Convert a typed array to a Buffer without a copy


Convert a typed array to a Buffer without a copy.

Say you're using the 'buffer' module on npm, or browserify and you're working with lots of binary data.

Unfortunately, sometimes the browser or someone else's API gives you an ArrayBuffer or a typed array like Uint8Array to work with and you need to convert it to a Buffer. What do you do?

Of course: new Buffer(uint8array)

But, alas, every time you do new Buffer(uint8array) the entire array gets copied. The Buffer constructor does a copy; this is defined by the node docs and the 'buffer' module matches the node API exactly.

So, how can we avoid this expensive copy in performance critical applications?

Simply use this module, of course!

npm install typedarray-to-buffer

To convert a typed array to a Buffer without a copy, do this:

var toBuffer = require('typedarray-to-buffer')
var arr = new Uint8Array([1, 2, 3])
arr = toBuffer(arr)
// arr is a buffer now! 
arr.toString()  // '\u0001\u0002\u0003' 
arr.readUInt16BE(0)  // 258 

If the browser supports typed arrays, then toBuffer will augment the typed array you pass in with the Buffer methods and return it. See how does Buffer work? for more about how augmentation works.

This module uses the typed array's underlying ArrayBuffer to back the new Buffer. This respects the "view" on the ArrayBuffer, i.e. byteOffset and byteLength. In other words, if you do toBuffer(new Uint32Array([1, 2, 3])), then the new Buffer will contain [1, 0, 0, 0, 2, 0, 0, 0, 3, 0, 0, 0], not [1, 2, 3]. And it still doesn't require a copy.

If the browser doesn't support typed arrays, then toBuffer will create a new Buffer object, copy the data into it, and return it. There's no simple performance optimization we can do for old browsers. Oh well.

If this module is used in node, then it will just call new Buffer. This is just for the convenience of modules that work in both node and the browser.

MIT. Copyright (C) Feross Aboukhadijeh.