ww/lib/bytes/bytes.ww

// bytes — slice operations over []u8. Mirrors Hare's bytes module
// (ref/hare/bytes/) for the in-tree subset: search/equality/prefix
// helpers used by lib/encoding, lib/bufio, lib/memio.
//
// Documented divergences from Hare:
//   - index_slice / rindex_slice use naive O(n·m); Hare specialises
//     2/3/4-byte needles and falls back to two_way (Crochemore-Perrin)
//     for longer (ref/hare/bytes/index.ha:61, ref/hare/bytes/two_way.ha).
//     Correctness equivalent.
//   - peek_token dispatches index/rindex by branching on `reverse`
//     rather than a function-pointer `ifunc` (ref/hare/bytes/tokenize.ha:97).
//     ww has no fn pointers in scope yet — same pattern as lib/strings
//     `move`. Outwardly identical.
//   - tokenize / rtokenize zero the `delim` field on the constructed
//     tokenizer when `in` is empty, rather than mutating the variadic
//     param before the struct write (ref/hare/bytes/tokenize.ha:26-28).
//     Semantically identical; the variadic param is borrowed and
//     captured-by-value into the struct, so mutating either side
//     yields the same observable state.

package bytes;

import os;
import types;

// done — iteration sentinel returned by next_token / peek_token at
// end-of-input. ref/hare/bytes/tokenize.ha uses the built-in `done`
// token; ww spells it per-package the same way lib/encoding/utf8 does
// (utf8.ww:36). Plain `void` (not `!void`): continuation signal.
export type done = void;

// tokenizer — cursor over an input slice. Layout mirrors
// ref/hare/bytes/tokenize.ha:6-10. `p` is the cached peek-position;
// I64_MAX (forward) / I64_MIN (reverse) are the unprimed sentinels.
// p < 0 also identifies a reverse-direction iterator.
export type tokenizer = struct {
	in:    []u8,
	delim: []u8,
	p:     i64,
};

// equal — true iff `a` and `b` have the same length and contents.
// ref/hare/bytes/equal.ha:9.
export fn equal(a: []u8, b: []u8) bool = {
	if (a.len != b.len) { return false; };
	let i: i32 = 0;
	for (i < a.len) {
		if (a[i] != b[i]) { return false; };
		i += 1;
	};
	return true;
};

// index — first offset of `needle` in `s`. u8 needle scans for the
// byte; []u8 needle scans for the substring. void if absent.
// ref/hare/bytes/index.ha:6.
export fn index(s: []u8, needle: (u8 | []u8)) (i32 | void) = {
	match (needle) {
	case let c: u8 => {
		let i: i32 = 0;
		for (i < s.len) {
			if (s[i] == c) { return i; };
			i += 1;
		};
		return;
	};
	case let sub: []u8 => {
		if (sub.len == 0) { return 0; };
		if (sub.len > s.len) { return; };
		let last: i32 = s.len - sub.len;
		let i: i32 = 0;
		for (i <= last) {
			let j: i32 = 0;
			let ok: bool = true;
			for (j < sub.len) {
				if (s[i + j] != sub[j]) { ok = false; j = sub.len; }
				else { j += 1; };
			};
			if (ok) { return i; };
			i += 1;
		};
		return;
	};
	};
	return;
};

// rindex — last offset of `needle` in `s`. Empty []u8 needle returns
// s.len (ref/hare/bytes/index.ha:103 — Hare's loop yields r-0 at i=0).
// ref/hare/bytes/index.ha:86.
export fn rindex(s: []u8, needle: (u8 | []u8)) (i32 | void) = {
	match (needle) {
	case let c: u8 => {
		let i: i32 = s.len - 1;
		for (i >= 0) {
			if (s[i] == c) { return i; };
			i -= 1;
		};
		return;
	};
	case let sub: []u8 => {
		if (sub.len == 0) { return s.len; };
		if (sub.len > s.len) { return; };
		let i: i32 = s.len - sub.len;
		for (i >= 0) {
			let j: i32 = 0;
			let ok: bool = true;
			for (j < sub.len) {
				if (s[i + j] != sub[j]) { ok = false; j = sub.len; }
				else { j += 1; };
			};
			if (ok) { return i; };
			i -= 1;
		};
		return;
	};
	};
	return;
};

