lua_vm/object.rs
1//! Generic functions over Lua objects.
2//!
3//! Ported from `reference/lua-5.4.7/src/lobject.c` (602 lines, ~20 functions).
4
5// TODO(port): resolve import paths — all `crate::*` paths below are speculative;
6// Phase B will reconcile against the actual module tree.
7use crate::state::LuaState;
8#[allow(unused_imports)] use crate::prelude::*;
9use lua_types::{LuaValue, GcRef, LuaString, StackIdx};
10use lua_types::error::LuaError;
11use lua_types::arith::ArithOp;
12use lua_types::value::F2Imod;
13
14// ──────────────────────────────────────────────────────────────────────────
15// Module-level constants
16// ──────────────────────────────────────────────────────────────────────────
17
18/// Maximum number of significant hex digits to read (avoids overflow even for
19/// single-precision floats).
20const MAX_SIG_DIG: usize = 30;
21
22/// Maximum length of a numeral string accepted for conversion to a number.
23const L_MAX_LEN_NUM: usize = 200;
24
25/// Maximum size of a number-to-string conversion buffer.
26/// Accommodates both `%.14g` float formatting and `%lld` integer formatting.
27pub const MAX_NUMBER_2_STR: usize = 44;
28
29/// Buffer size (bytes) for UTF-8 encoding; encoded backwards into this buffer.
30pub const UTF8_BUF_SZ: usize = 8;
31
32/// Maximum length of a chunk source identifier in error messages.
33// TODO(port): verify against luaconf.h; defaulting to 60 here.
34pub const LUA_ID_SIZE: usize = 60;
35
36/// Internal buffer size for `push_vfstring`.
37const BUF_VFS: usize = LUA_ID_SIZE + MAX_NUMBER_2_STR + 95;
38
39/// Truncation marker for long chunk source strings.
40const RETS: &[u8] = b"...";
41
42/// Prefix for [string "..."] chunk identifiers.
43const PRE: &[u8] = b"[string \"";
44
45/// Suffix for [string "..."] chunk identifiers.
46const POS: &[u8] = b"\"]";
47
48// ──────────────────────────────────────────────────────────────────────────
49// ceil_log2
50// ──────────────────────────────────────────────────────────────────────────
51
52/// Computes `ceil(log2(x))`; returns the minimum `k` such that `2^k >= x`.
53///
54pub fn ceil_log2(x: u32) -> i32 {
55 static LOG_2: [u8; 256] = [
56 0,1,2,2,3,3,3,3,4,4,4,4,4,4,4,4,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,
57 6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,
58 7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,
59 7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,
60 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,
61 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,
62 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,
63 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,
64 ];
65 let mut l: i32 = 0;
66 let mut x = x.wrapping_sub(1);
67 while x >= 256 {
68 l += 8;
69 x >>= 8;
70 }
71 l + LOG_2[x as usize] as i32
72}
73
74// ──────────────────────────────────────────────────────────────────────────
75// Integer arithmetic dispatcher
76// ──────────────────────────────────────────────────────────────────────────
77
78/// Performs integer arithmetic for opcode `op` on operands `v1`, `v2`.
79/// Returns `Result` because floor-mod and floor-div can raise on zero divisor.
80///
81fn int_arith(state: &mut LuaState, op: ArithOp, v1: i64, v2: i64) -> Result<i64, LuaError> {
82 match op {
83 ArithOp::Add => Ok((v1 as u64).wrapping_add(v2 as u64) as i64),
84 ArithOp::Sub => Ok((v1 as u64).wrapping_sub(v2 as u64) as i64),
85 ArithOp::Mul => Ok((v1 as u64).wrapping_mul(v2 as u64) as i64),
86 // TODO(port): confirm function path for integer floor-mod in lvm.rs
87 ArithOp::Mod => crate::vm::int_floor_mod(state, v1, v2),
88 // TODO(port): confirm function path for integer floor-div in lvm.rs
89 ArithOp::Idiv => crate::vm::int_floor_div(state, v1, v2),
90 ArithOp::Band => Ok(v1 & v2),
91 ArithOp::Bor => Ok(v1 | v2),
92 ArithOp::Bxor => Ok(v1 ^ v2),
93 // TODO(port): confirm function path for shift-left in lvm.rs
94 ArithOp::Shl => Ok(crate::vm::shiftl(v1, v2)),
95 ArithOp::Shr => Ok(crate::vm::shiftl(v1, -v2)),
96 ArithOp::Unm => Ok((0u64).wrapping_sub(v1 as u64) as i64),
97 // l_castS2U(0) → 0u64, ~0u64 = 0xFFFFFFFFFFFFFFFF = !0u64
98 ArithOp::Bnot => Ok((!0u64 ^ v1 as u64) as i64),
99 _ => {
100 debug_assert!(false, "int_arith called with non-integer op");
101 Ok(0)
102 }
103 }
104}
105
106// ──────────────────────────────────────────────────────────────────────────
107// Float arithmetic dispatcher
108// ──────────────────────────────────────────────────────────────────────────
109
110/// Performs float arithmetic for opcode `op` on operands `v1`, `v2`.
