base64-async

Non-blocking chunked Base64 encoding

Process large Base64 documents without blocking the event loop.

Configurable chunk size option to optimise for your use case.

Note:

Base64 in Node.js is already crazy fast. Breaking the work up into chunks and adding async logic adds overhead. If you aren't dealing with large files it will probably be more efficient to just block the event loop for the small amount of time it takes Node.js to process Base64 synchronously.

Install

npm install --save base64-async

Usage

const b64 = require('base64-async');
const fs = require('fs');
const buffer = fs.readFileSync('somehugefile.jpg');

b64.encode(buffer).then(b64String => console.log(b64String));
// aGkgbXVt...

b64.decode(b64String).then(buffer => console.log(buffer));
// <Buffer 68 69 20 6d 75 6d ... >

// or, for the cool kids
const b64String = await b64.encode(buffer);
const buffer = await b64.decode(b64String);

// which is equivalent to this
const b64String = await b64(buffer);
const buffer = await b64(b64String);
// If no method is specified, buffers are encoded, strings are decoded

Example

$ npm run example

Registering 4 asynchronous jobs...
Encoding 100 MB with default Node.js Buffer API...
Base64 encode complete
Hi, I'm an asynchronous job, and I'm late by 231ms
Hi, I'm an asynchronous job, and I'm late by 238ms
Hi, I'm an asynchronous job, and I'm late by 239ms
Hi, I'm an asynchronous job, and I'm late by 245ms

Registering 4 asynchronous jobs...
Encoding 100 MB with base64-async in chunks of 250 kB...
Hi, I'm an asynchronous job, and I'm on time
Hi, I'm an asynchronous job, and I'm on time
Hi, I'm an asynchronous job, and I'm on time
Hi, I'm an asynchronous job, and I'm on time
Base64 encode complete

(example source code)

Notice how none of the async jobs can start until the Buffer API has finished encoding and stops blocking the event loop? With base64-async the async jobs can execute in-between each chunk of data.

Performance

$ npm run bench

Benchmark completed with a chunk size of 250 kB
┌────────┬──────────────┬──────────────┬──────────────┬──────────────┐
│ Bytes  │ Encode Sync  │ Decode Sync  │ Encode Async │ Decode Async │
├────────┼──────────────┼──────────────┼──────────────┼──────────────┤
│ 10 kB  │ 0.097225ms   │ 0.383031ms   │ 1.276201ms   │ 0.537687ms   │
├────────┼──────────────┼──────────────┼──────────────┼──────────────┤
│ 100 kB │ 0.198161ms   │ 0.271577ms   │ 0.99799ms    │ 0.356765ms   │
├────────┼──────────────┼──────────────┼──────────────┼──────────────┤
│ 1 MB   │ 1.924415ms   │ 2.038406ms   │ 2.679117ms   │ 2.544993ms   │
├────────┼──────────────┼──────────────┼──────────────┼──────────────┤
│ 10 MB  │ 15.749204ms  │ 16.280246ms  │ 33.666111ms  │ 29.918725ms  │
├────────┼──────────────┼──────────────┼──────────────┼──────────────┤
│ 100 MB │ 165.189455ms │ 195.298199ms │ 246.359068ms │ 280.792751ms │
└────────┴──────────────┴──────────────┴──────────────┴──────────────┘

As you can see, the total processing time is longer with base64-async (as we spend some time paused waiting for the event loop). However, if you have an idea of the size of the data you'll be working with, you can play around with the chunk size to get better performance.

The included benchmarking tool accepts arguments to help you test this:

$ npm run bench -- --chunkSize=1000000 --bytesToBenchmark=50000000,100000000

Benchmark completed with a chunk size of 1 MB
┌────────┬──────────────┬──────────────┬──────────────┬──────────────┐
│ Bytes  │ Encode Sync  │ Decode Sync  │ Encode Async │ Decode Async │
├────────┼──────────────┼──────────────┼──────────────┼──────────────┤
│ 50 MB  │ 79.675533ms  │ 87.251079ms  │ 92.400367ms  │ 137.468082ms │
├────────┼──────────────┼──────────────┼──────────────┼──────────────┤
│ 100 MB │ 203.423705ms │ 173.567974ms │ 186.181857ms │ 264.123311ms │
└────────┴──────────────┴──────────────┴──────────────┴──────────────┘

API

b64(input, [options])

Returns a Promise that resolves to the Base64 encoded/decoded input.

input

Type: string, buffer

A Base64 string to decode, or a Buffer to encode.

options

Type: object

options.chunkSize

Type: number
Default: 250000

Size of the chunk of data to work on before deferring execution to the next iteration of the event loop.

If encoding, the number is interpreted as number of bytes. If decoding, the number is interpreted as number of characters.

b64.encode(input, [options])

Returns a Promise that resolves to a Base64 encoded string.

input

Type: buffer

A Buffer to encode.

b64.decode(input, [options])

Returns a Promise that resolves to a decoded Buffer.

input

Type: string

A Base64 string to decode.

License

MIT © Luke Childs