, you empty the store, populate it with the appropriate data for the request, and render the markup at the same time. And by 'at the same time', I mean synchronously.

In the browser

In the entry point to your app, right before you call ReactDOM.hydrate(), call initStore() with the data that you sent from the server:

import { initStore } from 'react-recollect';

// other stuff

initStore(window.__PRELOADED_STATE__);

ReactDOM.hydrate(<App />, document.getElementById('root'));

This will take the data that you saved in the DOM on the server and fill up the Recollect store with it. You should only init the store once, before the initial render.

Note that initStore will trigger a render of collected components where applicable, and will fire afterChange.

batch(callback)

<small>Added in 4.0.0</small>

The batch function allows you to update the store multiple times, and be guaranteed that components will only be updated after all updates are made.

The callback function will be called immediately and should only contain synchronous code.

import { batch } from 'react-recollect';

const fetchData = async () => {
  const { posts, users, meta } = await fetch('/api').then((response) =>
    response.json()
  );

  batch(() => {
    store.posts = posts;
    store.users = users;
    store.meta = meta;
  });

  // now a render will be triggered for any components that use this data
};

Note that React already does a good job of batching multiple updates into a single render cycle. So only clutter up your code with batch if it results in an actual performance improvement.

useProps(propArray)

<small>Added in 5.1.0</small>

In most cases, you can rely on Recollect to know what data your component requires to render. However, Recollect can't know that your component will require a property in the future. If you reference a property:

... then Recollect won't know about it and your component won't be subscribed to changes in that property.

You can tell Recollect “I want to know if any of these properties change” by passing an array of store objects to the useProps function, like so:

import { collect, useProps } from 'react-recollect';

const MyComponent = ({ store }) => {
  const [showHiddenMessage, setShowHiddenMessage] = useState(false);

  // "This component might read `store.hiddenMessage` in the future"
  useProps([store.hiddenMessage]);

  return (
    <div>
      {showHiddenMessage && <p>{store.hiddenMessage}</p>}

      <button onClick={() => setShowHiddenMessage(true)}>
        Show hidden message
      </button>
    </div>
  );
};

export default collect(MyComponent);

Although useProps starts with the word 'use', it doesn't require React's Hooks mechanism, so it works just fine in versions before React 16.8. (For the curious, the implementation is literally just propArray.includes(0).)

Check out these tests for more usage examples.

PropTypes

<small>Added in 5.2.0</small>

As you've learnt by now, Recollect works by 'recording' which properties your component reads from the store while it renders. This poses a problem if you use the prop-types library, because it is going to read every property that you define in your prop types.

This could result in your component being subscribed to changes in a property it doesn't use, potentially concealing a problem that would only become apparent in production (where prop types aren't checked).

For this reason, react-recollect exports a proxied version of prop-types. It's exactly the same as the normal prop-types library, except that Recollect will pause its recording while your props are being checked.

import { PropTypes } from 'react-recollect';

const MyComponent = (props) => <h1>{props.title}</h1>;

MyComponent.propTypes = {
  title: PropTypes.string.isRequired,
};

export default MyComponent;

We recommended that you uninstall prop-types from your project and replace its usages with the Recollect version. That way no one can accidentally use the 'wrong' prop-types (if they didn't get this far in the readme).

If you use @types/prop-types you can uninstall that too, the types are built into react-recollect.

window.__RR__

Use window.__RR__ to inspect or edit the Recollect store in your browser's console.

__RR__ does not form part of the official API and should not be used in production. It might change between versions without warning and without respecting semver.

It has these properties, available in development or production:

Via the internals object, you can get a reference to the store, which can be handy for troubleshooting. For example, typing __RR__.internals.store.loading = true in the console would update the store and re-render the appropriate components.

If you just log the store to the console, you will see a strange object littered with [[Handler]] and [[Target]] props. These are the proxies. All you need to know is that [[Target]] is the actual object you put in the store.

During development there are two more methods to help you inspect your app:

You can optionally pass a string or regular expression to filter the results.