111/// Returns `Result` because float floor-mod can raise on zero divisor.
112///
113fn float_arith(state: &mut LuaState, op: ArithOp, v1: f64, v2: f64) -> Result<f64, LuaError> {
114 match op {
115 ArithOp::Add => Ok(v1 + v2),
116 ArithOp::Sub => Ok(v1 - v2),
117 ArithOp::Mul => Ok(v1 * v2),
118 ArithOp::Div => Ok(v1 / v2),
119 ArithOp::Pow => Ok(if v2 == 2.0 { v1 * v1 } else { v1.powf(v2) }),
120 ArithOp::Idiv => Ok((v1 / v2).floor()),
121 ArithOp::Unm => Ok(-v1),
122 // TODO(port): confirm function path for float floor-mod in lvm.rs
123 ArithOp::Mod => crate::vm::float_floor_mod(state, v1, v2),
124 _ => {
125 debug_assert!(false, "float_arith called with non-float op");
126 Ok(0.0)
127 }
128 }
129}
130
131// ──────────────────────────────────────────────────────────────────────────
132// Raw arithmetic (no metamethods)
133// ──────────────────────────────────────────────────────────────────────────
134
135/// Attempts raw (no-metamethod) arithmetic on two Lua values.
136/// Writes the result to `res` and returns `true` on success, `false` if the
137/// operation cannot be performed with the given types (caller should invoke
138/// a metamethod instead).
139///
140pub fn raw_arith(
141 state: &mut LuaState,
142 op: ArithOp,
143 p1: &LuaValue,
144 p2: &LuaValue,
145 res: &mut LuaValue,
146) -> Result<bool, LuaError> {
147 match op {
148 // case LUA_OPSHL: case LUA_OPSHR: case LUA_OPBNOT: — integer-only ops
149 ArithOp::Band | ArithOp::Bor | ArithOp::Bxor
150 | ArithOp::Shl | ArithOp::Shr | ArithOp::Bnot => {
151 // setivalue(res, intarith(L, op, i1, i2)); return 1; }
152 // else return 0;
153 if let (Some(i1), Some(i2)) = (
154 p1.to_integer_no_strconv(),
155 p2.to_integer_no_strconv(),
156 ) {
157 *res = LuaValue::Int(int_arith(state, op, i1, i2)?);
158 Ok(true)
159 } else {
160 Ok(false)
161 }
162 }
163
164 ArithOp::Div | ArithOp::Pow => {
165 // setfltvalue(res, numarith(L, op, n1, n2)); return 1; }
166 // else return 0;
167 if let (Some(n1), Some(n2)) = (
168 p1.to_number_no_strconv(),
169 p2.to_number_no_strconv(),
170 ) {
171 *res = LuaValue::Float(float_arith(state, op, n1, n2)?);
172 Ok(true)
173 } else {
174 Ok(false)
175 }
176 }
177
178 _ => {
179 // setivalue(res, intarith(L, op, ivalue(p1), ivalue(p2))); return 1; }
180 if let (LuaValue::Int(i1), LuaValue::Int(i2)) = (p1, p2) {
181 *res = LuaValue::Int(int_arith(state, op, *i1, *i2)?);
182 return Ok(true);
183 }
184 if let (Some(n1), Some(n2)) = (
185 p1.to_number_no_strconv(),
186 p2.to_number_no_strconv(),
187 ) {
188 *res = LuaValue::Float(float_arith(state, op, n1, n2)?);
189 Ok(true)
190 } else {
191 Ok(false)
192 }
193 }
194 }
195}
196
197// ──────────────────────────────────────────────────────────────────────────
198// Arithmetic (with metamethod fallback)
199// ──────────────────────────────────────────────────────────────────────────
200
201/// Performs arithmetic for opcode `op`, writing the result to the stack slot
202/// `res`. Falls back to a binary tag-method if raw arithmetic is not possible.
203///
204pub fn arith(
205 state: &mut LuaState,
206 op: ArithOp,
207 p1: &LuaValue,
208 p2: &LuaValue,
209 res: StackIdx,
210) -> Result<(), LuaError> {
211 // luaT_trybinTM(L, p1, p2, res, cast(TMS, (op - LUA_OPADD) + TM_ADD)); }
212 //
213 // PORT NOTE: raw_arith writes to a local `temp` first; we then set the stack
214 // slot. This avoids holding a &mut borrow into the stack across try_bin_tm,
215 // which would violate the StackIdx rule (PORTING.md §2 #5).
