Does not encode checkmark properly

[alanhogan]

var btoa = require("btoa")
var atob = require("atob")


btoa('✓'); // "Ew==" (not right, should be "4pyT")
atob(btoa('✓')); // "\u0013" (not right)

[shirakaba]

This repo's btoa() implementation uses latin1 encoding

This repository's btoa() implementation currently essentially does this:

const btoa = function(str) { return new Buffer(str, 'latin1').toString('base64'); }

The latin1 flag – full documentation here – informs the Buffer that the input string is encoded in latin1 format (ie. single-byte characters). However, the checkbox character cannot be represented latin1 (which only represents a basic set of 256 characters).

How to configure your own btoa() to use utf8 encoding

The checkbox character can be represented in UTF-8, however: If you replace the latin1 flag with utf8 (or remove the flag altogether, because utf8 is Buffer's default), then the btoa() shim will encode the checkmark character as "4pyT" as you desire:

const btoa = function(str) { return new Buffer(str, 'utf8').toString('base64'); }

Remember that your atob() shim would also need to be set up to decode utf8 in turn:

const atob = function(b64Encoded) {return new Buffer(b64Encoded, 'base64').toString('utf8');}

latin1 is the correct encoding format for this repo

latin1 is the encoding format used by desktop Chrome (and all other desktop browsers, I would assume), which in fact throws an exception upon input of any string containing characters outside the latin1 range. Thus, it is the appropriate encoding format for this library. No change needed here.

[alanhogan]

I see. That’s fair enough. I didn’t realize the context of this package was to provide the same as a browser. I just searched for btoa on npmjs.org. Most of the other results attempt to provide full utf-8 support (solving the “unicode problem”) as far as I can tell.