Maducer

An experimental map-reduce concurrency over web workers using shared array buffer for handling large datasets.

Getting Started

Maducer has a similar interface to the native implementation – define your mapping and reducing functions, and additionally instruct Maducer how to split your data into logical chunks.

// Configure how to map and reduce your data.
const mapper = a => Number(a);
const reducer = (a, b) => a > b ? a : b;

// Load data from a file, API or some place else.
const data = await fetch('./dataset.csv').then(r => r.text());

// Define how your data is delimited.
const delimiter = ',';

// Put it all together and compute the highest value.
const compute = maducer(delimiter, mapper, reducer);
const result = await compute(data);
console.log('Highest value is:', result);

Maducer will distribute the concurrent map-reduce computation across x web workers using SharedArrayBuffer where x is the amount of CPU cores available to the client.

It's also possible to pass an ArrayBuffer to Maducer – you'll still get the decoded value in your map and reduce functions, but there is a performance gain in using a buffer as opposed to plain text – using fetch you can yield a buffer by using response.arrayBuffer().