// contains — true iff any of `needles` (byte or sub-slice) appears in `s`.
// ref/hare/bytes/contains.ha:6.
export fn contains(s: []u8, needles: (u8 | []u8)...) bool = {
	let i: i32 = 0;
	for (i < needles.len) {
		match (needles[i]) {
		case let b: u8   => {
			match (index(s, b)) {
			case let bo: i32 => return true;
			case void        => void;
			};
		};
		case let n: []u8 => {
			match (index(s, n)) {
			case let bo: i32 => return true;
			case void        => void;
			};
		};
		};
		i += 1;
	};
	return false;
};

// hasprefix — true iff `s` starts with `pre`.
// ref/hare/bytes/contains.ha:21.
export fn hasprefix(s: []u8, pre: []u8) bool = {
	if (pre.len > s.len) { return false; };
	let i: i32 = 0;
	for (i < pre.len) {
		if (s[i] != pre[i]) { return false; };
		i += 1;
	};
	return true;
};

// hassuffix — true iff `s` ends with `suf`.
// ref/hare/bytes/contains.ha:35.
export fn hassuffix(s: []u8, suf: []u8) bool = {
	if (suf.len > s.len) { return false; };
	let off: i32 = s.len - suf.len;
	let i: i32 = 0;
	for (i < suf.len) {
		if (s[off + i] != suf[i]) { return false; };
		i += 1;
	};
	return true;
};

// reverse — in-place reverse of `s`. ref/hare/bytes/reverse.ha:5.
export fn reverse(s: []u8) void = {
	let i: i32 = 0;
	let j: i32 = s.len - 1;
	for (i < j) {
		let t: u8 = s[i];
		s[i] = s[j];
		s[j] = t;
		i += 1;
		j -= 1;
	};
};

// zero — set every byte of `s` to 0. ref/hare/bytes/zero.ha:5.
export fn zero(s: []u8) void = {
	let i: i32 = 0;
	for (i < s.len) {
		s[i] = 0u8;
		i += 1;
	};
};

// tokenize — iterator yielding tokens from `in` separated by any byte
// in `delim`. Leading / trailing / adjacent delims yield empty tokens.
// `delim` is borrowed; caller keeps it valid for the tokenizer's
// lifetime. ref/hare/bytes/tokenize.ha:22.
export fn tokenize(in: []u8, delim: u8...) tokenizer = {
	os.assert(delim.len > 0, "bytes.tokenize called with empty slice");
	os.assert((in.len: i64) < types.I64_MAX,
		"bytes.tokenize: input length exceeds I64_MAX");
	let t: tokenizer;
	t.in = in;
	t.delim = delim;
	if (in.len == 0) {
		t.delim.len = 0;
		t.delim.cap = 0;
	};
	t.p = types.I64_MAX;
	return t;
};

// rtokenize — reverse-direction tokenize. First next_token yields the
// last token, last next_token yields the first. ref/hare/bytes/tokenize.ha:40.
export fn rtokenize(in: []u8, delim: u8...) tokenizer = {
	os.assert(delim.len > 0, "bytes.rtokenize called with empty slice");
	os.assert((in.len: i64) < types.I64_MAX,
		"bytes.rtokenize: input length exceeds I64_MAX");
	let t: tokenizer;
	t.in = in;
	t.delim = delim;
	if (in.len == 0) {
		t.delim.len = 0;
		t.delim.cap = 0;
	};
	t.p = types.I64_MIN;
	return t;
};

// peek_token — next token without advancing the cursor. Returns done
// once `s.delim` has been zeroed by a prior past-end next_token.
// ref/hare/bytes/tokenize.ha:91.
export fn peek_token(s: *tokenizer) ([]u8 | done) = {
	if (s.delim.len == 0) {
		let d: done; return d;
	};

	let reverse: bool = s.p < 0i64;
	let known: bool = false;
	if (reverse) {
		if (s.p != types.I64_MIN) { known = true; };
	} else {
		if (s.p != types.I64_MAX) { known = true; };
	};
	if (!known) {
		let i: i64 = types.I64_MAX;
		if (reverse) { i = types.I64_MIN; };
		let dlen: i64 = 0i64;
		let slen: i64 = s.in.len: i64;

