SnappyJS

A pure JavaScript implementation of Google's Snappy compression library.

This implementation is reasonably fast (see benchmark below). It takes advantage of ArrayBuffer.

Install

If using with Node.js,

npm install snappyjs

If using with Bower,

bower install snappyjs

Usage

Node.js

SnappyJS works with Node.js 10.x or later.

var SnappyJS = require('snappyjs')
var buffer = new ArrayBuffer(100)
// fill data in buffer
var compressed = SnappyJS.compress(buffer)
var uncompressed = SnappyJS.uncompress(compressed)

Browser

You can also use SnappyJS in browser. Adding dist/snappyjs.js or dist/snappyjs.min.js will introduce SnappyJS in the global scope.

SnappyJS relies on ArrayBuffer. All major browsers support it now (http://caniuse.com/#feat=typedarrays). Also, as I tested, SnappyJS has high performance on latest version of Google Chrome, Safari, Firefox, and Microsoft Edge.

When using webpack to build your project, and you plan to only use ArrayBuffer or Uint8Array as input parameters, make sure to put the following in your webpack config to prevent it from automatically bundling a Buffer polyfill:

  node: {
    Buffer: false,
  }

API

SnappyJS.compress(input)

Compress input, which must be type of ArrayBuffer, Buffer, or Uint8Array. Compressed byte stream is returned, with same type of input.

SnappyJS.uncompress(compressed, [maxLength])

Uncompress compressed, which must be type of ArrayBuffer, Buffer, or Uint8Array. Uncompressed byte stream is returned, with same type of compressed.

If maxLength is provided, uncompress function will throw an exception if the data length encoded in the header exceeds maxLength. This is a protection mechanism for malicious data stream.

Benchmark

Although JavaScript is dynamic-typing, all major JS engines are highly optimized. Thus well-crafted JavaScript code can have competitive performance even compared to native C++ code.

I benchmark SnappyJS against node-snappy (which is Node.js binding of native implementation).

Command for benchmark is node benchmark. Below is the result running on Node.js v5.5.0.

Real text #1 (length 618425, byte length 618425), repeated 100 times:
node-snappy#compress x 2.31 ops/sec ±1.47% (10 runs sampled)
snappyjs#compress x 0.91 ops/sec ±0.92% (7 runs sampled)
node-snappy#uncompress x 7.22 ops/sec ±4.07% (22 runs sampled)
snappyjs#uncompress x 2.45 ops/sec ±1.53% (11 runs sampled)

Real text #2 (length 3844590, byte length 3844591), repeated 10 times:
node-snappy#compress x 7.68 ops/sec ±2.78% (23 runs sampled)
snappyjs#compress x 3.56 ops/sec ±1.44% (13 runs sampled)
node-snappy#uncompress x 17.94 ops/sec ±4.71% (33 runs sampled)
snappyjs#uncompress x 7.24 ops/sec ±2.57% (22 runs sampled)

Random string (length 1000000, byte length 1500098), repeated 50 times:
node-snappy#compress x 6.69 ops/sec ±5.23% (21 runs sampled)
snappyjs#compress x 2.39 ops/sec ±2.54% (10 runs sampled)
node-snappy#uncompress x 14.94 ops/sec ±6.90% (40 runs sampled)
snappyjs#uncompress x 5.92 ops/sec ±4.28% (19 runs sampled)

Random string (length 100, byte length 147), repeated 100000 times:
node-snappy#compress x 4.17 ops/sec ±2.96% (15 runs sampled)
snappyjs#compress x 5.45 ops/sec ±1.51% (18 runs sampled)
node-snappy#uncompress x 4.39 ops/sec ±3.83% (15 runs sampled)
snappyjs#uncompress x 14.01 ops/sec ±2.06% (38 runs sampled)

From the result, we see that SnappyJS has 35%~45% performance of native implementation. If input size is small, SnappyJS may have better performance than node-snappy. It is because calling native function in JS is much more expensive than calling JS function.

License

MIT License