// Which components are subscribed to the user's status
__RR__.getComponentsByListener('user.status');

// What is <MyComponent> subscribed to?
__RR__.getListenersByComponent('MyComponent');

// What about the <Task> component where the prop `taskId` is 2?
__RR__.getListenersByComponent('Task2', (props) => props.taskId);

Check out the debug test suite for more examples.

Time travel

<small>Added in 5.2.3</small>

You can navigate through the history of changes to the store with the below functions (in your DevTools console):

If you update the store with initStore or execute multiple updates within the batch function callback, those changes are recorded as a single history event.

Note that these time travel functions are only available during development, not in the production build.

Usage with TypeScript

Your store

Define the shape of your recollect store like this:

declare module 'react-recollect' {
  interface Store {
    someProp?: string[];
    somethingElse?: string;
  }
}

Put this in a declarations file such as src/types/RecollectStore.ts.

Using collect

Components wrapped in collect must define store in props — use the WithStoreProp interface for this:

import { collect, WithStoreProp } from 'react-recollect';

interface Props extends WithStoreProp {
  someComponentProp: string;
}

const MyComponent = ({ store, someComponentProp }: Props) => (
  // < your awesome JSX here>
);

export default collect(MyComponent);

If the only prop your component needs is store, you can use WithStoreProp directly.

import { WithStoreProp } from 'react-recollect';

const MyComponent = ({ store }: WithStoreProp) => <div>Hello {store.name}</div>;

Recollect is written in TypeScript, so you can check out the integration tests if you're not sure how to implement something.

(If you've got Mad TypeScript Skillz and would like to contribute, see if you can work out how to resolve the @ts-ignore in the collect module).

Project structure guidelines

The ideas described in this section aren't part of the Recollect API, they're simply a guide.

Two concepts are described in this section (neither of them new):

Cycle of life

In a simple application, you don't need to explicitly think in terms of updaters and selectors. For example:

But as your app grows, it's important to keep your components focused on UI — you don't want 200 lines of logic in the onClick event of a button.

So there will come a point where moving code out of your components into dedicated files is necessary, and at this point, updaters and selectors will serve as useful concepts for organization.

In the examples below, I'll use a directory structure like this:

/my-app
 └─ src
    ├─ components
    ├─ store
    │  ├─ selectors
    │  └─ updaters
    └─ utils

(Fun fact: selector ends in 'or' because 'select' is derived from latin, while updater ends in 'er' because it was made up in 1941 and 'or' had gone out of style.)

Selectors

A simple case for a selector would be to return all incomplete tasks, sorted by due date.

export const getIncompleteTasksSortedByDueDate = (store) => {
  const tasks = store.tasks.slice();

  return tasks
    .sort((a, b) => a.dueDate - b.dueDate)
    .filter((task) => !task.done);
};

You would then use this function by importing it and referencing it in your component:

import { getIncompleteTasksSortedByDueDate } from '../store/selectors/taskSelectors';

const TaskList = ({ store }) => {
  const tasks = getIncompleteTasksSortedByDueDate(store);

  return (
    <div>
      {tasks.map((task) => (
        <Task key={task.id} task={task} />
      ))}
    </div>
  );
};

In this example, I'm passing the store object into the selector. But you could also do import { store } from 'react-recollect' in the selector file. (In Recollect version 4 and earlier, you had to pass the store through. From v5 onwards, you can use props.store or import the store.)

Maybe we want to conditionally show either all tasks or only incomplete tasks. Let's create a second selector. And while we're at it, move repeated sorting code out into its own function:

const getTasksSortedByDate = (tasks) => {
  const sortedTasks = tasks.slice();

  return sortedTasks.sort((a, b) => a.dueDate - b.dueDate);
};

export const getAllTasksSortedByDueDate = (store) =>
  getTasksSortedByDate(store.tasks);

export const getIncompleteTasksSortedByDueDate = (store) =>
  getTasksSortedByDate(store.tasks).filter((task) => !task.done);

