Skip to main content

Getting Started

Base Setup

The base setup is simple: just wrap your app into the GlobalStateProvider component, provided by this library, and you'll be able to use any library hooks within its child hierarchy.

/* The minimal example of the library setup and usage. */

import React from 'react';

import {
GlobalStateProvider,
useAsyncData,
useGlobalState,
} from '@dr.pogodin/react-global-state';

/* Example of component relying on the global state. */

function SampleComponent() {
const [value, setValue] = useGlobalState('sample.component', 0);
return (
<button onClick={() => setValue(1 + value)}>
{value}
</button>
);
}

/* Example of component relying on async data in the global state. */

async function sampleDataLoader() {
return new Promise((resolve) => {
setTimeout(() => resolve('Sample Data'), 500);
});
}

function SampleAsyncComponent() {
const { data, loading } = useAsyncData('sample.async-component', sampleDataLoader);
return data;
}

/* Example of the root app component, providing the state. */

export default function SampleApp() {
return (
<GlobalStateProvider>
<SampleComponent />
<SampleAsyncComponent />
</GlobalStateProvider>
);
}

Multiple, or nested GlobalStateProvider instances are allowed, and they will provide independent global states to its children (shadowing parent ones, in the case of nesting). However, the current SSR implementation assumes a single GlobalStateProvider at the app root. Multiple providers won't break it, but won't be a part of SSR data loading either.

This setup is fine to run both at the client, and at the server-side, but in the case of server-side rendering, the library won't run any async data fetching, thus rendering pages with the initial global state; e.g. in the example above the <SampleAsyncComponent> will be rendered as an empty node, as data will be undefined, and loading will be false. To handle SSR better, and to have <SampleAsyncComponent> rendered as Sample Data even at the server-side, you need the following, a bit more complex, setup.

Advanced SSR Setup

caution

This section still contains only the original description of SSR setup with JavaScript flavour of the library. The TypeScript flavour of SSR setup is to be documented...

Advanced setup with the server-side rendering (SSR) support is demonstrated below. It requires slightly different code depending on whether you use the new renderToPipeableStream() function, introduced in React 18 as the recommended API for SSR, or you still prefer the older renderToString(). Sure, it is also possible to implement similar setup with other SSR APIs offered by React.

info

The renderToPipeableStream() has advantage of full Suspense support and HTML streaming. However, as React's implementation of Suspense still does not cover data fetching multiple rendering passes are still required for full SSR integration of this library, because in each pass effectively collect global state updates which are applied to follow-up separate renders, something which is not natively supported by React SSR APIs yet.

Assume that <SampleComponent>, sampleDataLoader(..), and <SampleAsyncComponent> are defined the same way as in the Base Setup section, and <SampleApp> component itself does not include the GlobalStateProvider, i.e.

export default function SampleApp() {
return (
<React.Fragment>
<SampleComponent />
<SampleAsyncComponent />
</React.Fragment>
);
}

The server-side rendering code becomes:

/* Server-sider rendering. */

import React from 'react';
import ReactDOM from 'react-dom/server';

import { GlobalStateProvider } from '@dr.pogodin/react-global-state';

import SampleApp from 'path/to/app';

// As you want to keep server latency within a reason, it is a good idea
// to limit the maximum number of SSR rounds, or the maximum SSR time, or
// both, and perform any async operations which took too long at the client
// side.
const MAX_SSR_ROUNDS = 3;
const SSR_TIMEOUT = 1000; // 1 second (in milliseconds).

/**
* Executes a single rendering pass.
* @param {object} ssrContext SSR context object is used to pass information
* between separate rendering passes.
* @return {Promise<{
* abort: function;
* pipe: function;
* }>} Once all page content is ready it resolves to the stream with pipe(res)
* and abort() methods, returned by internal renderToPipeableStream() call.
*/
async function renderPass(ssrContext) {
return new Promise((resolve, reject) => {
const stream = ReactDOM.renderToPipeableStream(
<GlobalStateProvider
initialState={ssrContext.state}
ssrContext={ssrContext}
>
<SampleApp />
</GlobalStateProvider>,
{
onAllReady: () => resolve(stream),
onError: reject,
},
);
});
}

async function renderServerSide() {
let stream;
const start = Date.now();
const ssrContext = { state: {} };
for (let round = 0; round < MAX_SSR_ROUNDS; round += 1) {
// SSR round.
stream = await renderPass(ssrContext);

// SSR round did not altered the global state: we are done.
if (!ssrContext.dirty) break;

// Waiting for pending async operations to complete.
const timeout = SSR_TIMEOUT + start - Date.now();
const ok = timeout > 0 && await Promise.race([
Promise.allSettled(ssrContext.pending),
new Promise((x) => setTimeout(() => x(false), timeout)),
]);
if (!ok) break; // SSR timeout.
}

// At this point the "stream" should be used to pipe generated HTML markup
// into the server response, and "state" should be injected into the client
// side and used as the "initialState" of <GlobalStateProvider> there.
return { stream, state: ssrContext.state };
}

⇑ When ssrContext property is passed into the GlobalStateProvider, the corresponding global state object switches into the SSR mode. In this mode, if the app rendering modifies the state, the ssrContext.dirty flag is set true, and for any async operations, triggered by the library hooks, corresponding promises are added into the ssrContext.pending array. Thus, the block of code

if (!ssrContext.dirty) break;

const timeout = SSR_TIMEOUT + start - Date.now();
const ok = timeout > 0 && await Promise.race([
Promise.allSettled(ssrContext.pending),
new Promise((x) => setTimeout(() => x(false), timeout)),
]);
if (!ok) break;

in the case when last rendering pass triggered async operations, it waits for them to complete, and allows the rendering pass to be redone with the new initial value of the global state, which is written to ssrContext.state in this case. If no updates to the state happened in the last rendering pass, this block breaks out of the loop, leaving to you in the render variable the HTML markup to send to the client, and in the ssrContext.state the initial value of the global state to use for app initialization at the client side.

The outer for loop serves to protect against possible long re-rendering loops: if after several re-renders the state has not become stable, it is fine to send to the client side the latest render and state results, and finalize the rest of rendering at the client side. Similarly, the timeout condition interrupts SSR and cause the code to serve the current render and state, if any pending async operation takes too long, thus compromising server latency.

In case when some async operations are too long to wait for them during SSR, the async hooks accept noSSR option, to be ignored during SSR. Additional option is to wrap the rendering cycle into a timeout race codition, and if the desired rendering time has bit hit, the rendering loop can be interrupted, and the latest render and state can be sent to the client side.

The corresponding client-side rendering is simple, just pass the state calculated during the server-side rendering into the initialState prop of GlobalStateProvider at the client side:

/* Client-side rendering. */

import React from 'react';
import ReactDOM from 'react-dom';

import { GlobalStateProvider } from '@dr.pogodin/react-global-state';

import SampleApp from 'path/to/app';

function renderClientSide(stateFromServerSide) {
ReactDOM.hydrate((
<GlobalStateProvider initialState={stateFromServerSide}>
<SampleApp />
</GlobalStateProvider>
), document.getElementById('your-react-view'));
}