converting between Uint8Arrays and binary-encoded strings and word-arrays (for crypto purposes, with CryptoJS and NaCL)

gistfile1.js

/*
wordArray: { words: [..], sigBytes: words.length * 4 }
*/

// assumes wordArray is Big-Endian (because it comes from CryptoJS which is all BE)
// From: https://gist.github.com/creationix/07856504cf4d5cede5f9#file-encode-js
function convertWordArrayToUint8Array(wordArray) {
	var len = wordArray.words.length,
		u8_array = new Uint8Array(len << 2),
		offset = 0, word, i
	;
	for (i=0; i<len; i++) {
		word = wordArray.words[i];
		u8_array[offset++] = word >> 24;
		u8_array[offset++] = (word >> 16) & 0xff;
		u8_array[offset++] = (word >> 8) & 0xff;
		u8_array[offset++] = word & 0xff;
	}
	return u8_array;
}

// create a wordArray that is Big-Endian (because it's used with CryptoJS which is all BE)
// From: https://gist.github.com/creationix/07856504cf4d5cede5f9#file-encode-js
function convertUint8ArrayToWordArray(u8Array) {
	var words = [], i = 0, len = u8Array.length;

	while (i < len) {
		words.push(
			(u8Array[i++] << 24) |
			(u8Array[i++] << 16) |
			(u8Array[i++] << 8)  |
			(u8Array[i++])
		);
	}

	return {
		sigBytes: words.length * 4,
		words: words
	};
}

function convertUint8ArrayToBinaryString(u8Array) {
	var i, len = u8Array.length, b_str = "";
	for (i=0; i<len; i++) {
		b_str += String.fromCharCode(u8Array[i]);
	}
	return b_str;
}

function convertBinaryStringToUint8Array(bStr) {
	var i, len = bStr.length, u8_array = new Uint8Array(len);
	for (var i = 0; i < len; i++) {
		u8_array[i] = bStr.charCodeAt(i);
	}
	return u8_array;
}

For converting a Uint32Array to a normal Array, you can use the old Array.prototype.slice.call(arrayLike) trick.

@creationix are you talking about these lines?

Right.

w_array = {
  sigBytes: len,
  words: Array.prototype.slice.call(u32_array)
};

@creationix thanks, updated!

NOTE: had to update the gist to account for Big-Endianess coming out of CryptoJS lib (in its WordArray instances). Uint32Array instances subjected that data to a platform-dependent assumption of endianess, which was possibly unsafe. The above now goes directly from the explicitly (not assumed) BE of the WordArray to a non-endian Uint8Array (and vice versa).

convertWordArrayToUint8Array was creating an array that was slightly longer than the original wordarray for me. I replaced -
var len = wordArray.words.length, u8_array = new Uint8Array(len << 2),
with
u8_array = new Uint8Array(wordArray.sigBytes)
and this seems to have corrected the issue.

CryptoJS's WordArray has sigBytes for cases when length of uint8array is not divisible by 4. It should be the length of the original array and signals how much of the last word should be used.

This was also the source of @hygroovy's problem. Funfact: we're discussing this under a three-year-old topic :P

Thanks , Mr. Simpson. these help me a lot when play with msgpack and axios.

@pawelabrams Right, this caused troubles for me and should be changed to sigBytes: len

Big thanks for inspiration!

const formatU8Afrom = Uint8Array.from.bind(Uint8Array);

const formatFromCC = String.fromCharCode.bind(String);

export function stringToUint8Array(str: string) {
  return formatU8Afrom(str, (c) => c.charCodeAt(0));
}

export function uint8ArrayToString(u8a: Uint8Array) {
  // cf. https://stackoverflow.com/questions/12710001/how-to-convert-uint8-array-to-base64-encoded-string/12713326#12713326
  const maxargs = 0x1000;
  const strs = [];
  for (let i = 0, l = u8a.length; i < l; i += maxargs) {
    strs.push(formatFromCC(...u8a.subarray(i, i + maxargs)));
  }
  return strs.join('');
}