		let k: i32 = 0;
		for (k < s.delim.len) {
			let d: u8 = s.delim[k];
			let ix_found: bool = false;
			let ix_val: i32 = 0;
			if (reverse) {
				match (rindex(s.in, d)) {
				case let v: i32 => { ix_found = true; ix_val = v; };
				case void       => void;
				};
			} else {
				match (index(s.in, d)) {
				case let v: i32 => { ix_found = true; ix_val = v; };
				case void       => void;
				};
			};
			if (ix_found) {
				if (!reverse) {
					if ((ix_val: i64) < i) { i = ix_val: i64; dlen = 1i64; };
				} else {
					if ((ix_val: i64) > i) { i = ix_val: i64; dlen = 1i64; };
				};
			} else {
				if (!reverse) {
					if (slen < i) { i = slen; };
				} else {
					if (0i64 > i) { i = 0i64; };
				};
			};
			k += 1;
		};

		if (reverse) {
			if (i == slen) {
				s.p = -(slen + 1i64);
			} else {
				s.p = i + dlen - slen - 1i64;
			};
		} else {
			s.p = i;
		};
	};

	let r: []u8;
	if (reverse) {
		let start: i32 = (s.in.len: i64 + s.p + 1i64): i32;
		r.ptr = s.in.ptr + (start: u64);
		r.len = s.in.len - start;
		r.cap = r.len;
	} else {
		let end: i32 = s.p: i32;
		r.ptr = s.in.ptr;
		r.len = end;
		r.cap = end;
	};
	return r;
};

// next_token — current token, then advance past it and the delim.
// Once the input is exhausted, returns done and zeros `s.delim` so
// subsequent peeks short-circuit. ref/hare/bytes/tokenize.ha:59.
export fn next_token(s: *tokenizer) ([]u8 | done) = {
	let b: []u8;
	match (peek_token(s)) {
	case let v: []u8 => { b = v; };
	case done        => { let d: done; return d; };
	};

	let slen: i64 = s.in.len: i64;
	let reverse: bool = s.p < 0i64;
	if (reverse) {
		if (slen + s.p + 1i64 == 0i64) {
			s.delim.len = 0;
			s.delim.cap = 0;
			s.in.len = 0;
			s.in.cap = 0;
		} else {
			let end: i32 = (slen + s.p + 1i64 - 1i64): i32;
			s.in.len = end;
			s.in.cap = end;
		};
		s.p = types.I64_MIN;
	} else {
		if (s.p == slen) {
			s.delim.len = 0;
			s.delim.cap = 0;
			s.in.len = 0;
			s.in.cap = 0;
		} else {
			let adv: u64 = (s.p: u64) + 1u64;
			let adv_i32: i32 = (s.p: i32) + 1;
			s.in.ptr = s.in.ptr + adv;
			s.in.len = s.in.len - adv_i32;
			s.in.cap = s.in.cap - adv_i32;
		};
		s.p = types.I64_MAX;
	};
	return b;
};

// remaining_tokens — the unconsumed portion of `s.in`. Read-only view.
// ref/hare/bytes/tokenize.ha:145.
export fn remaining_tokens(s: *tokenizer) []u8 = {
	return s.in;
};

// rt_ensure is the runtime slice-growth helper invoked by the
// `append(s, v)` builtin. We bind it directly because the builtin's
// expansion stores only 8 bytes of the new element (cgen emits a
// single MOVQ), losing the .len/.cap fields of a []u8 element (24B).
// Mirrors the same workaround in lib/shlex.shlex (appendstr, 16B) and
// lib/getopt.getopt (appendoption, 24B); collapses in one go when the
// append builtin learns to store the full element width.
@symbol("rt_ensure") fn rtensure(s: *void, membsz: u64) void;

// appendslice — grow `*slice` by one and store `item` (24B). Mirror
// of [[shlex.appendstr]] / [[getopt.appendoption]]. Bypasses the
// `append` builtin's first-8B-only-store gap for a slice-element.
fn appendslice(slice: *[][]u8, item: []u8) void = {
	let newlen: i32 = slice.len + 1;
	slice.len = newlen;
	rtensure(slice: *void, 24u64);
	let dst: *[]u8 = &slice.ptr[newlen - 1];
	dst.ptr = item.ptr;
	dst.len = item.len;
	dst.cap = item.cap;
};