216 let mut temp = LuaValue::Nil;
217 if raw_arith(state, op, p1, p2, &mut temp)? {
218 state.set_at(res, temp);
219 } else {
220 let _ = (p1, p2);
221 return Err(LuaError::runtime(format_args!(
222 "arithmetic metamethod dispatch not yet implemented for opcode {:?}", op
223 )));
224 }
225 Ok(())
226}
227
228// ──────────────────────────────────────────────────────────────────────────
229// hex_value
230// ──────────────────────────────────────────────────────────────────────────
231
232/// Converts a hexadecimal digit byte to its numeric value (0–15).
233/// Caller must ensure `c` is a valid hex digit.
234///
235pub fn hex_value(c: u8) -> u8 {
236 if c.is_ascii_digit() {
237 c - b'0'
238 } else {
239 c.to_ascii_lowercase() - b'a' + 10
240 }
241}
242
243// ──────────────────────────────────────────────────────────────────────────
244// Sign helper
245// ──────────────────────────────────────────────────────────────────────────
246
247/// Checks for and consumes a leading sign byte (`+` or `-`) in `s` starting
248/// at `*idx`. Returns `true` if a minus sign was consumed.
249///
250fn is_neg(s: &[u8], idx: &mut usize) -> bool {
251 // else if (**s == '+') (*s)++;
252 // return 0;
253 if *idx < s.len() && s[*idx] == b'-' {
254 *idx += 1;
255 return true;
256 }
257 if *idx < s.len() && s[*idx] == b'+' {
258 *idx += 1;
259 }
260 false
261}
262
263// ──────────────────────────────────────────────────────────────────────────
264// Hexadecimal float parser
265// ──────────────────────────────────────────────────────────────────────────
266
267/// Converts a hexadecimal float literal (C99 `0x…p…` form) in `s` to `f64`.
268/// Returns `Some((value, end_index))` on success, `None` on failure.
269///
270/// (conditionally compiled when the platform doesn't provide it)
271fn str_x2number(s: &[u8]) -> Option<(f64, usize)> {
272 let mut idx = 0;
273 while idx < s.len() && s[idx].is_ascii_whitespace() {
274 idx += 1;
275 }
276 let neg = is_neg(s, &mut idx);
277 if idx + 1 >= s.len() || s[idx] != b'0' || (s[idx + 1] != b'x' && s[idx + 1] != b'X') {
278 return None;
279 }
280 idx += 2;
281 let mut r: f64 = 0.0;
282 let mut sigdig: usize = 0;
283 let mut nosigdig: usize = 0;
284 let mut e: i32 = 0;
285 let mut hasdot = false;
286
287 // PORT NOTE: `lua_getlocaledecpoint()` returns the locale decimal separator.
288 // Rust has no locale; we always treat '.' as the separator here.
289 let dot = b'.';
290
291 loop {
292 if idx >= s.len() {
293 break;
294 }
295 let ch = s[idx];
296 if ch == dot {
297 if hasdot {
298 break;
299 }
300 hasdot = true;
301 } else if ch.is_ascii_hexdigit() {
302 // else if (++sigdig <= MAXSIGDIG) r = (r * 16.0) + luaO_hexavalue(*s);
303 // else e++;
304 // if (hasdot) e--;
305 if sigdig == 0 && ch == b'0' {
306 nosigdig += 1;
307 } else if {
308 sigdig += 1;
309 sigdig <= MAX_SIG_DIG
310 } {
311 r = r * 16.0 + hex_value(ch) as f64;
312 } else {
313 e += 1;
314 }
315 if hasdot {
316 e -= 1;
317 }
318 } else {
319 break;
320 }
321 idx += 1;
322 }
323
324 if nosigdig + sigdig == 0 {
325 return None;
326 }
327 // `idx` is now the valid end so far
328 let valid_end = idx;
329 e *= 4;
330
331 if idx < s.len() && (s[idx] == b'p' || s[idx] == b'P') {
332 idx += 1; // skip 'p'/'P'
333 let neg1 = is_neg(s, &mut idx);
334 if idx >= s.len() || !s[idx].is_ascii_digit() {
335 return None;
336 }
337 let mut exp1: i32 = 0;
338 while idx < s.len() && s[idx].is_ascii_digit() {
339 exp1 = exp1 * 10 + (s[idx] - b'0') as i32;
340 idx += 1;
341 }
342 if neg1 {
343 exp1 = -exp1;
344 }
345 e += exp1;
346 // update valid end: the exponent consumed up to here
347 // (valid_end is updated to idx below)
348 }
349 let result = if neg { -r } else { r };
350 Some((result * (2.0f64).powi(e), idx))
351}
352
353// ──────────────────────────────────────────────────────────────────────────
354// String-to-float helpers
355// ──────────────────────────────────────────────────────────────────────────
356
357/// Inner conversion: tries to parse the bytes `s` as a float using the given
358/// `mode` (`b'x'` for hex, anything else for decimal).