And here's a more complex component with local state and a dropdown to show either all tasks or just those that aren't done:

class TaskList extends PureComponent {
  state = {
    filter: 'all',
  };

  render() {
    const { store } = this.props;

    const tasks =
      this.state.filter === 'all'
        ? getAllTasksSortedByDueDate(store)
        : getIncompleteTasksSortedByDueDate(store);

    return (
      <div>
        {tasks.map((task) => (
          <Task key={task.id} task={task} />
        ))}

        <select
          value={this.state.filter}
          onChange={(e) => {
            this.setState({ filter: e.target.value });
          }}
        >
          <option value="all">All tasks</option>
          <option value="incomplete">Incomplete tasks</option>
        </select>
      </div>
    );
  }
}

Now, when a user changes the dropdown, the component state will update, a re-render will be triggered, and as a result, a different selector will be used.

Updaters

An 'updater' is a function that updates the store in some way. As with selectors, you don't need to use updaters, they're just an organizational concept to minimize the amount of data logic you have in your component files.

A simple case for an updater would be to mark all tasks as done in a todo app:

import { store } from 'react-recollect';

export const markAllTasksAsDone = () => {
  store.tasks.forEach((task) => {
    task.done = true;
  });
};

You would reference this from a component by importing it then calling it in response to some user action:

import { markAllTasksAsDone } from '../store/updaters/taskUpdaters';

const Footer = () => (
  <button onClick={markAllTasksAsDone}>Mark all as done</button>
);

export default Footer;

You don't need to 'dispatch' an 'action' from an 'action creator' to a 'reducer'; you're just calling a function that updates the store.

And since these are just plain functions, they're 'composable'. Or in other words, if you want an updater that calls three other updaters, go for it.

Loading data with an updater

Let's create an updater that loads some tasks from an api when our app mounts. It will need to:

  1. Set a loading indicator to true
  2. Fetch some tasks from a server
  3. Save the data to the store
  4. Set the loading indicator to false
export const loadTasksFromServer = async () => {
  store.loading = true;

  store.tasks = await fetchJson('/api/get-my-tasks');

  store.loading = false;
};

You might call this function like so:

import { loadTasksFromServer } from '../store/updaters/taskUpdaters';

class TaskList extends React.Component {
  componentDidMount() {
    loadTasksFromServer();
  }

  render() {
    // just render stuff
  }
}

Asynchronous updaters

Did you notice that we've already covered the super-complex topic of asynchronicity? You can update the Recollect store whenever you like, so you don't need to do anything special to get asynchronous code to work.

Testing an updater

Let's write a unit test to call our updater and assert that it put the correct data in the store. The function we're testing is async, so our test will be async too:

test('loadTasksFromServer should update the store', async () => {
  // Execute the updater
  await loadTasksFromServer();

  // Check that the final state of the store is what we expected
  expect(store).toEqual(
    expect.objectContaining({
      loading: false,
      tasks: [
        {
          id: 1,
          name: 'Fetched task',
          done: false,
        },
      ],
    })
  );
});

Pretty easy, right?

We can make it less easy.

Maybe we want to assert that loading was set to true, then the tasks loaded, and then loading was set to false again.

Well, Recollect exports an afterChange function designed to call a callback every time the store changes. If we pass it a Jest mock function, Jest will conveniently keep a record of each time the store changed.

Also, no one likes half an example, so here's the entire test file:

import { afterChange, store } from 'react-recollect';
import { loadTasksFromServer } from './taskUpdaters';

jest.mock('../../utils/fetchJson', () => async () => [
  {
    id: 1,
    name: 'Fetched task',
    done: false,
  },
]);

test('loadTasksFromServer should update the store', async () => {
  // Create a mock
  const afterChangeHandler = jest.fn();

  // Pass the mock to afterChange. Jest will record calls to this function
  // and therefore record calls to update the store.
  afterChange(afterChangeHandler);

  // Execute our updater
  await loadTasksFromServer();