// splitn — split `in` on any byte in `delim`, returning up to `n`
// tokens via forward iteration. The trailing slot (when more than
// `n - 1` tokens exist) holds the unconsumed remainder.
//
// The caller frees the returned slice via
// `os.free(r.ptr: *void, (r.cap: u64) * 24u64)`. Element bytes are
// borrowed from `in`.
//
// Hare's `([][]u8 | nomem)` collapses to `[][]u8` here: ww os.alloc
// has no recoverable failure path. Same precedent as
// shlex.split / getopt.tryparse.
//
// ref/hare/bytes/tokenize.ha:156.
export fn splitn(in: []u8, delim: []u8, n: i32) [][]u8 = {
	os.assert(delim.len > 0,
		"bytes.splitn must not be called with an empty delimiter");
	let toks: [][]u8;
	toks.ptr = nil: *[]u8;
	toks.len = 0;
	toks.cap = 0;
	let tok: tokenizer = tokenize(in, delim...);
	let i: i32 = 0;
	for (i < n - 1) {
		match (next_token(&tok)) {
		case let s: []u8 => { appendslice(&toks, s); };
		case done        => { return toks; };
		};
		i += 1;
	};
	match (peek_token(&tok)) {
	case done         => void;
	case let pk: []u8 => {
		let r: []u8 = remaining_tokens(&tok);
		appendslice(&toks, r);
	};
	};
	return toks;
};

// rsplitn — reverse-direction counterpart to [[splitn]]: tokens are
// collected from the end of `in`. The trailing slot holds the
// unconsumed prefix (everything before the n-th-from-last delim hit).
//
// When the input has fewer than n tokens, the `done` short-circuit
// returns toks UN-reversed (in last-token-first order). Mirrors Hare
// at ref/hare/bytes/tokenize.ha:196-199 where the in-place reverse
// step is gated behind the n-1 loop running to completion. Only the
// "loop ran to completion AND peek saw a remainder" path applies the
// reverse; both early-exit paths skip it.
//
// ref/hare/bytes/tokenize.ha:186.
export fn rsplitn(in: []u8, delim: []u8, n: i32) [][]u8 = {
	os.assert(delim.len > 0,
		"bytes.rsplitn called with empty delimiter");
	let toks: [][]u8;
	toks.ptr = nil: *[]u8;
	toks.len = 0;
	toks.cap = 0;
	let tok: tokenizer = rtokenize(in, delim...);
	let i: i32 = 0;
	for (i < n - 1) {
		match (next_token(&tok)) {
		case let s: []u8 => { appendslice(&toks, s); };
		case done        => { return toks; };
		};
		i += 1;
	};
	match (peek_token(&tok)) {
	case done         => void;
	case let pk: []u8 => {
		let r: []u8 = remaining_tokens(&tok);
		appendslice(&toks, r);
	};
	};

	// In-place reverse so callers see argv-order, matching Hare
	// (ref/hare/bytes/tokenize.ha:207). Element copy is field-wise
	// through `*[]u8` because `toks[i] = toks[j]` (full 24B slice
	// store) lands in the multi-word-store gap noted at
	// cmd/w6c/cgen.c:6515-6523.
	let a: i32 = 0;
	let b: i32 = toks.len - 1;
	for (a < b) {
		let pa: *[]u8 = &toks.ptr[a];
		let pb: *[]u8 = &toks.ptr[b];
		let tp: *u8 = pa.ptr;
		let tl: i32 = pa.len;
		let tc: i32 = pa.cap;
		pa.ptr = pb.ptr;
		pa.len = pb.len;
		pa.cap = pb.cap;
		pb.ptr = tp;
		pb.len = tl;
		pb.cap = tc;
		a += 1;
		b -= 1;
	};
	return toks;
};

// split — full split of `in` on `delim` (no token cap). Mirrors
// `splitn(in, delim, types::SIZE_MAX)`. ww uses `types.I32_MAX`
// because the index type is i32 (lib/CLAUDE.md).
//
// ref/hare/bytes/tokenize.ha:225.
export fn split(in: []u8, delim: []u8) [][]u8 = {
	return splitn(in, delim, types.I32_MAX);
};