Skip to content

An autonomous parallel processing engine for the browser.

License

Notifications You must be signed in to change notification settings

GitSquared/rinzler

Repository files navigation

Rinzler project logo

npm version badge types included badge license badge


Rinzler is a turboramjet parallel processing engine for the browser.

It speeds up your web application by allowing recurring functions to execute in parallel taking full advantage of the host system's available processing power.

Check out the full docs, try the interactive demo or read on for a high-level overview and a quick start guide.

Concept

Most devices have a processor unit (CPU) with multiple cores, meaning that they are capable of working on multiple tasks at the same time. Modern operating systems with multi-tasking functionality (e.g the ability to run & manage multiple programs/windows) have a special component called a thread scheduler.

Each program you run can have multiple threads, and the scheduler's job is to distribute threads to the CPU's cores.

A web page's JavaScript normally executes on a single thread, meaning it will never use more than one CPU core. In most cases this is fine, and also helps ensures other tabs in the user's browser, or other programs, can also keep running smoothly.

However, some web applications might need to process a lot of data, or do a lot of expensive computing, and therefore can benefit from spreading work across all the available cores of the host machine.

Rinzler is a tool to do just that, in the simplest way possible - just define functions to run in parallel, and use native ES6 Promises to run & manage parallelized jobs.

Infographic explaining how Rinzler allows you to use more CPU cores

Internally, it leverages Web Workers, which is a standard Web API for executing code in separate threads. It also includes a custom scheduler that handles spawning new Workers when necessary, sharing work between them, and shutting them down when they're not used.

Install

npm i rinzler-engine

Both ES & UMD modules are bundled, as well as TypeScript types, so you should be all set.

Rinzler targets browsers with WebWorkers and Promises support (check the browserslistrc). Most modern evergreen browsers, including Edge, should be compatible.

Quick start

In the following example, we will set up a Rinzler-accelerated app that decodes ArrayBuffers of text.

0. Environment initialization (optional)

In most real-life use cases, the job processing you will offload to Rinzler will depend on some dynamic variable in the context of your app: in this example, the original encoding of the text we want to decode.

The processing functions you will pass to Rinzler cannot contain references to external variables, because their source code will be extracted and printed in the Worker instances' source.

To work around this limitation, Rinzler allows you to setup an "initialization" function and pass it a payload. This function & payload will be run on each new Web Worker instance before it starts processing your jobs.

const initOptions = [{
  encoding: 'utf-8' // We're just going to print this here, but in real life you would probably get this option from user input.
}]

function init(options) {
  // This will run once in new Web Worker contexts. We can use the `self` global to store data for later.
  self.params = {
    encoding: options.encoding
  }
}

1. Job processing function

We need to setup a function that will actually do the job we need to parallelize, in this case, decoding text buffers.

function processJob(message) {
  // We expect to receive an object with an `encodedText` prop that is an ArrayBuffer.
  const buffer = message.encodedText

  // Get the encoding parameter we stored earlier, or default to ASCII.
  const encoding = self.params?.encoding || 'ascii'

  const text = new TextDecoder(encoding).decode(buffer)
  return [text]
}

2. Engine start

Next we will import Rinzler and start the engine by passing the function(s) we defined above.

The following code is written for asynchronous contexts, but you can translate it to synchronous by using .then() with a callback instead of await.

import RinzlerEngine from 'rinzler-engine'

const engine = await new RinzlerEngine().configureAndStart(processJob, init, initOptions)

3. Running jobs

Now we can actually run jobs! We'll use the runJob() method, which returns a Promise that will resolve when the job is completed.

Since we need to pass an ArrayBuffer, we'll use the second argument as a Transferable[] - much like in the native worker.postMessage() API.

// Encode some text to try our decoder with
const encodedText = new TextEncoder('utf-8').encode('hello Rinzler!')

// Pass the encoded text to our decoding engine
const decodedResult = await engine.runJob({ encodedText }, [encodedText])

console.log(decodedResult) // "hello Rinzler!"

You can start as many jobs as you want, and take full advantage of ES6's asynchronous syntax (for example, Promise.all()).

If you use TypeScript, you can pass return types with the runJob<T>(): Promise<T> signature.

Under the hood, Rinzler will take care of launching Web Workers, balancing their load, and gracefully shutting them down when needed to reduce your app's memory footprint.

4. Shutting down

Web Worker instances will be destroyed by the browser when the page exits, but you can schedule a graceful shutdown yourself using engine.shutdown(), which returns a Promise that will resolve once all currently active jobs have completed and all workers have been stopped.

Licensing

Rinzler is licensed under the MIT License. You may integrate it in commercial applications.


© 2020-2022 Gabriel Saillard gabriel@saillard.dev