As it is, if you have some Uint8Array that is a view into a larger ArrayBuffer then a copy of the data is made in encode because new Uint8Array(someOtherUint8Array) makes new ArrayBuffer with a copy of the data. Where as you could just use the Uint8Array passed in?

Basically this would change

  exports.encode = function(arraybuffer) {
    var bytes = new Uint8Array(arraybuffer),

To something like

  exports.encode = function(arraybuffer) {
    var bytes = (arrayBuffer instanceof Uint8Array || arrayBuffer instanceof Uint8ClampArray)
            ? arrayBuffer
            : new Uint8Array(arraybuffer),

console.log(encode(decode('P-vo4HxWwOyJEZsd-7dYER4L7VLMZzDD0Hs6IdiMT_k')))
// Output: PAvo4HxWwOyJEZsdA7dYER4L7VLMZzDD0Hs6IdiMTAk=

Am I doing something wrong?

Hi! It seems the sources field in the sourcemap base64-arraybuffer.es5.js.map is invalid:

{"version":3,"file":"base64-arraybuffer.es5.js","sources":["../../src/index.ts"],"sourcesContent":[null],"names":[], ...

It should be ../src/index.ts I think.

node_modules/
└── base64-arraybuffer
    ├── dist
    │   ├── base64-arraybuffer.es5.js
    │   ├── base64-arraybuffer.es5.js.map
    │ ...
    ├── package.json
    ├── README.md
    └── rollup.config.ts

Not sure why though.

Related: socketio/socket.io-client#1520

For repeated invocations with temporary data, there's an opportunity (not sure how great) for memory allocation savings if instead of always creating a new ArrayBuffer, the decode function allows an optional 2nd parameter for a reusable ArrayBuffer that is console.assert()'ed to be at least the length needed for decoding the base64 string.

Perhaps the decode function could return the number of bytes written instead of an ArrayBuffer if a buffer is provided - speculating usefulness for handling variable-byte input strings.

Would you be open to a PR for this?

Hi,

encode method for buffer is giving me base64 strng.. But reverse is not happening.
Its giving me empty ArrayBuffer.

Am I doing anything wrong?

Here is my code

var b64toBuff = require('base64-arraybuffer');
let data = fs.readFileSync(imagePath);
let c = b64toBuff.encode(data);
let d = b64toBuff.decode(c)

c gives me a base64 string.
But d gives me empty ArrayBuffer{}

Hi,
Thanks a lot for your work on this library! I'm curious if you perhaps benchmarked your solution against browser built-in functions btoa and atob?

If not, I could volunteer to set up a benchmark and create a PR to include the results in the readme if that's desirable? Do you have any preferance on how to benchmark?

We internally used this package for encoding/decoding large pdf files and noticed our instance crashing on around 2gb of files. The instance had free memory of 12gb! After hours of debugging we narrowed it down to the encode function of this library. I read the encode function code and tried to figure out why? But couldn't find a leak. But from our testing and debugging, I am sure there's a memory leak in encode function (in nodejs v8) @niklasvh

For example:

Use this words as an input:

Base 64 \u2014 Mozilla Developer Network

Add a test case such as:

it('encode "Base 64 \u2014 Mozilla Developer Network"', () => equal(encode(stringArrayBuffer("Base 64 \u2014 Mozilla Developer Network")), "QmFzZSA2NCDigJQgTW96aWxsYSBEZXZlbG9wZXIgTmV0d29yaw=="))

The result is

      + expected - actual

      +QmFzZSA2NCDigJQgTW96aWxsYSBEZXZlbG9wZXIgTmV0d29yaw==
      -QmFzZSA2NCAUIE1vemlsbGEgRGV2ZWxvcGVyIE5ldHdvcms=

Is it possible to publish version 1.1 on npm? The currently available one no longer works with current ArrayBuffer API.

I want to Encode my file in C# to base64 and decode it with this module in my Ionic App
I use this code to Encode in C# but it seems not get the true file in my Client when I decode it:

Byte[] bytes = File.ReadAllBytes(localFilePath); String file = Convert.ToBase64String(bytes);

An install of version 0.1.2 from npm adds two temporary backup files to system. These files are README.md~ and package.json~.

These seem to be deleted from git and hence similar update should be pushed to npm.