ww/lib/encoding/base32/base32.ww

// Mirrors Hare's encoding::base32 surface, modulo Hare's stream-based
// encoder/decoder. ww ships the in-memory subset only: `encode(dst,
// src)` writes the encoded bytes into `dst`, returning the count;
// `decode(dst, src)` writes the decoded bytes into `dst`, returning a
// count or invalid.
//
// std uses 'A'-'Z' and '2'-'7' for indexes 0..31 (RFC 4648 §6); hex
// uses '0'-'9' and 'A'-'V' (the base32hex alphabet, RFC 4648 §7).
// Both pad encoded output with '=' to a multiple of 8 bytes.

package base32;

export type invalid = !i32;

// encodedsize — bytes required to encode `n` source bytes (including
// '=' padding). Hare names it the same.
export fn encodedsize(n: i32) i32 = {
	if (n == 0) { return 0; };
	return ((n - 1) / 5 + 1) * 8;
};

// decodedsize — upper bound on the number of bytes decoded from `n`
// encoded bytes.
export fn decodedsize(n: i32) i32 = {
	return (n / 8) * 5;
};

// encchar — map a 5-bit value to its alphabet character. `hex` picks
// the base32hex alphabet instead of std.
fn encchar(v: u8, hex: bool) u8 = {
	if (hex) {
		if (v < 10u8) { return v + 48u8; };	// '0' + v
		return v + 55u8;			// 'A' + (v - 10) = v + 55
	};
	if (v < 26u8) { return v + 65u8; };		// 'A' + v
	return v + 24u8;				// '2' + (v - 26) = v + 24
};

// decchar — inverse of encchar. Returns 0..31 or 255 on invalid char.
// '=' is handled in the decode loop, not here.
fn decchar(c: u8, hex: bool) u8 = {
	if (hex) {
		if (c >= 48u8) { if (c <= 57u8) { return c - 48u8; }; };	// '0'..'9'
		if (c >= 65u8) { if (c <= 86u8) { return c - 55u8; }; };	// 'A'..'V'
		return 255u8;
	};
	if (c >= 65u8) { if (c <= 90u8) { return c - 65u8; }; };		// 'A'..'Z'
	if (c >= 50u8) { if (c <= 55u8) { return c - 24u8; }; };		// '2'..'7'
	return 255u8;
};

fn encgroup(dst: []u8, j: i32, src: []u8, i: i32, n: i32, hex: bool) void = {
	let b0: u8 = 0u8;
	let b1: u8 = 0u8;
	let b2: u8 = 0u8;
	let b3: u8 = 0u8;
	let b4: u8 = 0u8;
	if (n > 0) { b0 = src[i]; };
	if (n > 1) { b1 = src[i + 1]; };
	if (n > 2) { b2 = src[i + 2]; };
	if (n > 3) { b3 = src[i + 3]; };
	if (n > 4) { b4 = src[i + 4]; };
	dst[j]     = encchar(b0 >> 3u8, hex);
	dst[j + 1] = encchar(((b0 & 7u8) << 2u8) | (b1 >> 6u8), hex);
	dst[j + 2] = encchar((b1 >> 1u8) & 31u8, hex);
	dst[j + 3] = encchar(((b1 & 1u8) << 4u8) | (b2 >> 4u8), hex);
	dst[j + 4] = encchar(((b2 & 15u8) << 1u8) | (b3 >> 7u8), hex);
	dst[j + 5] = encchar((b3 >> 2u8) & 31u8, hex);
	dst[j + 6] = encchar(((b3 & 3u8) << 3u8) | (b4 >> 5u8), hex);
	dst[j + 7] = encchar(b4 & 31u8, hex);
	// Pad the encoded slots that map past the source bytes.
	if (n < 5) {
		// n=1 → 2 chars then 6 '='. n=2 → 4 chars. n=3 → 5. n=4 → 7.
		let keep: i32 = 2;
		if (n == 2) { keep = 4; };
		if (n == 3) { keep = 5; };
		if (n == 4) { keep = 7; };
		let k: i32 = keep;
		for (k < 8) { dst[j + k] = '='; k += 1; };
	};
};