  // afterChangeHandler will be called with the new version of the store and the path that was changed
  const firstChange = afterChangeHandler.mock.calls[0][0];
  const secondChange = afterChangeHandler.mock.calls[1][0];
  const thirdChange = afterChangeHandler.mock.calls[2][0];

  expect(firstChange.changedProps[0]).toBe('loading');
  expect(firstChange.store.loading).toBe(true);

  expect(secondChange.changedProps[0]).toBe('tasks');
  expect(secondChange.store.tasks.length).toBe(1);

  expect(thirdChange.changedProps[0]).toBe('loading');
  expect(thirdChange.store.loading).toBe(false);

  // Check that the final state of the store is what we expected
  expect(store).toEqual(
    expect.objectContaining({
      loading: false,
      tasks: [
        {
          id: 1,
          name: 'Fetched task',
          done: false,
        },
      ],
    })
  );
});

FAQ

How does it work?

Every object you add to the Recollect store gets wrapped in a Proxy. These proxies allow Recollect to intercept reads and writes. It's similar to defining getters and setters, but far more powerful.

If you were to execute the code below, that site object would be wrapped in a proxy.

store.site = {
  title: 'Page one',
};

(Items are deeply/recursively wrapped, not just the top level object you add.)

Now, if you execute the code store.site.title = 'Page two', Recollect won't mutate the site object to set the title property. Recollect will block the operation and instead create a clone of the object where title is 'Page two'. Recollect keeps a reference between the old and the new site objects, so any attempt to read from or write to the 'old version' will be redirected to the 'new version' of that object.

In addition to intercepting write operations, the proxies also allow Recollect to know when data is being read from the store. When you wrap a component in collect, you're instructing Recollect to monitor when that component starts and stops rendering. Any read from the store while a component is rendering results in that component being 'subscribed' to the property that was read.

Bringing it all together: when some of your code attempts to write to the store, Recollect will clone as described above, then notify all the components that use the property that was updated, passing those components the 'next' version of the store.

What sort of stuff can go in the store?

You can store anything that's valid JSON. If that's all you want to do, you can skip the rest of this section.

<small>The below applies to 4.0.0 and up</small>

Recollect will store data of any type, including (but not limited to):

Recollect will monitor changes to:

Recollect will store, but not monitor attempted mutations to other objects. For example:

The same applies to WeakMap, DataView, ArrayBuffer and any other object you can think of.

If there's a data type you want to store and mutate that isn't supported, log an issue and we'll chat.

Other things that aren't supported (or haven't been tested):

You can even store components in the store if you feel the need. This hasn't been performance tested, so proceed with caution.

const Page = collect(({ store }) => {
  const { Header, Footer, Button } = store.components;

  return (
    <React.Fragment>
      <Header title="Page one" />

      <Button onClick={doSomething} />

      <Footer />
    </React.Fragment>
  );
});

Map and Set limitations

Updating an object key of a map entry will not always trigger a render of components using that object. But in most cases you'd be storing your data in the value of a map entry, and that works fine.

Similarly, updating an object in a set may not trigger an update to components using that object (adding/removing items from a set works fine).

When will my components be re-rendered?

Short version: when they need to be, don't worry about it.

Longer version: if a property is changed (e.g. store.page.title = 'Page two'), any component that read that property when it last rendered will be updated. If an object, array, map or set is changed (e.g. store.tasks.pop()) any component that read that object/array/map/set will be updated.

Check out tests/integration/updating for the full suite of scenarios.

How many components should I wrap in collect?

You can wrap every component in collect if you feel like it. As a general rule, the more you wrap in collect, the fewer unnecessary renders you'll get, and the less you'll have to pass props down through your component tree.

There is one rule you must follow though: do not pass part of the store into a collected component as props. Don't worry if you're not sure what that means, you'll get a development-time error if you try. This issue explains why.

When dealing with components that are rendered as array items (e.g. <Product>s in a <ProductList>), you'll probably get the best performance with the following setup:

With this arrangement, when an item in the array changes (e.g. a product is starred), Recollect will immutably update only that item in the store, and trigger the parent component to update. React will skip the update on all the children that didn't change, so only the one child will re-render.

For a working example, see the Recollect demo on CodeSandbox.

