I found that although the shims from answers above worked, they did not match the behaviour of desktop browsers' implementations of btoa()
and atob()
:
const btoa = function(str){ return Buffer.from(str).toString('base64'); }
// returns "4pyT", yet in desktop Chrome would throw an error.
btoa('?');
// returns "fsO1w6bCvA==", yet in desktop Chrome would return "fvXmvA=="
btoa(String.fromCharCode.apply(null, new Uint8Array([0x7e, 0xf5, 0xe6, 0xbc])));
As it turns out, Buffer
instances represent/interpret strings encoded in UTF-8 by default. By contrast, in desktop Chrome, you can't even input a string that contains characters outside of the latin1 range into btoa()
, as it will throw an exception: Uncaught DOMException: Failed to execute 'btoa' on 'Window': The string to be encoded contains characters outside of the Latin1 range.
Therefore, you need to explicitly set the encoding type to latin1
in order for your Node.js shim to match the encoding type of desktop Chrome:
const btoaLatin1 = function(str) { return Buffer.from(str, 'latin1').toString('base64'); }
const atobLatin1 = function(b64Encoded) {return Buffer.from(b64Encoded, 'base64').toString('latin1');}
const btoaUTF8 = function(str) { return Buffer.from(str, 'utf8').toString('base64'); }
const atobUTF8 = function(b64Encoded) {return Buffer.from(b64Encoded, 'base64').toString('utf8');}
btoaLatin1('?'); // returns "Ew==" (would be preferable for it to throw error because this is undecodable)
atobLatin1(btoa('?')); // returns "\u0019" (END OF MEDIUM)
btoaUTF8('?'); // returns "4pyT"
atobUTF8(btoa('?')); // returns "?"
// returns "fvXmvA==", just like desktop Chrome
btoaLatin1(String.fromCharCode.apply(null, new Uint8Array([0x7e, 0xf5, 0xe6, 0xbc])));
// returns "fsO1w6bCvA=="
btoaUTF8(String.fromCharCode.apply(null, new Uint8Array([0x7e, 0xf5, 0xe6, 0xbc])));