fn encodeinto(dst: []u8, src: []u8, hex: bool) i32 = {
	let i: i32 = 0;
	let j: i32 = 0;
	for (i + 5 <= src.len) {
		encgroup(dst, j, src, i, 5, hex);
		i += 5;
		j += 8;
	};
	let rem: i32 = src.len - i;
	if (rem > 0) {
		encgroup(dst, j, src, i, rem, hex);
		j += 8;
	};
	return j;
};

// encode — encode `src` into `dst` using the std (RFC 4648 §6)
// alphabet. Returns bytes written. `dst` must hold at least
// encodedsize(src.len) bytes.
export fn encode(dst: []u8, src: []u8) i32 = {
	return encodeinto(dst, src, false);
};

// encodehex — same as encode but uses the base32hex alphabet
// (RFC 4648 §7).
export fn encodehex(dst: []u8, src: []u8) i32 = {
	return encodeinto(dst, src, true);
};

// padcount — number of bytes encoded in the last group, given the
// count `p` of trailing '=' chars. RFC 4648 lists the legal mapping:
// 6=>1, 4=>2, 3=>3, 1=>4, 0=>5. Returns -1 if `p` isn't legal.
fn padcount(p: i32) i32 = {
	if (p == 0) { return 5; };
	if (p == 1) { return 4; };
	if (p == 3) { return 3; };
	if (p == 4) { return 2; };
	if (p == 6) { return 1; };
	return -1;
};

fn decodeinto(dst: []u8, src: []u8, hex: bool) (i32 | invalid) = {
	if (src.len == 0) { return 0; };
	if ((src.len & 7) != 0) { return src.len: invalid; };
	let i: i32 = 0;
	let j: i32 = 0;
	let end: i32 = src.len;
	for (i < end) {
		let v: [8]u8;
		let last: bool = false;
		let p: i32 = 0;
		let k: i32 = 0;
		for (k < 8) {
			let c: u8 = src[i + k];
			if (c == '=') {
				if (i + 8 != end) { return (i + k): invalid; };
				last = true;
				v[k] = 0u8;
				p += 1;
			} else {
				if (last) { return (i + k): invalid; };
				let d: u8 = decchar(c, hex);
				if (d == 255u8) { return (i + k): invalid; };
				v[k] = d;
			};
			k += 1;
		};
		let nb: i32 = 5;
		if (last) {
			nb = padcount(p);
			if (nb < 0) { return (i + 8 - p): invalid; };
		};
		// First two chars cover byte[0] (5 + 3 bits).
		if (nb > 0) {
			dst[j] = (v[0] << 3u8) | (v[1] >> 2u8);
		};
		if (nb > 1) {
			dst[j + 1] = (v[1] << 6u8) | (v[2] << 1u8) | (v[3] >> 4u8);
		};
		if (nb > 2) {
			dst[j + 2] = (v[3] << 4u8) | (v[4] >> 1u8);
		};
		if (nb > 3) {
			dst[j + 3] = (v[4] << 7u8) | (v[5] << 2u8) | (v[6] >> 3u8);
		};
		if (nb > 4) {
			dst[j + 4] = (v[6] << 5u8) | v[7];
		};
		j += nb;
		i += 8;
	};
	return j;
};

// decode — decode std-alphabet base32 from `src` into `dst`. Returns
// count of decoded bytes, or invalid on malformed input.
export fn decode(dst: []u8, src: []u8) (i32 | invalid) = {
	return decodeinto(dst, src, false);
};

// decodehex — same as decode but accepts base32hex.
export fn decodehex(dst: []u8, src: []u8) (i32 | invalid) = {
	return decodeinto(dst, src, true);
};