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 a typed array like
Uint8Array
to work with and you need to convert it to a Buffer
. What do you do?
Of course: Buffer.from(uint8array)
But, alas, every time you do Buffer.from(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!
If you have an ArrayBuffer
, you don't need this module, because
Buffer.from(arrayBuffer)
is already efficient.
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 Buffer.from
. This is just for
the convenience of modules that work in both node and the browser.
MIT. Copyright (C) Feross Aboukhadijeh.