Can I use this with class-based components and functional components?

Yep and yep.

Hooks?

Yep.

Will component state still work?

Yes, but be careful. If a change in state reveals some new UI, and a property from the store is only read in that UI (not elsewhere in the component) then Recollect won't be aware of it, and won't update your component if it changes.

Use the useProps function to make sure your component is subscribed to changes in this property.

Do lifecycle methods still fire?

Yep. Recollect has no effect on componentDidMount, componentDidUpdate and friends.

Why isn't my componentDidUpdate code firing?

If you have a store prop that you only refer to in componentDidUpdate (e.g. store.loaded), then your component won't be subscribed to changes in that prop. So when store.loaded changes, your component might not be updated.

Use the useProps function to make sure your component is subscribed to changes in this property.

Can I use this with shouldComponentUpdate()?

Yes, but no, but you probably don't need to.

The React docs say of shouldComponentUpdate():

This method only exists as a performance optimization. Do not rely on it to “prevent” a rendering, as this can lead to bugs ... In the future React may treat shouldComponentUpdate() as a hint rather than a strict directive, and returning false may still result in a re-rendering of the component

So, if you're using shouldComponentUpdate for performance reasons, then you don't need it anymore. If the shouldComponentUpdate method is executing, it's because Recollect has told React to update the component, which means a value that it needs to render has changed.

Can I wrap a PureComponent or React.memo in collect?

There's no need. The collect function wraps your component in a PureComponent and there's no benefit to having two of them.

It's a good idea to wrap other components in PureComponent or React.memo though — especially components that are rendered in an array, like <Todo>. If you have a hundred todos, and add one to the list, you can skip a render for all the existing <Todo> components if they're marked as pure.

Can I use this with Context?

Yes. Recollect doesn't interfere with other libraries that use Context.

You shouldn't need to use Context yourself though. You have a global store object that you can read from and write to anywhere.

Can I use this with refs?

Yes, refs just work, as long as you don't use the reserved name 'ref' (React strips this out). You can use something like inputRef instead. For an example, see this test

Can I have multiple stores?

No. There is no performance improvement to be had, so the desire for multiple stores is just an organizational preference. For this, you can use 'selectors' to focus on a subset of your store.

Can I use Recollect without React?

Yes! You can use store without using collect. Pair this with afterChange to have an object that notifies you when its changed. For an example, check out tests/integration/nodeJs.js

I'm getting a no-param-reassign ESLint error

You can add 'store' as a special case in your ESLint config so that the rule allows you to mutate the properties of store.

{
  "rules": {
    "no-param-reassign": [
      "error",
      {
        "props": true,
        "ignorePropertyModificationsFor": ["store"]
      }
    ]
  }
}

Check out the no-param-reassign rule in this repo's eslint config for the syntax.

Tell me about your tests

How big is it?

3—5 KB, depending on what else you've got installed. If you're coming from Redux land, you'll save about 1 KB in library size, but the big savings come from getting rid of all your reducers.

Is reading/writing via a proxy slow?

Slow, no. Slower than vanilla object operations, yes.

Let's quantify with a case study: in an app that has a store with ~80,000 properties, ~30,000 of them proxied objects and ~1,000 component listeners, updating a chunk of data requiring 50 new proxies takes ~2 milliseconds (on desktop). For that app, 2ms was considered insignificant compared to the ~90ms spent on the resulting render cycle.

If you're processing big data and facing performance troubles, open an issue and we'll chat.

Dependencies

Recollect has a peer dependency of React >=15.3.

Alternatives

If you want IE support, use Redux.

If you want explicit 'observables' and multiple stores, use MobX.

If you want a walk down memory lane, use Flux.

Also there is a library that is very similar to this one (I didn't copy, promise) called react-easy-state.

Is it really OK to drop support for IE?

Sure, why not! Imagine: all that time you spend getting stuff to work for a few users in crappy old browsers could instead be spent making awesome new features for the vast majority of your users.

For inspiration, these brave websites have dropped the hammer and dropped support for IE: