1 /* $NetBSD: lstrlib.c,v 1.21 2023/06/08 21:12:08 nikita Exp $ */
2
3 /*
4 ** Id: lstrlib.c
5 ** Standard library for string operations and pattern-matching
6 ** See Copyright Notice in lua.h
7 */
8
9 #define lstrlib_c
10 #define LUA_LIB
11
12 #include "lprefix.h"
13
14
15 #ifndef _KERNEL
16 #include <ctype.h>
17 #include <float.h>
18 #include <limits.h>
19 #include <locale.h>
20 #include <math.h>
21 #include <stddef.h>
22 #include <stdio.h>
23 #include <stdlib.h>
24 #include <string.h>
25 #endif /* _KERNEL */
26
27 #include "lua.h"
28
29 #include "lauxlib.h"
30 #include "lualib.h"
31
32
33 /*
34 ** maximum number of captures that a pattern can do during
35 ** pattern-matching. This limit is arbitrary, but must fit in
36 ** an unsigned char.
37 */
38 #if !defined(LUA_MAXCAPTURES)
39 #define LUA_MAXCAPTURES 32
40 #endif
41
42
43 /* macro to 'unsign' a character */
44 #define uchar(c) ((unsigned char)(c))
45
46
47 /*
48 ** Some sizes are better limited to fit in 'int', but must also fit in
49 ** 'size_t'. (We assume that 'lua_Integer' cannot be smaller than 'int'.)
50 */
51 #define MAX_SIZET ((size_t)(~(size_t)0))
52
53 #define MAXSIZE \
54 (sizeof(size_t) < sizeof(int) ? MAX_SIZET : (size_t)(INT_MAX))
55
56
57
58
str_len(lua_State * L)59 static int str_len (lua_State *L) {
60 size_t l;
61 luaL_checklstring(L, 1, &l);
62 lua_pushinteger(L, (lua_Integer)l);
63 return 1;
64 }
65
66
67 /*
68 ** translate a relative initial string position
69 ** (negative means back from end): clip result to [1, inf).
70 ** The length of any string in Lua must fit in a lua_Integer,
71 ** so there are no overflows in the casts.
72 ** The inverted comparison avoids a possible overflow
73 ** computing '-pos'.
74 */
posrelatI(lua_Integer pos,size_t len)75 static size_t posrelatI (lua_Integer pos, size_t len) {
76 if (pos > 0)
77 return (size_t)pos;
78 else if (pos == 0)
79 return 1;
80 else if (pos < -(lua_Integer)len) /* inverted comparison */
81 return 1; /* clip to 1 */
82 else return len + (size_t)pos + 1;
83 }
84
85
86 /*
87 ** Gets an optional ending string position from argument 'arg',
88 ** with default value 'def'.
89 ** Negative means back from end: clip result to [0, len]
90 */
getendpos(lua_State * L,int arg,lua_Integer def,size_t len)91 static size_t getendpos (lua_State *L, int arg, lua_Integer def,
92 size_t len) {
93 lua_Integer pos = luaL_optinteger(L, arg, def);
94 if (pos > (lua_Integer)len)
95 return len;
96 else if (pos >= 0)
97 return (size_t)pos;
98 else if (pos < -(lua_Integer)len)
99 return 0;
100 else return len + (size_t)pos + 1;
101 }
102
103
str_sub(lua_State * L)104 static int str_sub (lua_State *L) {
105 size_t l;
106 const char *s = luaL_checklstring(L, 1, &l);
107 size_t start = posrelatI(luaL_checkinteger(L, 2), l);
108 size_t end = getendpos(L, 3, -1, l);
109 if (start <= end)
110 lua_pushlstring(L, s + start - 1, (end - start) + 1);
111 else lua_pushliteral(L, "");
112 return 1;
113 }
114
115
str_reverse(lua_State * L)116 static int str_reverse (lua_State *L) {
117 size_t l, i;
118 luaL_Buffer b;
119 const char *s = luaL_checklstring(L, 1, &l);
120 char *p = luaL_buffinitsize(L, &b, l);
121 for (i = 0; i < l; i++)
122 p[i] = s[l - i - 1];
123 luaL_pushresultsize(&b, l);
124 return 1;
125 }
126
127
str_lower(lua_State * L)128 static int str_lower (lua_State *L) {
129 size_t l;
130 size_t i;
131 luaL_Buffer b;
132 const char *s = luaL_checklstring(L, 1, &l);
133 char *p = luaL_buffinitsize(L, &b, l);
134 for (i=0; i<l; i++)
135 p[i] = tolower(uchar(s[i]));
136 luaL_pushresultsize(&b, l);
137 return 1;
138 }
139
140
str_upper(lua_State * L)141 static int str_upper (lua_State *L) {
142 size_t l;
143 size_t i;
144 luaL_Buffer b;
145 const char *s = luaL_checklstring(L, 1, &l);
146 char *p = luaL_buffinitsize(L, &b, l);
147 for (i=0; i<l; i++)
148 p[i] = toupper(uchar(s[i]));
149 luaL_pushresultsize(&b, l);
150 return 1;
151 }
152
153
str_rep(lua_State * L)154 static int str_rep (lua_State *L) {
155 size_t l, lsep;
156 const char *s = luaL_checklstring(L, 1, &l);
157 lua_Integer n = luaL_checkinteger(L, 2);
158 const char *sep = luaL_optlstring(L, 3, "", &lsep);
159 if (n <= 0)
160 lua_pushliteral(L, "");
161 else if (l_unlikely(l + lsep < l || l + lsep > MAXSIZE / n))
162 return luaL_error(L, "resulting string too large");
163 else {
164 size_t totallen = (size_t)n * l + (size_t)(n - 1) * lsep;
165 luaL_Buffer b;
166 char *p = luaL_buffinitsize(L, &b, totallen);
167 while (n-- > 1) { /* first n-1 copies (followed by separator) */
168 memcpy(p, s, l * sizeof(char)); p += l;
169 if (lsep > 0) { /* empty 'memcpy' is not that cheap */
170 memcpy(p, sep, lsep * sizeof(char));
171 p += lsep;
172 }
173 }
174 memcpy(p, s, l * sizeof(char)); /* last copy (not followed by separator) */
175 luaL_pushresultsize(&b, totallen);
176 }
177 return 1;
178 }
179
180
str_byte(lua_State * L)181 static int str_byte (lua_State *L) {
182 size_t l;
183 const char *s = luaL_checklstring(L, 1, &l);
184 lua_Integer pi = luaL_optinteger(L, 2, 1);
185 size_t posi = posrelatI(pi, l);
186 size_t pose = getendpos(L, 3, pi, l);
187 int n, i;
188 if (posi > pose) return 0; /* empty interval; return no values */
189 if (l_unlikely(pose - posi >= (size_t)INT_MAX)) /* arithmetic overflow? */
190 return luaL_error(L, "string slice too long");
191 n = (int)(pose - posi) + 1;
192 luaL_checkstack(L, n, "string slice too long");
193 for (i=0; i<n; i++)
194 lua_pushinteger(L, uchar(s[posi+i-1]));
195 return n;
196 }
197
198
str_char(lua_State * L)199 static int str_char (lua_State *L) {
200 int n = lua_gettop(L); /* number of arguments */
201 int i;
202 luaL_Buffer b;
203 char *p = luaL_buffinitsize(L, &b, n);
204 for (i=1; i<=n; i++) {
205 lua_Unsigned c = (lua_Unsigned)luaL_checkinteger(L, i);
206 luaL_argcheck(L, c <= (lua_Unsigned)UCHAR_MAX, i, "value out of range");
207 p[i - 1] = uchar(c);
208 }
209 luaL_pushresultsize(&b, n);
210 return 1;
211 }
212
213
214 /*
215 ** Buffer to store the result of 'string.dump'. It must be initialized
216 ** after the call to 'lua_dump', to ensure that the function is on the
217 ** top of the stack when 'lua_dump' is called. ('luaL_buffinit' might
218 ** push stuff.)
219 */
220 struct str_Writer {
221 int init; /* true iff buffer has been initialized */
222 luaL_Buffer B;
223 };
224
225
writer(lua_State * L,const void * b,size_t size,void * ud)226 static int writer (lua_State *L, const void *b, size_t size, void *ud) {
227 struct str_Writer *state = (struct str_Writer *)ud;
228 if (!state->init) {
229 state->init = 1;
230 luaL_buffinit(L, &state->B);
231 }
232 luaL_addlstring(&state->B, (const char *)b, size);
233 return 0;
234 }
235
236
str_dump(lua_State * L)237 static int str_dump (lua_State *L) {
238 struct str_Writer state;
239 int strip = lua_toboolean(L, 2);
240 luaL_checktype(L, 1, LUA_TFUNCTION);
241 lua_settop(L, 1); /* ensure function is on the top of the stack */
242 state.init = 0;
243 if (l_unlikely(lua_dump(L, writer, &state, strip) != 0))
244 return luaL_error(L, "unable to dump given function");
245 luaL_pushresult(&state.B);
246 return 1;
247 }
248
249
250
251 /*
252 ** {======================================================
253 ** METAMETHODS
254 ** =======================================================
255 */
256
257 #if defined(LUA_NOCVTS2N) /* { */
258
259 /* no coercion from strings to numbers */
260
261 static const luaL_Reg stringmetamethods[] = {
262 {"__index", NULL}, /* placeholder */
263 {NULL, NULL}
264 };
265
266 #else /* }{ */
267
tonum(lua_State * L,int arg)268 static int tonum (lua_State *L, int arg) {
269 if (lua_type(L, arg) == LUA_TNUMBER) { /* already a number? */
270 lua_pushvalue(L, arg);
271 return 1;
272 }
273 else { /* check whether it is a numerical string */
274 size_t len;
275 const char *s = lua_tolstring(L, arg, &len);
276 return (s != NULL && lua_stringtonumber(L, s) == len + 1);
277 }
278 }
279
280
trymt(lua_State * L,const char * mtname)281 static void trymt (lua_State *L, const char *mtname) {
282 lua_settop(L, 2); /* back to the original arguments */
283 if (l_unlikely(lua_type(L, 2) == LUA_TSTRING ||
284 !luaL_getmetafield(L, 2, mtname)))
285 luaL_error(L, "attempt to %s a '%s' with a '%s'", mtname + 2,
286 luaL_typename(L, -2), luaL_typename(L, -1));
287 lua_insert(L, -3); /* put metamethod before arguments */
288 lua_call(L, 2, 1); /* call metamethod */
289 }
290
291
arith(lua_State * L,int op,const char * mtname)292 static int arith (lua_State *L, int op, const char *mtname) {
293 if (tonum(L, 1) && tonum(L, 2))
294 lua_arith(L, op); /* result will be on the top */
295 else
296 trymt(L, mtname);
297 return 1;
298 }
299
300
arith_add(lua_State * L)301 static int arith_add (lua_State *L) {
302 return arith(L, LUA_OPADD, "__add");
303 }
304
arith_sub(lua_State * L)305 static int arith_sub (lua_State *L) {
306 return arith(L, LUA_OPSUB, "__sub");
307 }
308
arith_mul(lua_State * L)309 static int arith_mul (lua_State *L) {
310 return arith(L, LUA_OPMUL, "__mul");
311 }
312
arith_mod(lua_State * L)313 static int arith_mod (lua_State *L) {
314 return arith(L, LUA_OPMOD, "__mod");
315 }
316
317 #ifndef _KERNEL
arith_pow(lua_State * L)318 static int arith_pow (lua_State *L) {
319 return arith(L, LUA_OPPOW, "__pow");
320 }
321
arith_div(lua_State * L)322 static int arith_div (lua_State *L) {
323 return arith(L, LUA_OPDIV, "__div");
324 }
325 #endif /* _KERNEL */
326
arith_idiv(lua_State * L)327 static int arith_idiv (lua_State *L) {
328 return arith(L, LUA_OPIDIV, "__idiv");
329 }
330
arith_unm(lua_State * L)331 static int arith_unm (lua_State *L) {
332 return arith(L, LUA_OPUNM, "__unm");
333 }
334
335
336 static const luaL_Reg stringmetamethods[] = {
337 {"__add", arith_add},
338 {"__sub", arith_sub},
339 {"__mul", arith_mul},
340 {"__mod", arith_mod},
341 #ifndef _KERNEL
342 {"__pow", arith_pow},
343 {"__div", arith_div},
344 #endif /* _KERNEL */
345 {"__idiv", arith_idiv},
346 {"__unm", arith_unm},
347 {"__index", NULL}, /* placeholder */
348 {NULL, NULL}
349 };
350
351 #endif /* } */
352
353 /* }====================================================== */
354
355 /*
356 ** {======================================================
357 ** PATTERN MATCHING
358 ** =======================================================
359 */
360
361
362 #define CAP_UNFINISHED (-1)
363 #define CAP_POSITION (-2)
364
365
366 typedef struct MatchState {
367 const char *src_init; /* init of source string */
368 const char *src_end; /* end ('\0') of source string */
369 const char *p_end; /* end ('\0') of pattern */
370 lua_State *L;
371 int matchdepth; /* control for recursive depth (to avoid C stack overflow) */
372 unsigned char level; /* total number of captures (finished or unfinished) */
373 struct {
374 const char *init;
375 ptrdiff_t len;
376 } capture[LUA_MAXCAPTURES];
377 } MatchState;
378
379
380 /* recursive function */
381 static const char *match (MatchState *ms, const char *s, const char *p);
382
383
384 /* maximum recursion depth for 'match' */
385 #if !defined(MAXCCALLS)
386 #define MAXCCALLS 200
387 #endif
388
389
390 #define L_ESC '%'
391 #define SPECIALS "^$*+?.([%-"
392
393
check_capture(MatchState * ms,int l)394 static int check_capture (MatchState *ms, int l) {
395 l -= '1';
396 if (l_unlikely(l < 0 || l >= ms->level ||
397 ms->capture[l].len == CAP_UNFINISHED))
398 return luaL_error(ms->L, "invalid capture index %%%d", l + 1);
399 return l;
400 }
401
402
capture_to_close(MatchState * ms)403 static int capture_to_close (MatchState *ms) {
404 int level = ms->level;
405 for (level--; level>=0; level--)
406 if (ms->capture[level].len == CAP_UNFINISHED) return level;
407 return luaL_error(ms->L, "invalid pattern capture");
408 }
409
410
classend(MatchState * ms,const char * p)411 static const char *classend (MatchState *ms, const char *p) {
412 switch (*p++) {
413 case L_ESC: {
414 if (l_unlikely(p == ms->p_end))
415 luaL_error(ms->L, "malformed pattern (ends with '%%')");
416 return p+1;
417 }
418 case '[': {
419 if (*p == '^') p++;
420 do { /* look for a ']' */
421 if (l_unlikely(p == ms->p_end))
422 luaL_error(ms->L, "malformed pattern (missing ']')");
423 if (*(p++) == L_ESC && p < ms->p_end)
424 p++; /* skip escapes (e.g. '%]') */
425 } while (*p != ']');
426 return p+1;
427 }
428 default: {
429 return p;
430 }
431 }
432 }
433
434
match_class(int c,int cl)435 static int match_class (int c, int cl) {
436 int res;
437 switch (tolower(cl)) {
438 case 'a' : res = isalpha(c); break;
439 case 'c' : res = iscntrl(c); break;
440 case 'd' : res = isdigit(c); break;
441 case 'g' : res = isgraph(c); break;
442 case 'l' : res = islower(c); break;
443 case 'p' : res = ispunct(c); break;
444 case 's' : res = isspace(c); break;
445 case 'u' : res = isupper(c); break;
446 case 'w' : res = isalnum(c); break;
447 case 'x' : res = isxdigit(c); break;
448 case 'z' : res = (c == 0); break; /* deprecated option */
449 default: return (cl == c);
450 }
451 return (islower(cl) ? res : !res);
452 }
453
454
matchbracketclass(int c,const char * p,const char * ec)455 static int matchbracketclass (int c, const char *p, const char *ec) {
456 int sig = 1;
457 if (*(p+1) == '^') {
458 sig = 0;
459 p++; /* skip the '^' */
460 }
461 while (++p < ec) {
462 if (*p == L_ESC) {
463 p++;
464 if (match_class(c, uchar(*p)))
465 return sig;
466 }
467 else if ((*(p+1) == '-') && (p+2 < ec)) {
468 p+=2;
469 if (uchar(*(p-2)) <= c && c <= uchar(*p))
470 return sig;
471 }
472 else if (uchar(*p) == c) return sig;
473 }
474 return !sig;
475 }
476
477
singlematch(MatchState * ms,const char * s,const char * p,const char * ep)478 static int singlematch (MatchState *ms, const char *s, const char *p,
479 const char *ep) {
480 if (s >= ms->src_end)
481 return 0;
482 else {
483 int c = uchar(*s);
484 switch (*p) {
485 case '.': return 1; /* matches any char */
486 case L_ESC: return match_class(c, uchar(*(p+1)));
487 case '[': return matchbracketclass(c, p, ep-1);
488 default: return (uchar(*p) == c);
489 }
490 }
491 }
492
493
matchbalance(MatchState * ms,const char * s,const char * p)494 static const char *matchbalance (MatchState *ms, const char *s,
495 const char *p) {
496 if (l_unlikely(p >= ms->p_end - 1))
497 luaL_error(ms->L, "malformed pattern (missing arguments to '%%b')");
498 if (*s != *p) return NULL;
499 else {
500 int b = *p;
501 int e = *(p+1);
502 int cont = 1;
503 while (++s < ms->src_end) {
504 if (*s == e) {
505 if (--cont == 0) return s+1;
506 }
507 else if (*s == b) cont++;
508 }
509 }
510 return NULL; /* string ends out of balance */
511 }
512
513
max_expand(MatchState * ms,const char * s,const char * p,const char * ep)514 static const char *max_expand (MatchState *ms, const char *s,
515 const char *p, const char *ep) {
516 ptrdiff_t i = 0; /* counts maximum expand for item */
517 while (singlematch(ms, s + i, p, ep))
518 i++;
519 /* keeps trying to match with the maximum repetitions */
520 while (i>=0) {
521 const char *res = match(ms, (s+i), ep+1);
522 if (res) return res;
523 i--; /* else didn't match; reduce 1 repetition to try again */
524 }
525 return NULL;
526 }
527
528
min_expand(MatchState * ms,const char * s,const char * p,const char * ep)529 static const char *min_expand (MatchState *ms, const char *s,
530 const char *p, const char *ep) {
531 for (;;) {
532 const char *res = match(ms, s, ep+1);
533 if (res != NULL)
534 return res;
535 else if (singlematch(ms, s, p, ep))
536 s++; /* try with one more repetition */
537 else return NULL;
538 }
539 }
540
541
start_capture(MatchState * ms,const char * s,const char * p,int what)542 static const char *start_capture (MatchState *ms, const char *s,
543 const char *p, int what) {
544 const char *res;
545 int level = ms->level;
546 if (level >= LUA_MAXCAPTURES) luaL_error(ms->L, "too many captures");
547 ms->capture[level].init = s;
548 ms->capture[level].len = what;
549 ms->level = level+1;
550 if ((res=match(ms, s, p)) == NULL) /* match failed? */
551 ms->level--; /* undo capture */
552 return res;
553 }
554
555
end_capture(MatchState * ms,const char * s,const char * p)556 static const char *end_capture (MatchState *ms, const char *s,
557 const char *p) {
558 int l = capture_to_close(ms);
559 const char *res;
560 ms->capture[l].len = s - ms->capture[l].init; /* close capture */
561 if ((res = match(ms, s, p)) == NULL) /* match failed? */
562 ms->capture[l].len = CAP_UNFINISHED; /* undo capture */
563 return res;
564 }
565
566
match_capture(MatchState * ms,const char * s,int l)567 static const char *match_capture (MatchState *ms, const char *s, int l) {
568 size_t len;
569 l = check_capture(ms, l);
570 len = ms->capture[l].len;
571 if ((size_t)(ms->src_end-s) >= len &&
572 memcmp(ms->capture[l].init, s, len) == 0)
573 return s+len;
574 else return NULL;
575 }
576
577
match(MatchState * ms,const char * s,const char * p)578 static const char *match (MatchState *ms, const char *s, const char *p) {
579 if (l_unlikely(ms->matchdepth-- == 0))
580 luaL_error(ms->L, "pattern too complex");
581 init: /* using goto to optimize tail recursion */
582 if (p != ms->p_end) { /* end of pattern? */
583 switch (*p) {
584 case '(': { /* start capture */
585 if (*(p + 1) == ')') /* position capture? */
586 s = start_capture(ms, s, p + 2, CAP_POSITION);
587 else
588 s = start_capture(ms, s, p + 1, CAP_UNFINISHED);
589 break;
590 }
591 case ')': { /* end capture */
592 s = end_capture(ms, s, p + 1);
593 break;
594 }
595 case '$': {
596 if ((p + 1) != ms->p_end) /* is the '$' the last char in pattern? */
597 goto dflt; /* no; go to default */
598 s = (s == ms->src_end) ? s : NULL; /* check end of string */
599 break;
600 }
601 case L_ESC: { /* escaped sequences not in the format class[*+?-]? */
602 switch (*(p + 1)) {
603 case 'b': { /* balanced string? */
604 s = matchbalance(ms, s, p + 2);
605 if (s != NULL) {
606 p += 4; goto init; /* return match(ms, s, p + 4); */
607 } /* else fail (s == NULL) */
608 break;
609 }
610 case 'f': { /* frontier? */
611 const char *ep; char previous;
612 p += 2;
613 if (l_unlikely(*p != '['))
614 luaL_error(ms->L, "missing '[' after '%%f' in pattern");
615 ep = classend(ms, p); /* points to what is next */
616 previous = (s == ms->src_init) ? '\0' : *(s - 1);
617 if (!matchbracketclass(uchar(previous), p, ep - 1) &&
618 matchbracketclass(uchar(*s), p, ep - 1)) {
619 p = ep; goto init; /* return match(ms, s, ep); */
620 }
621 s = NULL; /* match failed */
622 break;
623 }
624 case '0': case '1': case '2': case '3':
625 case '4': case '5': case '6': case '7':
626 case '8': case '9': { /* capture results (%0-%9)? */
627 s = match_capture(ms, s, uchar(*(p + 1)));
628 if (s != NULL) {
629 p += 2; goto init; /* return match(ms, s, p + 2) */
630 }
631 break;
632 }
633 default: goto dflt;
634 }
635 break;
636 }
637 default: dflt: { /* pattern class plus optional suffix */
638 const char *ep = classend(ms, p); /* points to optional suffix */
639 /* does not match at least once? */
640 if (!singlematch(ms, s, p, ep)) {
641 if (*ep == '*' || *ep == '?' || *ep == '-') { /* accept empty? */
642 p = ep + 1; goto init; /* return match(ms, s, ep + 1); */
643 }
644 else /* '+' or no suffix */
645 s = NULL; /* fail */
646 }
647 else { /* matched once */
648 switch (*ep) { /* handle optional suffix */
649 case '?': { /* optional */
650 const char *res;
651 if ((res = match(ms, s + 1, ep + 1)) != NULL)
652 s = res;
653 else {
654 p = ep + 1; goto init; /* else return match(ms, s, ep + 1); */
655 }
656 break;
657 }
658 case '+': /* 1 or more repetitions */
659 s++; /* 1 match already done */
660 /* FALLTHROUGH */
661 case '*': /* 0 or more repetitions */
662 s = max_expand(ms, s, p, ep);
663 break;
664 case '-': /* 0 or more repetitions (minimum) */
665 s = min_expand(ms, s, p, ep);
666 break;
667 default: /* no suffix */
668 s++; p = ep; goto init; /* return match(ms, s + 1, ep); */
669 }
670 }
671 break;
672 }
673 }
674 }
675 ms->matchdepth++;
676 return s;
677 }
678
679
680
lmemfind(const char * s1,size_t l1,const char * s2,size_t l2)681 static const char *lmemfind (const char *s1, size_t l1,
682 const char *s2, size_t l2) {
683 if (l2 == 0) return s1; /* empty strings are everywhere */
684 else if (l2 > l1) return NULL; /* avoids a negative 'l1' */
685 else {
686 const char *init; /* to search for a '*s2' inside 's1' */
687 l2--; /* 1st char will be checked by 'memchr' */
688 l1 = l1-l2; /* 's2' cannot be found after that */
689 while (l1 > 0 && (init = (const char *)memchr(s1, *s2, l1)) != NULL) {
690 init++; /* 1st char is already checked */
691 if (memcmp(init, s2+1, l2) == 0)
692 return init-1;
693 else { /* correct 'l1' and 's1' to try again */
694 l1 -= init-s1;
695 s1 = init;
696 }
697 }
698 return NULL; /* not found */
699 }
700 }
701
702
703 /*
704 ** get information about the i-th capture. If there are no captures
705 ** and 'i==0', return information about the whole match, which
706 ** is the range 's'..'e'. If the capture is a string, return
707 ** its length and put its address in '*cap'. If it is an integer
708 ** (a position), push it on the stack and return CAP_POSITION.
709 */
get_onecapture(MatchState * ms,int i,const char * s,const char * e,const char ** cap)710 static size_t get_onecapture (MatchState *ms, int i, const char *s,
711 const char *e, const char **cap) {
712 if (i >= ms->level) {
713 if (l_unlikely(i != 0))
714 luaL_error(ms->L, "invalid capture index %%%d", i + 1);
715 *cap = s;
716 return e - s;
717 }
718 else {
719 ptrdiff_t capl = ms->capture[i].len;
720 *cap = ms->capture[i].init;
721 if (l_unlikely(capl == CAP_UNFINISHED))
722 luaL_error(ms->L, "unfinished capture");
723 else if (capl == CAP_POSITION)
724 lua_pushinteger(ms->L, (ms->capture[i].init - ms->src_init) + 1);
725 return capl;
726 }
727 }
728
729
730 /*
731 ** Push the i-th capture on the stack.
732 */
push_onecapture(MatchState * ms,int i,const char * s,const char * e)733 static void push_onecapture (MatchState *ms, int i, const char *s,
734 const char *e) {
735 const char *cap;
736 ptrdiff_t l = get_onecapture(ms, i, s, e, &cap);
737 if (l != CAP_POSITION)
738 lua_pushlstring(ms->L, cap, l);
739 /* else position was already pushed */
740 }
741
742
push_captures(MatchState * ms,const char * s,const char * e)743 static int push_captures (MatchState *ms, const char *s, const char *e) {
744 int i;
745 int nlevels = (ms->level == 0 && s) ? 1 : ms->level;
746 luaL_checkstack(ms->L, nlevels, "too many captures");
747 for (i = 0; i < nlevels; i++)
748 push_onecapture(ms, i, s, e);
749 return nlevels; /* number of strings pushed */
750 }
751
752
753 /* check whether pattern has no special characters */
nospecials(const char * p,size_t l)754 static int nospecials (const char *p, size_t l) {
755 size_t upto = 0;
756 do {
757 if (strpbrk(p + upto, SPECIALS))
758 return 0; /* pattern has a special character */
759 upto += strlen(p + upto) + 1; /* may have more after \0 */
760 } while (upto <= l);
761 return 1; /* no special chars found */
762 }
763
764
prepstate(MatchState * ms,lua_State * L,const char * s,size_t ls,const char * p,size_t lp)765 static void prepstate (MatchState *ms, lua_State *L,
766 const char *s, size_t ls, const char *p, size_t lp) {
767 ms->L = L;
768 ms->matchdepth = MAXCCALLS;
769 ms->src_init = s;
770 ms->src_end = s + ls;
771 ms->p_end = p + lp;
772 }
773
774
reprepstate(MatchState * ms)775 static void reprepstate (MatchState *ms) {
776 ms->level = 0;
777 lua_assert(ms->matchdepth == MAXCCALLS);
778 }
779
780
str_find_aux(lua_State * L,int find)781 static int str_find_aux (lua_State *L, int find) {
782 size_t ls, lp;
783 const char *s = luaL_checklstring(L, 1, &ls);
784 const char *p = luaL_checklstring(L, 2, &lp);
785 size_t init = posrelatI(luaL_optinteger(L, 3, 1), ls) - 1;
786 if (init > ls) { /* start after string's end? */
787 luaL_pushfail(L); /* cannot find anything */
788 return 1;
789 }
790 /* explicit request or no special characters? */
791 if (find && (lua_toboolean(L, 4) || nospecials(p, lp))) {
792 /* do a plain search */
793 const char *s2 = lmemfind(s + init, ls - init, p, lp);
794 if (s2) {
795 lua_pushinteger(L, (s2 - s) + 1);
796 lua_pushinteger(L, (s2 - s) + lp);
797 return 2;
798 }
799 }
800 else {
801 MatchState ms;
802 const char *s1 = s + init;
803 int anchor = (*p == '^');
804 if (anchor) {
805 p++; lp--; /* skip anchor character */
806 }
807 prepstate(&ms, L, s, ls, p, lp);
808 do {
809 const char *res;
810 reprepstate(&ms);
811 if ((res=match(&ms, s1, p)) != NULL) {
812 if (find) {
813 lua_pushinteger(L, (s1 - s) + 1); /* start */
814 lua_pushinteger(L, res - s); /* end */
815 return push_captures(&ms, NULL, 0) + 2;
816 }
817 else
818 return push_captures(&ms, s1, res);
819 }
820 } while (s1++ < ms.src_end && !anchor);
821 }
822 luaL_pushfail(L); /* not found */
823 return 1;
824 }
825
826
str_find(lua_State * L)827 static int str_find (lua_State *L) {
828 return str_find_aux(L, 1);
829 }
830
831
str_match(lua_State * L)832 static int str_match (lua_State *L) {
833 return str_find_aux(L, 0);
834 }
835
836
837 /* state for 'gmatch' */
838 typedef struct GMatchState {
839 const char *src; /* current position */
840 const char *p; /* pattern */
841 const char *lastmatch; /* end of last match */
842 MatchState ms; /* match state */
843 } GMatchState;
844
845
gmatch_aux(lua_State * L)846 static int gmatch_aux (lua_State *L) {
847 GMatchState *gm = (GMatchState *)lua_touserdata(L, lua_upvalueindex(3));
848 const char *src;
849 gm->ms.L = L;
850 for (src = gm->src; src <= gm->ms.src_end; src++) {
851 const char *e;
852 reprepstate(&gm->ms);
853 if ((e = match(&gm->ms, src, gm->p)) != NULL && e != gm->lastmatch) {
854 gm->src = gm->lastmatch = e;
855 return push_captures(&gm->ms, src, e);
856 }
857 }
858 return 0; /* not found */
859 }
860
861
gmatch(lua_State * L)862 static int gmatch (lua_State *L) {
863 size_t ls, lp;
864 const char *s = luaL_checklstring(L, 1, &ls);
865 const char *p = luaL_checklstring(L, 2, &lp);
866 size_t init = posrelatI(luaL_optinteger(L, 3, 1), ls) - 1;
867 GMatchState *gm;
868 lua_settop(L, 2); /* keep strings on closure to avoid being collected */
869 gm = (GMatchState *)lua_newuserdatauv(L, sizeof(GMatchState), 0);
870 if (init > ls) /* start after string's end? */
871 init = ls + 1; /* avoid overflows in 's + init' */
872 prepstate(&gm->ms, L, s, ls, p, lp);
873 gm->src = s + init; gm->p = p; gm->lastmatch = NULL;
874 lua_pushcclosure(L, gmatch_aux, 3);
875 return 1;
876 }
877
878
add_s(MatchState * ms,luaL_Buffer * b,const char * s,const char * e)879 static void add_s (MatchState *ms, luaL_Buffer *b, const char *s,
880 const char *e) {
881 size_t l;
882 lua_State *L = ms->L;
883 const char *news = lua_tolstring(L, 3, &l);
884 const char *p;
885 while ((p = (char *)memchr(news, L_ESC, l)) != NULL) {
886 luaL_addlstring(b, news, p - news);
887 p++; /* skip ESC */
888 if (*p == L_ESC) /* '%%' */
889 luaL_addchar(b, *p);
890 else if (*p == '0') /* '%0' */
891 luaL_addlstring(b, s, e - s);
892 else if (isdigit(uchar(*p))) { /* '%n' */
893 const char *cap;
894 ptrdiff_t resl = get_onecapture(ms, *p - '1', s, e, &cap);
895 if (resl == CAP_POSITION)
896 luaL_addvalue(b); /* add position to accumulated result */
897 else
898 luaL_addlstring(b, cap, resl);
899 }
900 else
901 luaL_error(L, "invalid use of '%c' in replacement string", L_ESC);
902 l -= p + 1 - news;
903 news = p + 1;
904 }
905 luaL_addlstring(b, news, l);
906 }
907
908
909 /*
910 ** Add the replacement value to the string buffer 'b'.
911 ** Return true if the original string was changed. (Function calls and
912 ** table indexing resulting in nil or false do not change the subject.)
913 */
add_value(MatchState * ms,luaL_Buffer * b,const char * s,const char * e,int tr)914 static int add_value (MatchState *ms, luaL_Buffer *b, const char *s,
915 const char *e, int tr) {
916 lua_State *L = ms->L;
917 switch (tr) {
918 case LUA_TFUNCTION: { /* call the function */
919 int n;
920 lua_pushvalue(L, 3); /* push the function */
921 n = push_captures(ms, s, e); /* all captures as arguments */
922 lua_call(L, n, 1); /* call it */
923 break;
924 }
925 case LUA_TTABLE: { /* index the table */
926 push_onecapture(ms, 0, s, e); /* first capture is the index */
927 lua_gettable(L, 3);
928 break;
929 }
930 default: { /* LUA_TNUMBER or LUA_TSTRING */
931 add_s(ms, b, s, e); /* add value to the buffer */
932 return 1; /* something changed */
933 }
934 }
935 if (!lua_toboolean(L, -1)) { /* nil or false? */
936 lua_pop(L, 1); /* remove value */
937 luaL_addlstring(b, s, e - s); /* keep original text */
938 return 0; /* no changes */
939 }
940 else if (l_unlikely(!lua_isstring(L, -1)))
941 return luaL_error(L, "invalid replacement value (a %s)",
942 luaL_typename(L, -1));
943 else {
944 luaL_addvalue(b); /* add result to accumulator */
945 return 1; /* something changed */
946 }
947 }
948
949
str_gsub(lua_State * L)950 static int str_gsub (lua_State *L) {
951 size_t srcl, lp;
952 const char *src = luaL_checklstring(L, 1, &srcl); /* subject */
953 const char *p = luaL_checklstring(L, 2, &lp); /* pattern */
954 const char *lastmatch = NULL; /* end of last match */
955 int tr = lua_type(L, 3); /* replacement type */
956 lua_Integer max_s = luaL_optinteger(L, 4, srcl + 1); /* max replacements */
957 int anchor = (*p == '^');
958 lua_Integer n = 0; /* replacement count */
959 int changed = 0; /* change flag */
960 MatchState ms;
961 luaL_Buffer b;
962 luaL_argexpected(L, tr == LUA_TNUMBER || tr == LUA_TSTRING ||
963 tr == LUA_TFUNCTION || tr == LUA_TTABLE, 3,
964 "string/function/table");
965 luaL_buffinit(L, &b);
966 if (anchor) {
967 p++; lp--; /* skip anchor character */
968 }
969 prepstate(&ms, L, src, srcl, p, lp);
970 while (n < max_s) {
971 const char *e;
972 reprepstate(&ms); /* (re)prepare state for new match */
973 if ((e = match(&ms, src, p)) != NULL && e != lastmatch) { /* match? */
974 n++;
975 changed = add_value(&ms, &b, src, e, tr) | changed;
976 src = lastmatch = e;
977 }
978 else if (src < ms.src_end) /* otherwise, skip one character */
979 luaL_addchar(&b, *src++);
980 else break; /* end of subject */
981 if (anchor) break;
982 }
983 if (!changed) /* no changes? */
984 lua_pushvalue(L, 1); /* return original string */
985 else { /* something changed */
986 luaL_addlstring(&b, src, ms.src_end-src);
987 luaL_pushresult(&b); /* create and return new string */
988 }
989 lua_pushinteger(L, n); /* number of substitutions */
990 return 2;
991 }
992
993 /* }====================================================== */
994
995
996
997 /*
998 ** {======================================================
999 ** STRING FORMAT
1000 ** =======================================================
1001 */
1002
1003 #if !defined(lua_number2strx) /* { */
1004
1005 /*
1006 ** Hexadecimal floating-point formatter
1007 */
1008
1009 #define SIZELENMOD (sizeof(LUA_NUMBER_FRMLEN)/sizeof(char))
1010
1011
1012 /*
1013 ** Number of bits that goes into the first digit. It can be any value
1014 ** between 1 and 4; the following definition tries to align the number
1015 ** to nibble boundaries by making what is left after that first digit a
1016 ** multiple of 4.
1017 */
1018 #define L_NBFD ((l_floatatt(MANT_DIG) - 1)%4 + 1)
1019
1020
1021 /*
1022 ** Add integer part of 'x' to buffer and return new 'x'
1023 */
adddigit(char * buff,int n,lua_Number x)1024 static lua_Number adddigit (char *buff, int n, lua_Number x) {
1025 lua_Number dd = l_mathop(floor)(x); /* get integer part from 'x' */
1026 int d = (int)dd;
1027 buff[n] = (d < 10 ? d + '0' : d - 10 + 'a'); /* add to buffer */
1028 return x - dd; /* return what is left */
1029 }
1030
1031
num2straux(char * buff,int sz,lua_Number x)1032 static int num2straux (char *buff, int sz, lua_Number x) {
1033 /* if 'inf' or 'NaN', format it like '%g' */
1034 if (x != x || x == (lua_Number)HUGE_VAL || x == -(lua_Number)HUGE_VAL)
1035 return l_sprintf(buff, sz, LUA_NUMBER_FMT, (LUAI_UACNUMBER)x);
1036 else if (x == 0) { /* can be -0... */
1037 /* create "0" or "-0" followed by exponent */
1038 return l_sprintf(buff, sz, LUA_NUMBER_FMT "x0p+0", (LUAI_UACNUMBER)x);
1039 }
1040 else {
1041 int e;
1042 lua_Number m = l_mathop(frexp)(x, &e); /* 'x' fraction and exponent */
1043 int n = 0; /* character count */
1044 if (m < 0) { /* is number negative? */
1045 buff[n++] = '-'; /* add sign */
1046 m = -m; /* make it positive */
1047 }
1048 buff[n++] = '0'; buff[n++] = 'x'; /* add "0x" */
1049 m = adddigit(buff, n++, m * (1 << L_NBFD)); /* add first digit */
1050 e -= L_NBFD; /* this digit goes before the radix point */
1051 if (m > 0) { /* more digits? */
1052 buff[n++] = lua_getlocaledecpoint(); /* add radix point */
1053 do { /* add as many digits as needed */
1054 m = adddigit(buff, n++, m * 16);
1055 } while (m > 0);
1056 }
1057 n += l_sprintf(buff + n, sz - n, "p%+d", e); /* add exponent */
1058 lua_assert(n < sz);
1059 return n;
1060 }
1061 }
1062
1063
lua_number2strx(lua_State * L,char * buff,int sz,const char * fmt,lua_Number x)1064 static int lua_number2strx (lua_State *L, char *buff, int sz,
1065 const char *fmt, lua_Number x) {
1066 int n = num2straux(buff, sz, x);
1067 if (fmt[SIZELENMOD] == 'A') {
1068 int i;
1069 for (i = 0; i < n; i++)
1070 buff[i] = toupper(uchar(buff[i]));
1071 }
1072 else if (l_unlikely(fmt[SIZELENMOD] != 'a'))
1073 return luaL_error(L, "modifiers for format '%%a'/'%%A' not implemented");
1074 return n;
1075 }
1076
1077 #endif /* } */
1078
1079
1080 /*
1081 ** Maximum size for items formatted with '%f'. This size is produced
1082 ** by format('%.99f', -maxfloat), and is equal to 99 + 3 ('-', '.',
1083 ** and '\0') + number of decimal digits to represent maxfloat (which
1084 ** is maximum exponent + 1). (99+3+1, adding some extra, 110)
1085 */
1086 #define MAX_ITEMF (110 + l_floatatt(MAX_10_EXP))
1087
1088
1089 /*
1090 ** All formats except '%f' do not need that large limit. The other
1091 ** float formats use exponents, so that they fit in the 99 limit for
1092 ** significant digits; 's' for large strings and 'q' add items directly
1093 ** to the buffer; all integer formats also fit in the 99 limit. The
1094 ** worst case are floats: they may need 99 significant digits, plus
1095 ** '0x', '-', '.', 'e+XXXX', and '\0'. Adding some extra, 120.
1096 */
1097 #define MAX_ITEM 120
1098
1099
1100 /* valid flags in a format specification */
1101 #if !defined(L_FMTFLAGSF)
1102
1103 /* valid flags for a, A, e, E, f, F, g, and G conversions */
1104 #define L_FMTFLAGSF "-+#0 "
1105
1106 /* valid flags for o, x, and X conversions */
1107 #define L_FMTFLAGSX "-#0"
1108
1109 /* valid flags for d and i conversions */
1110 #define L_FMTFLAGSI "-+0 "
1111
1112 /* valid flags for u conversions */
1113 #define L_FMTFLAGSU "-0"
1114
1115 /* valid flags for c, p, and s conversions */
1116 #define L_FMTFLAGSC "-"
1117
1118 #endif
1119
1120
1121 /*
1122 ** Maximum size of each format specification (such as "%-099.99d"):
1123 ** Initial '%', flags (up to 5), width (2), period, precision (2),
1124 ** length modifier (8), conversion specifier, and final '\0', plus some
1125 ** extra.
1126 */
1127 #define MAX_FORMAT 32
1128
1129
addquoted(luaL_Buffer * b,const char * s,size_t len)1130 static void addquoted (luaL_Buffer *b, const char *s, size_t len) {
1131 luaL_addchar(b, '"');
1132 while (len--) {
1133 if (*s == '"' || *s == '\\' || *s == '\n') {
1134 luaL_addchar(b, '\\');
1135 luaL_addchar(b, *s);
1136 }
1137 else if (iscntrl(uchar(*s))) {
1138 char buff[10];
1139 if (!isdigit(uchar(*(s+1))))
1140 l_sprintf(buff, sizeof(buff), "\\%d", (int)uchar(*s));
1141 else
1142 l_sprintf(buff, sizeof(buff), "\\%03d", (int)uchar(*s));
1143 luaL_addstring(b, buff);
1144 }
1145 else
1146 luaL_addchar(b, *s);
1147 s++;
1148 }
1149 luaL_addchar(b, '"');
1150 }
1151
1152
1153 #ifndef _KERNEL
1154 /*
1155 ** Serialize a floating-point number in such a way that it can be
1156 ** scanned back by Lua. Use hexadecimal format for "common" numbers
1157 ** (to preserve precision); inf, -inf, and NaN are handled separately.
1158 ** (NaN cannot be expressed as a numeral, so we write '(0/0)' for it.)
1159 */
quotefloat(lua_State * L,char * buff,lua_Number n)1160 static int quotefloat (lua_State *L, char *buff, lua_Number n) {
1161 const char *s; /* for the fixed representations */
1162 if (n == (lua_Number)HUGE_VAL) /* inf? */
1163 s = "1e9999";
1164 else if (n == -(lua_Number)HUGE_VAL) /* -inf? */
1165 s = "-1e9999";
1166 else if (n != n) /* NaN? */
1167 s = "(0/0)";
1168 else { /* format number as hexadecimal */
1169 int nb = lua_number2strx(L, buff, MAX_ITEM,
1170 "%" LUA_NUMBER_FRMLEN "a", n);
1171 /* ensures that 'buff' string uses a dot as the radix character */
1172 if (memchr(buff, '.', nb) == NULL) { /* no dot? */
1173 char point = lua_getlocaledecpoint(); /* try locale point */
1174 char *ppoint = (char *)memchr(buff, point, nb);
1175 if (ppoint) *ppoint = '.'; /* change it to a dot */
1176 }
1177 return nb;
1178 }
1179 /* for the fixed representations */
1180 return l_sprintf(buff, MAX_ITEM, "%s", s);
1181 }
1182 #endif /* _KERNEL */
1183
1184
addliteral(lua_State * L,luaL_Buffer * b,int arg)1185 static void addliteral (lua_State *L, luaL_Buffer *b, int arg) {
1186 switch (lua_type(L, arg)) {
1187 case LUA_TSTRING: {
1188 size_t len;
1189 const char *s = lua_tolstring(L, arg, &len);
1190 addquoted(b, s, len);
1191 break;
1192 }
1193 case LUA_TNUMBER: {
1194 char *buff = luaL_prepbuffsize(b, MAX_ITEM);
1195 int nb;
1196 #ifndef _KERNEL
1197 if (!lua_isinteger(L, arg)) /* float? */
1198 nb = quotefloat(L, buff, lua_tonumber(L, arg));
1199 else { /* integers */
1200 #endif /* _KERNEL */
1201 lua_Integer n = lua_tointeger(L, arg);
1202 const char *format = (n == LUA_MININTEGER) /* corner case? */
1203 ? "0x%" LUA_INTEGER_FRMLEN "x" /* use hex */
1204 : LUA_INTEGER_FMT; /* else use default format */
1205 nb = l_sprintf(buff, MAX_ITEM, format, (LUAI_UACINT)n);
1206 #ifndef _KERNEL
1207 }
1208 #endif /* _KERNEL */
1209 luaL_addsize(b, nb);
1210 break;
1211 }
1212 case LUA_TNIL: case LUA_TBOOLEAN: {
1213 luaL_tolstring(L, arg, NULL);
1214 luaL_addvalue(b);
1215 break;
1216 }
1217 default: {
1218 luaL_argerror(L, arg, "value has no literal form");
1219 }
1220 }
1221 }
1222
1223
get2digits(const char * s)1224 static const char *get2digits (const char *s) {
1225 if (isdigit(uchar(*s))) {
1226 s++;
1227 if (isdigit(uchar(*s))) s++; /* (2 digits at most) */
1228 }
1229 return s;
1230 }
1231
1232
1233 /*
1234 ** Check whether a conversion specification is valid. When called,
1235 ** first character in 'form' must be '%' and last character must
1236 ** be a valid conversion specifier. 'flags' are the accepted flags;
1237 ** 'precision' signals whether to accept a precision.
1238 */
checkformat(lua_State * L,const char * form,const char * flags,int precision)1239 static void checkformat (lua_State *L, const char *form, const char *flags,
1240 int precision) {
1241 const char *spec = form + 1; /* skip '%' */
1242 spec += strspn(spec, flags); /* skip flags */
1243 if (*spec != '0') { /* a width cannot start with '0' */
1244 spec = get2digits(spec); /* skip width */
1245 if (*spec == '.' && precision) {
1246 spec++;
1247 spec = get2digits(spec); /* skip precision */
1248 }
1249 }
1250 if (!isalpha(uchar(*spec))) /* did not go to the end? */
1251 luaL_error(L, "invalid conversion specification: '%s'", form);
1252 }
1253
1254
1255 /*
1256 ** Get a conversion specification and copy it to 'form'.
1257 ** Return the address of its last character.
1258 */
getformat(lua_State * L,const char * strfrmt,char * form)1259 static const char *getformat (lua_State *L, const char *strfrmt,
1260 char *form) {
1261 /* spans flags, width, and precision ('0' is included as a flag) */
1262 size_t len = strspn(strfrmt, L_FMTFLAGSF "123456789.");
1263 len++; /* adds following character (should be the specifier) */
1264 /* still needs space for '%', '\0', plus a length modifier */
1265 if (len >= MAX_FORMAT - 10)
1266 luaL_error(L, "invalid format (too long)");
1267 *(form++) = '%';
1268 memcpy(form, strfrmt, len * sizeof(char));
1269 *(form + len) = '\0';
1270 return strfrmt + len - 1;
1271 }
1272
1273
1274 /*
1275 ** add length modifier into formats
1276 */
addlenmod(char * form,const char * lenmod)1277 static void addlenmod (char *form, const char *lenmod) {
1278 size_t l = strlen(form);
1279 size_t lm = strlen(lenmod);
1280 char spec = form[l - 1];
1281 strcpy(form + l - 1, lenmod);
1282 form[l + lm - 1] = spec;
1283 form[l + lm] = '\0';
1284 }
1285
1286
str_format(lua_State * L)1287 static int str_format (lua_State *L) {
1288 int top = lua_gettop(L);
1289 int arg = 1;
1290 size_t sfl;
1291 const char *strfrmt = luaL_checklstring(L, arg, &sfl);
1292 const char *strfrmt_end = strfrmt+sfl;
1293 const char *flags;
1294 luaL_Buffer b;
1295 luaL_buffinit(L, &b);
1296 while (strfrmt < strfrmt_end) {
1297 if (*strfrmt != L_ESC)
1298 luaL_addchar(&b, *strfrmt++);
1299 else if (*++strfrmt == L_ESC)
1300 luaL_addchar(&b, *strfrmt++); /* %% */
1301 else { /* format item */
1302 char form[MAX_FORMAT]; /* to store the format ('%...') */
1303 int maxitem = MAX_ITEM; /* maximum length for the result */
1304 char *buff = luaL_prepbuffsize(&b, maxitem); /* to put result */
1305 int nb = 0; /* number of bytes in result */
1306 if (++arg > top)
1307 return luaL_argerror(L, arg, "no value");
1308 strfrmt = getformat(L, strfrmt, form);
1309 switch (*strfrmt++) {
1310 case 'c': {
1311 checkformat(L, form, L_FMTFLAGSC, 0);
1312 nb = l_sprintf(buff, maxitem, form, (int)luaL_checkinteger(L, arg));
1313 break;
1314 }
1315 case 'd': case 'i':
1316 flags = L_FMTFLAGSI;
1317 goto intcase;
1318 case 'u':
1319 flags = L_FMTFLAGSU;
1320 goto intcase;
1321 case 'o': case 'x': case 'X':
1322 flags = L_FMTFLAGSX;
1323 intcase: {
1324 lua_Integer n = luaL_checkinteger(L, arg);
1325 checkformat(L, form, flags, 1);
1326 addlenmod(form, LUA_INTEGER_FRMLEN);
1327 nb = l_sprintf(buff, maxitem, form, (LUAI_UACINT)n);
1328 break;
1329 }
1330 #ifndef _KERNEL
1331 case 'a': case 'A':
1332 checkformat(L, form, L_FMTFLAGSF, 1);
1333 addlenmod(form, LUA_NUMBER_FRMLEN);
1334 nb = lua_number2strx(L, buff, maxitem, form,
1335 luaL_checknumber(L, arg));
1336 break;
1337 case 'f':
1338 maxitem = MAX_ITEMF; /* extra space for '%f' */
1339 buff = luaL_prepbuffsize(&b, maxitem);
1340 /* FALLTHROUGH */
1341 case 'e': case 'E': case 'g': case 'G': {
1342 lua_Number n = luaL_checknumber(L, arg);
1343 checkformat(L, form, L_FMTFLAGSF, 1);
1344 addlenmod(form, LUA_NUMBER_FRMLEN);
1345 nb = l_sprintf(buff, maxitem, form, (LUAI_UACNUMBER)n);
1346 break;
1347 }
1348 case 'p': {
1349 const void *p = lua_topointer(L, arg);
1350 checkformat(L, form, L_FMTFLAGSC, 0);
1351 if (p == NULL) { /* avoid calling 'printf' with argument NULL */
1352 p = "(null)"; /* result */
1353 form[strlen(form) - 1] = 's'; /* format it as a string */
1354 }
1355 nb = l_sprintf(buff, maxitem, form, p);
1356 break;
1357 }
1358 #endif /* _KERNEL */
1359 case 'q': {
1360 if (form[2] != '\0') /* modifiers? */
1361 return luaL_error(L, "specifier '%%q' cannot have modifiers");
1362 addliteral(L, &b, arg);
1363 break;
1364 }
1365 case 's': {
1366 size_t l;
1367 const char *s = luaL_tolstring(L, arg, &l);
1368 if (form[2] == '\0') /* no modifiers? */
1369 luaL_addvalue(&b); /* keep entire string */
1370 else {
1371 luaL_argcheck(L, l == strlen(s), arg, "string contains zeros");
1372 checkformat(L, form, L_FMTFLAGSC, 1);
1373 if (strchr(form, '.') == NULL && l >= 100) {
1374 /* no precision and string is too long to be formatted */
1375 luaL_addvalue(&b); /* keep entire string */
1376 }
1377 else { /* format the string into 'buff' */
1378 nb = l_sprintf(buff, maxitem, form, s);
1379 lua_pop(L, 1); /* remove result from 'luaL_tolstring' */
1380 }
1381 }
1382 break;
1383 }
1384 default: { /* also treat cases 'pnLlh' */
1385 return luaL_error(L, "invalid conversion '%s' to 'format'", form);
1386 }
1387 }
1388 lua_assert(nb < maxitem);
1389 luaL_addsize(&b, nb);
1390 }
1391 }
1392 luaL_pushresult(&b);
1393 return 1;
1394 }
1395
1396 /* }====================================================== */
1397
1398
1399 /*
1400 ** {======================================================
1401 ** PACK/UNPACK
1402 ** =======================================================
1403 */
1404
1405
1406 /* value used for padding */
1407 #if !defined(LUAL_PACKPADBYTE)
1408 #define LUAL_PACKPADBYTE 0x00
1409 #endif
1410
1411 /* maximum size for the binary representation of an integer */
1412 #define MAXINTSIZE 16
1413
1414 /* number of bits in a character */
1415 #define NB CHAR_BIT
1416
1417 /* mask for one character (NB 1's) */
1418 #define MC ((1 << NB) - 1)
1419
1420 /* size of a lua_Integer */
1421 #define SZINT ((int)sizeof(lua_Integer))
1422
1423
1424 /* dummy union to get native endianness */
1425 static const union {
1426 int dummy;
1427 char little; /* true iff machine is little endian */
1428 } nativeendian = {1};
1429
1430
1431 /*
1432 ** information to pack/unpack stuff
1433 */
1434 typedef struct Header {
1435 lua_State *L;
1436 int islittle;
1437 int maxalign;
1438 } Header;
1439
1440
1441 /*
1442 ** options for pack/unpack
1443 */
1444 typedef enum KOption {
1445 Kint, /* signed integers */
1446 Kuint, /* unsigned integers */
1447 #ifndef _KERNEL
1448 Kfloat, /* single-precision floating-point numbers */
1449 Knumber, /* Lua "native" floating-point numbers */
1450 Kdouble, /* double-precision floating-point numbers */
1451 #endif /* _KERNEL */
1452 Kchar, /* fixed-length strings */
1453 Kstring, /* strings with prefixed length */
1454 Kzstr, /* zero-terminated strings */
1455 Kpadding, /* padding */
1456 Kpaddalign, /* padding for alignment */
1457 Knop /* no-op (configuration or spaces) */
1458 } KOption;
1459
1460
1461 /*
1462 ** Read an integer numeral from string 'fmt' or return 'df' if
1463 ** there is no numeral
1464 */
digit(int c)1465 static int digit (int c) { return '0' <= c && c <= '9'; }
1466
getnum(const char ** fmt,int df)1467 static int getnum (const char **fmt, int df) {
1468 if (!digit(**fmt)) /* no number? */
1469 return df; /* return default value */
1470 else {
1471 int a = 0;
1472 do {
1473 a = a*10 + (*((*fmt)++) - '0');
1474 } while (digit(**fmt) && a <= ((int)MAXSIZE - 9)/10);
1475 return a;
1476 }
1477 }
1478
1479
1480 /*
1481 ** Read an integer numeral and raises an error if it is larger
1482 ** than the maximum size for integers.
1483 */
getnumlimit(Header * h,const char ** fmt,int df)1484 static int getnumlimit (Header *h, const char **fmt, int df) {
1485 int sz = getnum(fmt, df);
1486 if (l_unlikely(sz > MAXINTSIZE || sz <= 0))
1487 return luaL_error(h->L, "integral size (%d) out of limits [1,%d]",
1488 sz, MAXINTSIZE);
1489 return sz;
1490 }
1491
1492
1493 /*
1494 ** Initialize Header
1495 */
initheader(lua_State * L,Header * h)1496 static void initheader (lua_State *L, Header *h) {
1497 h->L = L;
1498 h->islittle = nativeendian.little;
1499 h->maxalign = 1;
1500 }
1501
1502
1503 /*
1504 ** Read and classify next option. 'size' is filled with option's size.
1505 */
getoption(Header * h,const char ** fmt,int * size)1506 static KOption getoption (Header *h, const char **fmt, int *size) {
1507 /* dummy structure to get native alignment requirements */
1508 struct cD { char c; union { LUAI_MAXALIGN; } u; };
1509 int opt = *((*fmt)++);
1510 *size = 0; /* default */
1511 switch (opt) {
1512 case 'b': *size = sizeof(char); return Kint;
1513 case 'B': *size = sizeof(char); return Kuint;
1514 case 'h': *size = sizeof(short); return Kint;
1515 case 'H': *size = sizeof(short); return Kuint;
1516 case 'l': *size = sizeof(long); return Kint;
1517 case 'L': *size = sizeof(long); return Kuint;
1518 case 'j': *size = sizeof(lua_Integer); return Kint;
1519 case 'J': *size = sizeof(lua_Integer); return Kuint;
1520 case 'T': *size = sizeof(size_t); return Kuint;
1521 #ifndef _KERNEL
1522 case 'f': *size = sizeof(float); return Kfloat;
1523 case 'd': *size = sizeof(double); return Kdouble;
1524 #endif
1525 #ifndef _KERNEL
1526 case 'n': *size = sizeof(lua_Number); return Knumber;
1527 #else /* _KERNEL */
1528 case 'n': *size = sizeof(lua_Number); return Kint;
1529 #endif
1530 case 'i': *size = getnumlimit(h, fmt, sizeof(int)); return Kint;
1531 case 'I': *size = getnumlimit(h, fmt, sizeof(int)); return Kuint;
1532 case 's': *size = getnumlimit(h, fmt, sizeof(size_t)); return Kstring;
1533 case 'c':
1534 *size = getnum(fmt, -1);
1535 if (l_unlikely(*size == -1))
1536 luaL_error(h->L, "missing size for format option 'c'");
1537 return Kchar;
1538 case 'z': return Kzstr;
1539 case 'x': *size = 1; return Kpadding;
1540 case 'X': return Kpaddalign;
1541 case ' ': break;
1542 case '<': h->islittle = 1; break;
1543 case '>': h->islittle = 0; break;
1544 case '=': h->islittle = nativeendian.little; break;
1545 case '!': {
1546 const int maxalign = offsetof(struct cD, u);
1547 h->maxalign = getnumlimit(h, fmt, maxalign);
1548 break;
1549 }
1550 default: luaL_error(h->L, "invalid format option '%c'", opt);
1551 }
1552 return Knop;
1553 }
1554
1555
1556 /*
1557 ** Read, classify, and fill other details about the next option.
1558 ** 'psize' is filled with option's size, 'notoalign' with its
1559 ** alignment requirements.
1560 ** Local variable 'size' gets the size to be aligned. (Kpadal option
1561 ** always gets its full alignment, other options are limited by
1562 ** the maximum alignment ('maxalign'). Kchar option needs no alignment
1563 ** despite its size.
1564 */
getdetails(Header * h,size_t totalsize,const char ** fmt,int * psize,int * ntoalign)1565 static KOption getdetails (Header *h, size_t totalsize,
1566 const char **fmt, int *psize, int *ntoalign) {
1567 KOption opt = getoption(h, fmt, psize);
1568 int align = *psize; /* usually, alignment follows size */
1569 if (opt == Kpaddalign) { /* 'X' gets alignment from following option */
1570 if (**fmt == '\0' || getoption(h, fmt, &align) == Kchar || align == 0)
1571 luaL_argerror(h->L, 1, "invalid next option for option 'X'");
1572 }
1573 if (align <= 1 || opt == Kchar) /* need no alignment? */
1574 *ntoalign = 0;
1575 else {
1576 if (align > h->maxalign) /* enforce maximum alignment */
1577 align = h->maxalign;
1578 if (l_unlikely((align & (align - 1)) != 0)) /* not a power of 2? */
1579 luaL_argerror(h->L, 1, "format asks for alignment not power of 2");
1580 *ntoalign = (align - (int)(totalsize & (align - 1))) & (align - 1);
1581 }
1582 return opt;
1583 }
1584
1585
1586 /*
1587 ** Pack integer 'n' with 'size' bytes and 'islittle' endianness.
1588 ** The final 'if' handles the case when 'size' is larger than
1589 ** the size of a Lua integer, correcting the extra sign-extension
1590 ** bytes if necessary (by default they would be zeros).
1591 */
packint(luaL_Buffer * b,lua_Unsigned n,int islittle,int size,int neg)1592 static void packint (luaL_Buffer *b, lua_Unsigned n,
1593 int islittle, int size, int neg) {
1594 char *buff = luaL_prepbuffsize(b, size);
1595 int i;
1596 buff[islittle ? 0 : size - 1] = (char)(n & MC); /* first byte */
1597 for (i = 1; i < size; i++) {
1598 n >>= NB;
1599 buff[islittle ? i : size - 1 - i] = (char)(n & MC);
1600 }
1601 if (neg && size > SZINT) { /* negative number need sign extension? */
1602 for (i = SZINT; i < size; i++) /* correct extra bytes */
1603 buff[islittle ? i : size - 1 - i] = (char)MC;
1604 }
1605 luaL_addsize(b, size); /* add result to buffer */
1606 }
1607
1608
1609 #ifndef _KERNEL
1610 /*
1611 ** Copy 'size' bytes from 'src' to 'dest', correcting endianness if
1612 ** given 'islittle' is different from native endianness.
1613 */
copywithendian(char * dest,const char * src,int size,int islittle)1614 static void copywithendian (char *dest, const char *src,
1615 int size, int islittle) {
1616 if (islittle == nativeendian.little)
1617 memcpy(dest, src, size);
1618 else {
1619 dest += size - 1;
1620 while (size-- != 0)
1621 *(dest--) = *(src++);
1622 }
1623 }
1624 #endif /* _KERNEL */
1625
1626
str_pack(lua_State * L)1627 static int str_pack (lua_State *L) {
1628 luaL_Buffer b;
1629 Header h;
1630 const char *fmt = luaL_checkstring(L, 1); /* format string */
1631 int arg = 1; /* current argument to pack */
1632 size_t totalsize = 0; /* accumulate total size of result */
1633 initheader(L, &h);
1634 lua_pushnil(L); /* mark to separate arguments from string buffer */
1635 luaL_buffinit(L, &b);
1636 while (*fmt != '\0') {
1637 int size, ntoalign;
1638 KOption opt = getdetails(&h, totalsize, &fmt, &size, &ntoalign);
1639 totalsize += ntoalign + size;
1640 while (ntoalign-- > 0)
1641 luaL_addchar(&b, LUAL_PACKPADBYTE); /* fill alignment */
1642 arg++;
1643 switch (opt) {
1644 case Kint: { /* signed integers */
1645 lua_Integer n = luaL_checkinteger(L, arg);
1646 if (size < SZINT) { /* need overflow check? */
1647 lua_Integer lim = (lua_Integer)1 << ((size * NB) - 1);
1648 luaL_argcheck(L, -lim <= n && n < lim, arg, "integer overflow");
1649 }
1650 packint(&b, (lua_Unsigned)n, h.islittle, size, (n < 0));
1651 break;
1652 }
1653 case Kuint: { /* unsigned integers */
1654 lua_Integer n = luaL_checkinteger(L, arg);
1655 if (size < SZINT) /* need overflow check? */
1656 luaL_argcheck(L, (lua_Unsigned)n < ((lua_Unsigned)1 << (size * NB)),
1657 arg, "unsigned overflow");
1658 packint(&b, (lua_Unsigned)n, h.islittle, size, 0);
1659 break;
1660 }
1661 #ifndef _KERNEL
1662 case Kfloat: { /* C float */
1663 float f = (float)luaL_checknumber(L, arg); /* get argument */
1664 char *buff = luaL_prepbuffsize(&b, sizeof(f));
1665 /* move 'f' to final result, correcting endianness if needed */
1666 copywithendian(buff, (char *)&f, sizeof(f), h.islittle);
1667 luaL_addsize(&b, size);
1668 break;
1669 }
1670 case Knumber: { /* Lua float */
1671 lua_Number f = luaL_checknumber(L, arg); /* get argument */
1672 char *buff = luaL_prepbuffsize(&b, sizeof(f));
1673 /* move 'f' to final result, correcting endianness if needed */
1674 copywithendian(buff, (char *)&f, sizeof(f), h.islittle);
1675 luaL_addsize(&b, size);
1676 break;
1677 }
1678 case Kdouble: { /* C double */
1679 double f = (double)luaL_checknumber(L, arg); /* get argument */
1680 char *buff = luaL_prepbuffsize(&b, sizeof(f));
1681 /* move 'f' to final result, correcting endianness if needed */
1682 copywithendian(buff, (char *)&f, sizeof(f), h.islittle);
1683 luaL_addsize(&b, size);
1684 break;
1685 }
1686 #endif /* _KERNEL */
1687 case Kchar: { /* fixed-size string */
1688 size_t len;
1689 const char *s = luaL_checklstring(L, arg, &len);
1690 luaL_argcheck(L, len <= (size_t)size, arg,
1691 "string longer than given size");
1692 luaL_addlstring(&b, s, len); /* add string */
1693 while (len++ < (size_t)size) /* pad extra space */
1694 luaL_addchar(&b, LUAL_PACKPADBYTE);
1695 break;
1696 }
1697 case Kstring: { /* strings with length count */
1698 size_t len;
1699 const char *s = luaL_checklstring(L, arg, &len);
1700 luaL_argcheck(L, size >= (int)sizeof(size_t) ||
1701 len < ((size_t)1 << (size * NB)),
1702 arg, "string length does not fit in given size");
1703 packint(&b, (lua_Unsigned)len, h.islittle, size, 0); /* pack length */
1704 luaL_addlstring(&b, s, len);
1705 totalsize += len;
1706 break;
1707 }
1708 case Kzstr: { /* zero-terminated string */
1709 size_t len;
1710 const char *s = luaL_checklstring(L, arg, &len);
1711 luaL_argcheck(L, strlen(s) == len, arg, "string contains zeros");
1712 luaL_addlstring(&b, s, len);
1713 luaL_addchar(&b, '\0'); /* add zero at the end */
1714 totalsize += len + 1;
1715 break;
1716 }
1717 case Kpadding: luaL_addchar(&b, LUAL_PACKPADBYTE); /* FALLTHROUGH */
1718 case Kpaddalign: case Knop:
1719 arg--; /* undo increment */
1720 break;
1721 }
1722 }
1723 luaL_pushresult(&b);
1724 return 1;
1725 }
1726
1727
str_packsize(lua_State * L)1728 static int str_packsize (lua_State *L) {
1729 Header h;
1730 const char *fmt = luaL_checkstring(L, 1); /* format string */
1731 size_t totalsize = 0; /* accumulate total size of result */
1732 initheader(L, &h);
1733 while (*fmt != '\0') {
1734 int size, ntoalign;
1735 KOption opt = getdetails(&h, totalsize, &fmt, &size, &ntoalign);
1736 luaL_argcheck(L, opt != Kstring && opt != Kzstr, 1,
1737 "variable-length format");
1738 size += ntoalign; /* total space used by option */
1739 luaL_argcheck(L, totalsize <= MAXSIZE - size, 1,
1740 "format result too large");
1741 totalsize += size;
1742 }
1743 lua_pushinteger(L, (lua_Integer)totalsize);
1744 return 1;
1745 }
1746
1747
1748 /*
1749 ** Unpack an integer with 'size' bytes and 'islittle' endianness.
1750 ** If size is smaller than the size of a Lua integer and integer
1751 ** is signed, must do sign extension (propagating the sign to the
1752 ** higher bits); if size is larger than the size of a Lua integer,
1753 ** it must check the unread bytes to see whether they do not cause an
1754 ** overflow.
1755 */
unpackint(lua_State * L,const char * str,int islittle,int size,int issigned)1756 static lua_Integer unpackint (lua_State *L, const char *str,
1757 int islittle, int size, int issigned) {
1758 lua_Unsigned res = 0;
1759 int i;
1760 int limit = (size <= SZINT) ? size : SZINT;
1761 for (i = limit - 1; i >= 0; i--) {
1762 res <<= NB;
1763 res |= (lua_Unsigned)(unsigned char)str[islittle ? i : size - 1 - i];
1764 }
1765 if (size < SZINT) { /* real size smaller than lua_Integer? */
1766 if (issigned) { /* needs sign extension? */
1767 lua_Unsigned mask = (lua_Unsigned)1 << (size*NB - 1);
1768 res = ((res ^ mask) - mask); /* do sign extension */
1769 }
1770 }
1771 else if (size > SZINT) { /* must check unread bytes */
1772 int mask = (!issigned || (lua_Integer)res >= 0) ? 0 : MC;
1773 for (i = limit; i < size; i++) {
1774 if (l_unlikely((unsigned char)str[islittle ? i : size - 1 - i] != mask))
1775 luaL_error(L, "%d-byte integer does not fit into Lua Integer", size);
1776 }
1777 }
1778 return (lua_Integer)res;
1779 }
1780
1781
str_unpack(lua_State * L)1782 static int str_unpack (lua_State *L) {
1783 Header h;
1784 const char *fmt = luaL_checkstring(L, 1);
1785 size_t ld;
1786 const char *data = luaL_checklstring(L, 2, &ld);
1787 size_t pos = posrelatI(luaL_optinteger(L, 3, 1), ld) - 1;
1788 int n = 0; /* number of results */
1789 luaL_argcheck(L, pos <= ld, 3, "initial position out of string");
1790 initheader(L, &h);
1791 while (*fmt != '\0') {
1792 int size, ntoalign;
1793 KOption opt = getdetails(&h, pos, &fmt, &size, &ntoalign);
1794 luaL_argcheck(L, (size_t)ntoalign + size <= ld - pos, 2,
1795 "data string too short");
1796 pos += ntoalign; /* skip alignment */
1797 /* stack space for item + next position */
1798 luaL_checkstack(L, 2, "too many results");
1799 n++;
1800 switch (opt) {
1801 case Kint:
1802 case Kuint: {
1803 lua_Integer res = unpackint(L, data + pos, h.islittle, size,
1804 (opt == Kint));
1805 lua_pushinteger(L, res);
1806 break;
1807 }
1808 #ifndef _KERNEL
1809 case Kfloat: {
1810 float f;
1811 copywithendian((char *)&f, data + pos, sizeof(f), h.islittle);
1812 lua_pushnumber(L, (lua_Number)f);
1813 break;
1814 }
1815 case Knumber: {
1816 lua_Number f;
1817 copywithendian((char *)&f, data + pos, sizeof(f), h.islittle);
1818 lua_pushnumber(L, f);
1819 break;
1820 }
1821 case Kdouble: {
1822 double f;
1823 copywithendian((char *)&f, data + pos, sizeof(f), h.islittle);
1824 lua_pushnumber(L, (lua_Number)f);
1825 break;
1826 }
1827 #endif /* _KERNEL */
1828 case Kchar: {
1829 lua_pushlstring(L, data + pos, size);
1830 break;
1831 }
1832 case Kstring: {
1833 size_t len = (size_t)unpackint(L, data + pos, h.islittle, size, 0);
1834 luaL_argcheck(L, len <= ld - pos - size, 2, "data string too short");
1835 lua_pushlstring(L, data + pos + size, len);
1836 pos += len; /* skip string */
1837 break;
1838 }
1839 case Kzstr: {
1840 size_t len = strlen(data + pos);
1841 luaL_argcheck(L, pos + len < ld, 2,
1842 "unfinished string for format 'z'");
1843 lua_pushlstring(L, data + pos, len);
1844 pos += len + 1; /* skip string plus final '\0' */
1845 break;
1846 }
1847 case Kpaddalign: case Kpadding: case Knop:
1848 n--; /* undo increment */
1849 break;
1850 }
1851 pos += size;
1852 }
1853 lua_pushinteger(L, pos + 1); /* next position */
1854 return n + 1;
1855 }
1856
1857 /* }====================================================== */
1858
1859
1860 static const luaL_Reg strlib[] = {
1861 {"byte", str_byte},
1862 {"char", str_char},
1863 {"dump", str_dump},
1864 {"find", str_find},
1865 {"format", str_format},
1866 {"gmatch", gmatch},
1867 {"gsub", str_gsub},
1868 {"len", str_len},
1869 {"lower", str_lower},
1870 {"match", str_match},
1871 {"rep", str_rep},
1872 {"reverse", str_reverse},
1873 {"sub", str_sub},
1874 {"upper", str_upper},
1875 {"pack", str_pack},
1876 {"packsize", str_packsize},
1877 {"unpack", str_unpack},
1878 {NULL, NULL}
1879 };
1880
1881
createmetatable(lua_State * L)1882 static void createmetatable (lua_State *L) {
1883 /* table to be metatable for strings */
1884 luaL_newlibtable(L, stringmetamethods);
1885 luaL_setfuncs(L, stringmetamethods, 0);
1886 lua_pushliteral(L, ""); /* dummy string */
1887 lua_pushvalue(L, -2); /* copy table */
1888 lua_setmetatable(L, -2); /* set table as metatable for strings */
1889 lua_pop(L, 1); /* pop dummy string */
1890 lua_pushvalue(L, -2); /* get string library */
1891 lua_setfield(L, -2, "__index"); /* metatable.__index = string */
1892 lua_pop(L, 1); /* pop metatable */
1893 }
1894
1895
1896 /*
1897 ** Open string library
1898 */
luaopen_string(lua_State * L)1899 LUAMOD_API int luaopen_string (lua_State *L) {
1900 luaL_newlib(L, strlib);
1901 createmetatable(L);
1902 return 1;
1903 }
1904
1905