359/// Returns `Some((value, end_index))` or `None`.
360///
361fn str2dloc(s: &[u8], mode: u8) -> Option<(f64, usize)> {
362 let (result, end) = if mode == b'x' {
363 str_x2number(s)?
364 } else {
365 // PORT NOTE: from_utf8 used here because numeric string literals are
366 // guaranteed to be ASCII (a strict subset of UTF-8).
367 // TODO(port): replace with a bytes-native float parser in Phase B
368 // (e.g., the `fast-float` crate) to satisfy the from_utf8 ban fully.
369 let text = core::str::from_utf8(s).ok()?;
370 let trimmed = text.trim();
371 // Reject "inf", "infinity", "nan" — Lua does not accept these.
372 let lower = trimmed.to_ascii_lowercase();
373 if lower.starts_with("inf") || lower.starts_with("nan") {
374 return None;
375 }
376 let f: f64 = trimmed.parse().ok()?;
377 (f, s.len()) // strtod parses as many chars as possible; we consumed all
378 };
379 if end == 0 {
380 return None;
381 }
382 let mut end2 = end;
383 while end2 < s.len() && s[end2].is_ascii_whitespace() {
384 end2 += 1;
385 }
386 if end2 == s.len() {
387 Some((result, end2))
388 } else {
389 None
390 }
391}
392
393/// Converts bytes `s` to a Lua float value.
394/// Returns `Some((value, end_index))` on success, `None` on failure.
395///
396fn str2d(s: &[u8]) -> Option<(f64, usize)> {
397 // int mode = pmode ? ltolower(cast_uchar(*pmode)) : 0;
398 let pmode = s.iter().position(|&b| {
399 b == b'.' || b == b'x' || b == b'X' || b == b'n' || b == b'N'
400 });
401 let mode = pmode.map(|i| s[i].to_ascii_lowercase()).unwrap_or(0);
402
403 if mode == b'n' {
404 return None;
405 }
406
407 if let Some(result) = str2dloc(s, mode) {
408 return Some(result);
409 }
410
411 // PORT NOTE: Lua retries by replacing '.' with the locale decimal separator.
412 // Rust has no locale support; we skip this retry path and always use '.'.
413 // TODO(port): add locale retry if locale-aware float parsing is needed.
414
415 None
416}
417
418// ──────────────────────────────────────────────────────────────────────────
419// String-to-integer helper
420// ──────────────────────────────────────────────────────────────────────────
421
422/// Converts bytes `s` to a Lua integer value (decimal or `0x` hex).
423/// Returns `Some(value)` on success (the entire byte slice was consumed),
424/// `None` on failure or overflow.
425///
426fn str2int(s: &[u8]) -> Option<i64> {
427 let mut idx = 0;
428 while idx < s.len() && s[idx].is_ascii_whitespace() {
429 idx += 1;
430 }
431 let neg = is_neg(s, &mut idx);
432
433 let mut a: u64 = 0;
434 let mut empty = true;
435
436 if idx + 1 < s.len() && s[idx] == b'0' && (s[idx + 1] == b'x' || s[idx + 1] == b'X') {
437 idx += 2;
438 while idx < s.len() && s[idx].is_ascii_hexdigit() {
439 a = a.wrapping_mul(16).wrapping_add(hex_value(s[idx]) as u64);
440 empty = false;
441 idx += 1;
442 }
443 } else {
444 // MAXBY10 = cast(lua_Unsigned, LUA_MAXINTEGER / 10)
445 // MAXLASTD = cast_int(LUA_MAXINTEGER % 10)
446 // if (a >= MAXBY10 && (a > MAXBY10 || d > MAXLASTD + neg)) return NULL;
447 const MAX_BY10: u64 = (i64::MAX / 10) as u64;
448 const MAX_LAST_D: u64 = (i64::MAX % 10) as u64;
449 while idx < s.len() && s[idx].is_ascii_digit() {
450 let d = (s[idx] - b'0') as u64;
451 if a >= MAX_BY10 && (a > MAX_BY10 || d > MAX_LAST_D + if neg { 1 } else { 0 }) {
452 return None; // overflow
453 }
454 a = a.wrapping_mul(10).wrapping_add(d);
455 empty = false;
456 idx += 1;
457 }
458 }
459
460 while idx < s.len() && s[idx].is_ascii_whitespace() {
461 idx += 1;
462 }
463 if empty || idx != s.len() {
464 return None;
465 }
466 let result = if neg { (0u64).wrapping_sub(a) as i64 } else { a as i64 };
467 Some(result)
468}
469
470// ──────────────────────────────────────────────────────────────────────────
471// str2num — main public string-to-number conversion
472// ──────────────────────────────────────────────────────────────────────────
473
474/// Tries to convert the byte string `s` to a Lua number (integer first, then
475/// float). Writes the result to `o` and returns `consumed_bytes + 1` on
476/// success (matching the C convention of including the null terminator in the
477/// count), or `0` on failure.
478///
479pub fn str2num(s: &[u8], o: &mut LuaValue) -> usize {
480 if let Some(i) = str2int(s) {
481 *o = LuaValue::Int(i);
482 return s.len() + 1; // entire string consumed; +1 for C null-terminator convention
483 }
484 if let Some((n, end)) = str2d(s) {
485 *o = LuaValue::Float(n);
486 return end + 1;
487 }
488 0
489}
490
491// ──────────────────────────────────────────────────────────────────────────
492// UTF-8 encoder
493// ──────────────────────────────────────────────────────────────────────────
494
495/// Encodes Unicode codepoint `x` as UTF-8 into `buff` (filled backwards from
496/// index `UTF8_BUF_SZ - 1`). Returns the number of bytes written.
497/// The valid bytes occupy `buff[UTF8_BUF_SZ - n .. UTF8_BUF_SZ]`.
498///
499pub fn utf8_esc(buff: &mut [u8; UTF8_BUF_SZ], x: u32) -> usize {
500 debug_assert!(x <= 0x7FFF_FFFF, "codepoint out of range");
501 let mut n: usize = 1;
502 if x < 0x80 {
503 buff[UTF8_BUF_SZ - 1] = x as u8;
504 } else {
505 let mut mfb: u32 = 0x3f;
506 let mut x = x;
507 loop {
508 buff[UTF8_BUF_SZ - n] = 0x80 | (x & 0x3f) as u8;
509 n += 1;
510 x >>= 6;
511 mfb >>= 1;
512 if x <= mfb {
513 break;
514 }
515 }
516 buff[UTF8_BUF_SZ - n] = ((!mfb << 1) | x) as u8;
517 }
518 n
519}
520
521// ──────────────────────────────────────────────────────────────────────────
522// Number → string conversion
523// ──────────────────────────────────────────────────────────────────────────
524
525/// Formats `f` as C's `printf("%.14g", f)` would, returning the bytes.
526///
527/// PORT NOTE: Rust has no built-in `%g` format. This replicates the C99
528/// `%g` algorithm with precision 14: pick scientific or fixed-point based
529/// on the value's exponent, strip trailing zeros, normalize the exponent
530/// to `e[+-]NN` with at least two digits (matching C's output).
531fn fmt_g14(f: f64) -> Vec<u8> {
532 if f.is_nan() {
533 return b"nan".to_vec();
534 }
535 if f.is_infinite() {
536 return if f > 0.0 { b"inf".to_vec() } else { b"-inf".to_vec() };
537 }
538 if f == 0.0 {
539 return if f.is_sign_negative() { b"-0".to_vec() } else { b"0".to_vec() };
540 }
541
542 let precision: i32 = 14;
543 let abs = f.abs();
544 let exp = abs.log10().floor() as i32;
545
546 let s = if exp < -4 || exp >= precision {
547 let mantissa_decimals = (precision - 1) as usize;
548 let raw = format!("{:.*e}", mantissa_decimals, f);
549 let e_idx = raw.find('e').expect("Rust scientific format always contains 'e'");
550 let mantissa = strip_fixed_trailing_zeros(&raw[..e_idx]);
551 let exp_num: i32 = raw[e_idx + 1..].parse().expect("Rust formats integer exponents");
552 let sign = if exp_num < 0 { '-' } else { '+' };
553 let abs_exp = exp_num.abs();
554 if abs_exp < 10 {
555 format!("{}e{}0{}", mantissa, sign, abs_exp)
556 } else {
557 format!("{}e{}{}", mantissa, sign, abs_exp)
558 }
559 } else {
560 let decimals = (precision - 1 - exp).max(0) as usize;
561 let raw = format!("{:.*}", decimals, f);
562 strip_fixed_trailing_zeros(&raw)
563 };
564
565 s.into_bytes()
566}
567
568fn strip_fixed_trailing_zeros(s: &str) -> String {
569 if !s.contains('.') {
570 return s.to_string();
571 }
572 let mut out = s.to_string();
573 while out.ends_with('0') {
574 out.pop();
575 }
576 if out.ends_with('.') {
577 out.pop();
578 }
579 out
580}
581
582/// Formats the numeric `LuaValue` `val` (must be Int or Float) into a byte
583/// buffer and returns it.
584///
585fn number_to_str_buf(val: &LuaValue) -> Vec<u8> {
586 debug_assert!(
587 matches!(val, LuaValue::Int(_) | LuaValue::Float(_)),
588 "number_to_str_buf: value is not a number"
589 );
590
591 match val {
592 LuaValue::Int(i) => {
593 // lua_integer2str → l_sprintf with LUA_INTEGER_FMT ("%lld")
594 // PORT NOTE: using Rust's default i64 Display formatting, which
595 // matches C's `%lld` for all values in [i64::MIN, i64::MAX].
596 let s = format!("{}", i);
597 s.into_bytes()
598 }
599 LuaValue::Float(f) => {
600 let mut bytes = fmt_g14(*f);
601
602 let looks_like_int = bytes.iter().all(|&b| b == b'-' || b.is_ascii_digit());
603 if looks_like_int {
604 bytes.push(b'.');
605 bytes.push(b'0');
606 }
607 bytes
608 }
609 // Unreachable — guarded by debug_assert above.
610 _ => Vec::new(),
611 }
612}
613
614/// Converts a numeric `LuaValue` to an interned `LuaString`, returning a
615/// `GcRef<LuaString>` handle. Callers are responsible for updating the
616/// `LuaValue` (or stack slot) with `LuaValue::Str(s)`.
617///
618/// in place; in Rust we return the string because holding `&mut LuaValue`
619/// across a `state.intern_str` call would borrow `state` twice.
620pub fn num_to_string(state: &mut LuaState, val: &LuaValue) -> Result<GcRef<LuaString>, LuaError> {
621 // int len = tostringbuff(obj, buff);
622 // setsvalue(L, obj, luaS_newlstr(L, buff, len));
623 let bytes = number_to_str_buf(val);
624 // TODO(port): state.intern_str path needs to be confirmed in lua-vm
625 state.intern_str(&bytes)
626}
627
628// ──────────────────────────────────────────────────────────────────────────
629// push_vfstring infrastructure
630// ──────────────────────────────────────────────────────────────────────────
631
632/// Typed format argument for `push_vfstring`.
633///
634/// PORT NOTE: replaces the C `va_list` variadic interface. C callers of
635/// `luaO_pushfstring(L, fmt, ...)` must be updated to pass structured
636/// `FmtArg` slices. The format-string scanning logic is preserved in
637/// `push_vfstring`; only the argument-list type changes.
638pub enum FmtArg<'a> {
639 /// `%s` — a byte string (replaces `const char *` from va_list).
640 Str(&'a [u8]),
641 /// `%c` — a single byte character.
642 Char(u8),
643 /// `%d` — a 32-bit integer.
644 Int(i32),
645 /// `%I` — a Lua integer (i64).
646 LuaInt(i64),
647 /// `%f` — a Lua float (f64).
648 Float(f64),
649 /// `%U` — a Unicode codepoint (u32), encoded as UTF-8.
650 Utf8Codepoint(u32),
651 // TODO(port): %p (pointer) omitted — raw pointer in safe Rust is not allowed
652 // outside explicit unsafe-budget crates. Callers that need pointer formatting must handle
653 // it separately and pass the pre-formatted bytes as FmtArg::Str.
654}
655
656/// Internal accumulator for `push_vfstring`.
657///
658///
659/// PORT NOTE: `space` is a `Vec<u8>` rather than a fixed-size array; the
660/// BUF_VFS threshold is still respected for flushing behaviour.
661struct BufFs {
662 /// Whether at least one partial result has been pushed onto the stack.
663 pushed: bool,
664 /// Accumulated bytes not yet pushed to the stack.
665 space: Vec<u8>,
666}
667
668impl BufFs {
669 fn new() -> Self {
670 BufFs {
671 pushed: false,
672 space: Vec::with_capacity(BUF_VFS),
673 }
674 }
675}
676
677/// Pushes the byte string `str_bytes` to the Lua stack and concatenates with
678/// any prior partial result.
679///
680fn pushstr(buf: &mut BufFs, state: &mut LuaState, str_bytes: &[u8]) -> Result<(), LuaError> {
681 // L->top.p++;
682 // if (!buff->pushed) buff->pushed = 1;
683 // else luaV_concat(L, 2);
684 let s = state.intern_str(str_bytes)?;
685 state.push(LuaValue::Str(s));
686 if !buf.pushed {
687 buf.pushed = true;
688 } else {
689 // TODO(port): confirm path to string concatenation helper in lvm.rs
690 crate::vm::concat(state, 2)?;
691 }
692 Ok(())
693}
694
695/// Flushes the internal buffer to the Lua stack.
696///
697fn clearbuff(buf: &mut BufFs, state: &mut LuaState) -> Result<(), LuaError> {
698 let bytes: Vec<u8> = buf.space.drain(..).collect();
699 pushstr(buf, state, &bytes)
700}
701
702/// Adds `str_bytes` to the internal buffer, flushing first if it won't fit.
703///
704fn addstr2buff(buf: &mut BufFs, state: &mut LuaState, str_bytes: &[u8]) -> Result<(), LuaError> {
705 // else { clearbuff; pushstr directly; }
706 if str_bytes.len() <= BUF_VFS {
707 if str_bytes.len() > BUF_VFS - buf.space.len() {
708 clearbuff(buf, state)?;
709 }
710 buf.space.extend_from_slice(str_bytes);
711 } else {
712 clearbuff(buf, state)?;
713 pushstr(buf, state, str_bytes)?;
714 }
715 Ok(())
716}
717
718/// Formats the numeric value `num` and appends it to the buffer.
719///
720fn addnum2buff(buf: &mut BufFs, state: &mut LuaState, num: &LuaValue) -> Result<(), LuaError> {
721 // int len = tostringbuff(num, numbuff);
722 // addsize(buff, len);
723 let bytes = number_to_str_buf(num);
724 addstr2buff(buf, state, &bytes)
725}
726
727// ──────────────────────────────────────────────────────────────────────────
728// push_vfstring / push_fstring
729// ──────────────────────────────────────────────────────────────────────────
730
731/// Builds a formatted Lua string from a format byte string and structured
732/// arguments, pushes it onto the stack, and returns the top-of-stack value.
733///
734/// Supported format specifiers (same subset as C's `luaO_pushvfstring`):
735/// `%s`, `%c`, `%d`, `%I`, `%f`, `%U`, `%%`.
736/// `%p` is **not** supported; see [`FmtArg`] documentation.
737///
738///
739/// PORT NOTE: `va_list` replaced by `&[FmtArg]`. Call sites that previously
740/// passed variadic arguments must be updated to build a `&[FmtArg]` slice.
741pub fn push_vfstring<'a>(
742 state: &mut LuaState,
743 fmt: &[u8],
744 args: &[FmtArg<'a>],
745) -> Result<GcRef<LuaString>, LuaError> {
746 let mut buf = BufFs::new();
747 let mut arg_idx = 0usize;
748 let mut pos = 0usize;
749
750 while let Some(rel) = fmt[pos..].iter().position(|&b| b == b'%') {
751 let e = pos + rel;
752 addstr2buff(&mut buf, state, &fmt[pos..e])?;
753
754 let spec = if e + 1 < fmt.len() { fmt[e + 1] } else { 0 };
755 match spec {
756 b's' => {
757 // addstr2buff(&buff, s, strlen(s));
758 let s = match args.get(arg_idx) {
759 Some(FmtArg::Str(b)) => *b,
760 None => b"(null)",
761 _ => b"(null)",
762 };
763 arg_idx += 1;
764 addstr2buff(&mut buf, state, s)?;
765 }
766 b'c' => {
767 // addstr2buff(&buff, &c, sizeof(char));
768 let c = match args.get(arg_idx) {
769 Some(FmtArg::Char(b)) => *b,
770 _ => b'?',
771 };
772 arg_idx += 1;
773 addstr2buff(&mut buf, state, &[c])?;
774 }
775 b'd' => {
776 let n = match args.get(arg_idx) {
777 Some(FmtArg::Int(i)) => *i as i64,
778 _ => 0,
779 };
780 arg_idx += 1;
781 addnum2buff(&mut buf, state, &LuaValue::Int(n))?;
782 }
783 b'I' => {
784 // addnum2buff(&buff, &num);
785 let n = match args.get(arg_idx) {
786 Some(FmtArg::LuaInt(i)) => *i,
787 _ => 0,
788 };
789 arg_idx += 1;
790 addnum2buff(&mut buf, state, &LuaValue::Int(n))?;
791 }
792 b'f' => {
793 // addnum2buff(&buff, &num);
794 let f = match args.get(arg_idx) {
795 Some(FmtArg::Float(f)) => *f,
796 _ => 0.0,
797 };
798 arg_idx += 1;
799 addnum2buff(&mut buf, state, &LuaValue::Float(f))?;
800 }
801 b'p' => {
802 // TODO(port): %p pointer formatting not implemented in safe Rust;
803 // callers that need it should pre-format the pointer and pass FmtArg::Str.
804 arg_idx += 1; // consume the argument slot
805 addstr2buff(&mut buf, state, b"<ptr>")?;
806 }
807 b'U' => {
808 // addstr2buff(&buff, bf + UTF8BUFFSZ - len, len);
809 let cp = match args.get(arg_idx) {
810 Some(FmtArg::Utf8Codepoint(u)) => *u,
811 _ => b'?' as u32,
812 };
813 arg_idx += 1;
814 let mut bf = [0u8; UTF8_BUF_SZ];
815 let n = utf8_esc(&mut bf, cp);
816 addstr2buff(&mut buf, state, &bf[UTF8_BUF_SZ - n..])?;
817 }
818 b'%' => {
819 addstr2buff(&mut buf, state, b"%")?;
820 }
821 other => {
822 return Err(LuaError::runtime(format_args!(
823 "invalid option '%%{}' to 'lua_pushfstring'",
824 other as char
825 )));
826 }
827 }
828 pos = e + 2;
829 }
830
831 addstr2buff(&mut buf, state, &fmt[pos..])?;
832 clearbuff(&mut buf, state)?;
833 debug_assert!(buf.pushed, "push_vfstring: no string was pushed");
834
835 // Return the interned string at the top of the stack.
836 // PORT NOTE: in C this returns a `const char *` into the TString; in Rust
837 // we return the GcRef<LuaString> directly.
838 // TODO(port): state.peek_string_at_top() path needs to be confirmed.
839 Ok(state.peek_string_at_top())
840}
841
842/// Variadic entry point; delegates to `push_vfstring`.
843///
844///
845/// PORT NOTE: callers that previously used `luaO_pushfstring` for error
846/// messages should collapse the call into `LuaError::runtime(format_args!(...))`;
847/// see PORTING.md §4.2 and error_sites.tsv.
848pub fn push_fstring<'a>(
849 state: &mut LuaState,
850 fmt: &[u8],
851 args: &[FmtArg<'a>],
852) -> Result<GcRef<LuaString>, LuaError> {
853 push_vfstring(state, fmt, args)
854}
855
856// ──────────────────────────────────────────────────────────────────────────
857// chunk_id — human-readable chunk identifier
858// ──────────────────────────────────────────────────────────────────────────
859
860/// Fills `out` with a human-readable identifier derived from `source` and
861/// returns the number of bytes written (not including any null terminator).
862///
863/// Rules (matching C):
864/// - `=...` → literal text (everything after `=`), truncated to `LUA_ID_SIZE - 1`.
865/// - `@...` → file name (everything after `@`), prefixed with `...` if too long.
866/// - anything else → `[string "..."]`, with the first line truncated.
867///
868pub fn chunk_id(out: &mut [u8], source: &[u8]) -> usize {
869 let bufflen = LUA_ID_SIZE;
870 let mut written = 0usize;
871
872 let mut write_bytes = |out: &mut [u8], written: &mut usize, bytes: &[u8]| {
873 let avail = out.len().saturating_sub(*written);
874 let n = bytes.len().min(avail);
875 out[*written..*written + n].copy_from_slice(&bytes[..n]);
876 *written += n;
877 };
878
879 let first = source.first().copied();
880 let srclen = source.len();
881
882 match first {
883 Some(b'=') => {
884 let body = &source[1..];
885 if srclen <= bufflen {
886 write_bytes(out, &mut written, body);
887 } else {
888 write_bytes(out, &mut written, &body[..bufflen - 1]);
889 if written < out.len() {
890 out[written] = 0;
891 }
892 }
893 }
894 Some(b'@') => {
895 let body = &source[1..];
896 if srclen <= bufflen {
897 write_bytes(out, &mut written, body);
898 } else {
899 write_bytes(out, &mut written, RETS);
900 let tail_len = bufflen - RETS.len() - 1;
901 let tail_start = body.len() - tail_len;
902 write_bytes(out, &mut written, &body[tail_start..tail_start + tail_len]);
903 }
904 }
905 _ => {
906 let nl_pos = source.iter().position(|&b| b == b'\n');
907 write_bytes(out, &mut written, PRE);
908 let reserved = PRE.len() + RETS.len() + POS.len() + 1;
909 let inner_limit = bufflen.saturating_sub(reserved);
910
911 if srclen < inner_limit && nl_pos.is_none() {
912 write_bytes(out, &mut written, source);
913 } else {
914 let take = nl_pos.unwrap_or(srclen).min(inner_limit);
915 write_bytes(out, &mut written, &source[..take]);
916 write_bytes(out, &mut written, RETS);
917 }
918 write_bytes(out, &mut written, POS);
919 }
920 }
921
922 written
923}
924
925// ──────────────────────────────────────────────────────────────────────────
926// PORT STATUS
927// source: src/lobject.c (602 lines, ~20 functions)
928// target_crate: lua-vm
929// confidence: medium
930// todos: 15
931// port_notes: 12
932// unsafe_blocks: 0
933// notes: All import paths are speculative (crate::state, lua_types::*);
934// Phase B must reconcile. va_list replaced by FmtArg enum —
935// call sites of push_fstring/push_vfstring need updating.
936// Float formatting (%.14g) is approximated with {:.14e}; needs
937// proper %g in Phase B. Locale decimal-point handling is
938// stubbed (always '.'). str2dloc uses from_utf8 for ASCII
939// number strings (flagged TODO). int_floor_mod, int_floor_div,
940// shiftl, float_floor_mod, concat are assumed to exist in
941// crate::vm; Phase B must confirm or create them.
942// ──────────────────────────────────────────────────────────────────────────