xref: /freebsd/contrib/lua/src/lstrlib.c (revision 07d397d74651435ff6f4d474876589e85967350e)
18e3e3a7aSWarner Losh /*
28e3e3a7aSWarner Losh ** $Id: lstrlib.c,v 1.254 2016/12/22 13:08:50 roberto Exp $
38e3e3a7aSWarner Losh ** Standard library for string operations and pattern-matching
48e3e3a7aSWarner Losh ** See Copyright Notice in lua.h
58e3e3a7aSWarner Losh */
68e3e3a7aSWarner Losh 
78e3e3a7aSWarner Losh #define lstrlib_c
88e3e3a7aSWarner Losh #define LUA_LIB
98e3e3a7aSWarner Losh 
108e3e3a7aSWarner Losh #include "lprefix.h"
118e3e3a7aSWarner Losh 
128e3e3a7aSWarner Losh 
138e3e3a7aSWarner Losh #include <ctype.h>
148e3e3a7aSWarner Losh #include <float.h>
158e3e3a7aSWarner Losh #include <limits.h>
168e3e3a7aSWarner Losh #include <locale.h>
178e3e3a7aSWarner Losh #include <stddef.h>
188e3e3a7aSWarner Losh #include <stdio.h>
198e3e3a7aSWarner Losh #include <stdlib.h>
208e3e3a7aSWarner Losh #include <string.h>
218e3e3a7aSWarner Losh 
228e3e3a7aSWarner Losh #include "lua.h"
238e3e3a7aSWarner Losh 
248e3e3a7aSWarner Losh #include "lauxlib.h"
258e3e3a7aSWarner Losh #include "lualib.h"
268e3e3a7aSWarner Losh 
278e3e3a7aSWarner Losh 
288e3e3a7aSWarner Losh /*
298e3e3a7aSWarner Losh ** maximum number of captures that a pattern can do during
308e3e3a7aSWarner Losh ** pattern-matching. This limit is arbitrary, but must fit in
318e3e3a7aSWarner Losh ** an unsigned char.
328e3e3a7aSWarner Losh */
338e3e3a7aSWarner Losh #if !defined(LUA_MAXCAPTURES)
348e3e3a7aSWarner Losh #define LUA_MAXCAPTURES		32
358e3e3a7aSWarner Losh #endif
368e3e3a7aSWarner Losh 
378e3e3a7aSWarner Losh 
388e3e3a7aSWarner Losh /* macro to 'unsign' a character */
398e3e3a7aSWarner Losh #define uchar(c)	((unsigned char)(c))
408e3e3a7aSWarner Losh 
418e3e3a7aSWarner Losh 
428e3e3a7aSWarner Losh /*
438e3e3a7aSWarner Losh ** Some sizes are better limited to fit in 'int', but must also fit in
448e3e3a7aSWarner Losh ** 'size_t'. (We assume that 'lua_Integer' cannot be smaller than 'int'.)
458e3e3a7aSWarner Losh */
468e3e3a7aSWarner Losh #define MAX_SIZET	((size_t)(~(size_t)0))
478e3e3a7aSWarner Losh 
488e3e3a7aSWarner Losh #define MAXSIZE  \
498e3e3a7aSWarner Losh 	(sizeof(size_t) < sizeof(int) ? MAX_SIZET : (size_t)(INT_MAX))
508e3e3a7aSWarner Losh 
518e3e3a7aSWarner Losh 
528e3e3a7aSWarner Losh 
538e3e3a7aSWarner Losh 
548e3e3a7aSWarner Losh static int str_len (lua_State *L) {
558e3e3a7aSWarner Losh   size_t l;
568e3e3a7aSWarner Losh   luaL_checklstring(L, 1, &l);
578e3e3a7aSWarner Losh   lua_pushinteger(L, (lua_Integer)l);
588e3e3a7aSWarner Losh   return 1;
598e3e3a7aSWarner Losh }
608e3e3a7aSWarner Losh 
618e3e3a7aSWarner Losh 
628e3e3a7aSWarner Losh /* translate a relative string position: negative means back from end */
638e3e3a7aSWarner Losh static lua_Integer posrelat (lua_Integer pos, size_t len) {
648e3e3a7aSWarner Losh   if (pos >= 0) return pos;
658e3e3a7aSWarner Losh   else if (0u - (size_t)pos > len) return 0;
668e3e3a7aSWarner Losh   else return (lua_Integer)len + pos + 1;
678e3e3a7aSWarner Losh }
688e3e3a7aSWarner Losh 
698e3e3a7aSWarner Losh 
708e3e3a7aSWarner Losh static int str_sub (lua_State *L) {
718e3e3a7aSWarner Losh   size_t l;
728e3e3a7aSWarner Losh   const char *s = luaL_checklstring(L, 1, &l);
738e3e3a7aSWarner Losh   lua_Integer start = posrelat(luaL_checkinteger(L, 2), l);
748e3e3a7aSWarner Losh   lua_Integer end = posrelat(luaL_optinteger(L, 3, -1), l);
758e3e3a7aSWarner Losh   if (start < 1) start = 1;
768e3e3a7aSWarner Losh   if (end > (lua_Integer)l) end = l;
778e3e3a7aSWarner Losh   if (start <= end)
788e3e3a7aSWarner Losh     lua_pushlstring(L, s + start - 1, (size_t)(end - start) + 1);
798e3e3a7aSWarner Losh   else lua_pushliteral(L, "");
808e3e3a7aSWarner Losh   return 1;
818e3e3a7aSWarner Losh }
828e3e3a7aSWarner Losh 
838e3e3a7aSWarner Losh 
848e3e3a7aSWarner Losh static int str_reverse (lua_State *L) {
858e3e3a7aSWarner Losh   size_t l, i;
868e3e3a7aSWarner Losh   luaL_Buffer b;
878e3e3a7aSWarner Losh   const char *s = luaL_checklstring(L, 1, &l);
888e3e3a7aSWarner Losh   char *p = luaL_buffinitsize(L, &b, l);
898e3e3a7aSWarner Losh   for (i = 0; i < l; i++)
908e3e3a7aSWarner Losh     p[i] = s[l - i - 1];
918e3e3a7aSWarner Losh   luaL_pushresultsize(&b, l);
928e3e3a7aSWarner Losh   return 1;
938e3e3a7aSWarner Losh }
948e3e3a7aSWarner Losh 
958e3e3a7aSWarner Losh 
968e3e3a7aSWarner Losh static int str_lower (lua_State *L) {
978e3e3a7aSWarner Losh   size_t l;
988e3e3a7aSWarner Losh   size_t i;
998e3e3a7aSWarner Losh   luaL_Buffer b;
1008e3e3a7aSWarner Losh   const char *s = luaL_checklstring(L, 1, &l);
1018e3e3a7aSWarner Losh   char *p = luaL_buffinitsize(L, &b, l);
1028e3e3a7aSWarner Losh   for (i=0; i<l; i++)
1038e3e3a7aSWarner Losh     p[i] = tolower(uchar(s[i]));
1048e3e3a7aSWarner Losh   luaL_pushresultsize(&b, l);
1058e3e3a7aSWarner Losh   return 1;
1068e3e3a7aSWarner Losh }
1078e3e3a7aSWarner Losh 
1088e3e3a7aSWarner Losh 
1098e3e3a7aSWarner Losh static int str_upper (lua_State *L) {
1108e3e3a7aSWarner Losh   size_t l;
1118e3e3a7aSWarner Losh   size_t i;
1128e3e3a7aSWarner Losh   luaL_Buffer b;
1138e3e3a7aSWarner Losh   const char *s = luaL_checklstring(L, 1, &l);
1148e3e3a7aSWarner Losh   char *p = luaL_buffinitsize(L, &b, l);
1158e3e3a7aSWarner Losh   for (i=0; i<l; i++)
1168e3e3a7aSWarner Losh     p[i] = toupper(uchar(s[i]));
1178e3e3a7aSWarner Losh   luaL_pushresultsize(&b, l);
1188e3e3a7aSWarner Losh   return 1;
1198e3e3a7aSWarner Losh }
1208e3e3a7aSWarner Losh 
1218e3e3a7aSWarner Losh 
1228e3e3a7aSWarner Losh static int str_rep (lua_State *L) {
1238e3e3a7aSWarner Losh   size_t l, lsep;
1248e3e3a7aSWarner Losh   const char *s = luaL_checklstring(L, 1, &l);
1258e3e3a7aSWarner Losh   lua_Integer n = luaL_checkinteger(L, 2);
1268e3e3a7aSWarner Losh   const char *sep = luaL_optlstring(L, 3, "", &lsep);
1278e3e3a7aSWarner Losh   if (n <= 0) lua_pushliteral(L, "");
1288e3e3a7aSWarner Losh   else if (l + lsep < l || l + lsep > MAXSIZE / n)  /* may overflow? */
1298e3e3a7aSWarner Losh     return luaL_error(L, "resulting string too large");
1308e3e3a7aSWarner Losh   else {
1318e3e3a7aSWarner Losh     size_t totallen = (size_t)n * l + (size_t)(n - 1) * lsep;
1328e3e3a7aSWarner Losh     luaL_Buffer b;
1338e3e3a7aSWarner Losh     char *p = luaL_buffinitsize(L, &b, totallen);
1348e3e3a7aSWarner Losh     while (n-- > 1) {  /* first n-1 copies (followed by separator) */
1358e3e3a7aSWarner Losh       memcpy(p, s, l * sizeof(char)); p += l;
1368e3e3a7aSWarner Losh       if (lsep > 0) {  /* empty 'memcpy' is not that cheap */
1378e3e3a7aSWarner Losh         memcpy(p, sep, lsep * sizeof(char));
1388e3e3a7aSWarner Losh         p += lsep;
1398e3e3a7aSWarner Losh       }
1408e3e3a7aSWarner Losh     }
1418e3e3a7aSWarner Losh     memcpy(p, s, l * sizeof(char));  /* last copy (not followed by separator) */
1428e3e3a7aSWarner Losh     luaL_pushresultsize(&b, totallen);
1438e3e3a7aSWarner Losh   }
1448e3e3a7aSWarner Losh   return 1;
1458e3e3a7aSWarner Losh }
1468e3e3a7aSWarner Losh 
1478e3e3a7aSWarner Losh 
1488e3e3a7aSWarner Losh static int str_byte (lua_State *L) {
1498e3e3a7aSWarner Losh   size_t l;
1508e3e3a7aSWarner Losh   const char *s = luaL_checklstring(L, 1, &l);
1518e3e3a7aSWarner Losh   lua_Integer posi = posrelat(luaL_optinteger(L, 2, 1), l);
1528e3e3a7aSWarner Losh   lua_Integer pose = posrelat(luaL_optinteger(L, 3, posi), l);
1538e3e3a7aSWarner Losh   int n, i;
1548e3e3a7aSWarner Losh   if (posi < 1) posi = 1;
1558e3e3a7aSWarner Losh   if (pose > (lua_Integer)l) pose = l;
1568e3e3a7aSWarner Losh   if (posi > pose) return 0;  /* empty interval; return no values */
1578e3e3a7aSWarner Losh   if (pose - posi >= INT_MAX)  /* arithmetic overflow? */
1588e3e3a7aSWarner Losh     return luaL_error(L, "string slice too long");
1598e3e3a7aSWarner Losh   n = (int)(pose -  posi) + 1;
1608e3e3a7aSWarner Losh   luaL_checkstack(L, n, "string slice too long");
1618e3e3a7aSWarner Losh   for (i=0; i<n; i++)
1628e3e3a7aSWarner Losh     lua_pushinteger(L, uchar(s[posi+i-1]));
1638e3e3a7aSWarner Losh   return n;
1648e3e3a7aSWarner Losh }
1658e3e3a7aSWarner Losh 
1668e3e3a7aSWarner Losh 
1678e3e3a7aSWarner Losh static int str_char (lua_State *L) {
1688e3e3a7aSWarner Losh   int n = lua_gettop(L);  /* number of arguments */
1698e3e3a7aSWarner Losh   int i;
1708e3e3a7aSWarner Losh   luaL_Buffer b;
1718e3e3a7aSWarner Losh   char *p = luaL_buffinitsize(L, &b, n);
1728e3e3a7aSWarner Losh   for (i=1; i<=n; i++) {
1738e3e3a7aSWarner Losh     lua_Integer c = luaL_checkinteger(L, i);
1748e3e3a7aSWarner Losh     luaL_argcheck(L, uchar(c) == c, i, "value out of range");
1758e3e3a7aSWarner Losh     p[i - 1] = uchar(c);
1768e3e3a7aSWarner Losh   }
1778e3e3a7aSWarner Losh   luaL_pushresultsize(&b, n);
1788e3e3a7aSWarner Losh   return 1;
1798e3e3a7aSWarner Losh }
1808e3e3a7aSWarner Losh 
1818e3e3a7aSWarner Losh 
1828e3e3a7aSWarner Losh static int writer (lua_State *L, const void *b, size_t size, void *B) {
1838e3e3a7aSWarner Losh   (void)L;
1848e3e3a7aSWarner Losh   luaL_addlstring((luaL_Buffer *) B, (const char *)b, size);
1858e3e3a7aSWarner Losh   return 0;
1868e3e3a7aSWarner Losh }
1878e3e3a7aSWarner Losh 
1888e3e3a7aSWarner Losh 
1898e3e3a7aSWarner Losh static int str_dump (lua_State *L) {
1908e3e3a7aSWarner Losh   luaL_Buffer b;
1918e3e3a7aSWarner Losh   int strip = lua_toboolean(L, 2);
1928e3e3a7aSWarner Losh   luaL_checktype(L, 1, LUA_TFUNCTION);
1938e3e3a7aSWarner Losh   lua_settop(L, 1);
1948e3e3a7aSWarner Losh   luaL_buffinit(L,&b);
1958e3e3a7aSWarner Losh   if (lua_dump(L, writer, &b, strip) != 0)
1968e3e3a7aSWarner Losh     return luaL_error(L, "unable to dump given function");
1978e3e3a7aSWarner Losh   luaL_pushresult(&b);
1988e3e3a7aSWarner Losh   return 1;
1998e3e3a7aSWarner Losh }
2008e3e3a7aSWarner Losh 
2018e3e3a7aSWarner Losh 
2028e3e3a7aSWarner Losh 
2038e3e3a7aSWarner Losh /*
2048e3e3a7aSWarner Losh ** {======================================================
2058e3e3a7aSWarner Losh ** PATTERN MATCHING
2068e3e3a7aSWarner Losh ** =======================================================
2078e3e3a7aSWarner Losh */
2088e3e3a7aSWarner Losh 
2098e3e3a7aSWarner Losh 
2108e3e3a7aSWarner Losh #define CAP_UNFINISHED	(-1)
2118e3e3a7aSWarner Losh #define CAP_POSITION	(-2)
2128e3e3a7aSWarner Losh 
2138e3e3a7aSWarner Losh 
2148e3e3a7aSWarner Losh typedef struct MatchState {
2158e3e3a7aSWarner Losh   const char *src_init;  /* init of source string */
2168e3e3a7aSWarner Losh   const char *src_end;  /* end ('\0') of source string */
2178e3e3a7aSWarner Losh   const char *p_end;  /* end ('\0') of pattern */
2188e3e3a7aSWarner Losh   lua_State *L;
2198e3e3a7aSWarner Losh   int matchdepth;  /* control for recursive depth (to avoid C stack overflow) */
2208e3e3a7aSWarner Losh   unsigned char level;  /* total number of captures (finished or unfinished) */
2218e3e3a7aSWarner Losh   struct {
2228e3e3a7aSWarner Losh     const char *init;
2238e3e3a7aSWarner Losh     ptrdiff_t len;
2248e3e3a7aSWarner Losh   } capture[LUA_MAXCAPTURES];
2258e3e3a7aSWarner Losh } MatchState;
2268e3e3a7aSWarner Losh 
2278e3e3a7aSWarner Losh 
2288e3e3a7aSWarner Losh /* recursive function */
2298e3e3a7aSWarner Losh static const char *match (MatchState *ms, const char *s, const char *p);
2308e3e3a7aSWarner Losh 
2318e3e3a7aSWarner Losh 
2328e3e3a7aSWarner Losh /* maximum recursion depth for 'match' */
2338e3e3a7aSWarner Losh #if !defined(MAXCCALLS)
2348e3e3a7aSWarner Losh #define MAXCCALLS	200
2358e3e3a7aSWarner Losh #endif
2368e3e3a7aSWarner Losh 
2378e3e3a7aSWarner Losh 
2388e3e3a7aSWarner Losh #define L_ESC		'%'
2398e3e3a7aSWarner Losh #define SPECIALS	"^$*+?.([%-"
2408e3e3a7aSWarner Losh 
2418e3e3a7aSWarner Losh 
2428e3e3a7aSWarner Losh static int check_capture (MatchState *ms, int l) {
2438e3e3a7aSWarner Losh   l -= '1';
2448e3e3a7aSWarner Losh   if (l < 0 || l >= ms->level || ms->capture[l].len == CAP_UNFINISHED)
2458e3e3a7aSWarner Losh     return luaL_error(ms->L, "invalid capture index %%%d", l + 1);
2468e3e3a7aSWarner Losh   return l;
2478e3e3a7aSWarner Losh }
2488e3e3a7aSWarner Losh 
2498e3e3a7aSWarner Losh 
2508e3e3a7aSWarner Losh static int capture_to_close (MatchState *ms) {
2518e3e3a7aSWarner Losh   int level = ms->level;
2528e3e3a7aSWarner Losh   for (level--; level>=0; level--)
2538e3e3a7aSWarner Losh     if (ms->capture[level].len == CAP_UNFINISHED) return level;
2548e3e3a7aSWarner Losh   return luaL_error(ms->L, "invalid pattern capture");
2558e3e3a7aSWarner Losh }
2568e3e3a7aSWarner Losh 
2578e3e3a7aSWarner Losh 
2588e3e3a7aSWarner Losh static const char *classend (MatchState *ms, const char *p) {
2598e3e3a7aSWarner Losh   switch (*p++) {
2608e3e3a7aSWarner Losh     case L_ESC: {
2618e3e3a7aSWarner Losh       if (p == ms->p_end)
2628e3e3a7aSWarner Losh         luaL_error(ms->L, "malformed pattern (ends with '%%')");
2638e3e3a7aSWarner Losh       return p+1;
2648e3e3a7aSWarner Losh     }
2658e3e3a7aSWarner Losh     case '[': {
2668e3e3a7aSWarner Losh       if (*p == '^') p++;
2678e3e3a7aSWarner Losh       do {  /* look for a ']' */
2688e3e3a7aSWarner Losh         if (p == ms->p_end)
2698e3e3a7aSWarner Losh           luaL_error(ms->L, "malformed pattern (missing ']')");
2708e3e3a7aSWarner Losh         if (*(p++) == L_ESC && p < ms->p_end)
2718e3e3a7aSWarner Losh           p++;  /* skip escapes (e.g. '%]') */
2728e3e3a7aSWarner Losh       } while (*p != ']');
2738e3e3a7aSWarner Losh       return p+1;
2748e3e3a7aSWarner Losh     }
2758e3e3a7aSWarner Losh     default: {
2768e3e3a7aSWarner Losh       return p;
2778e3e3a7aSWarner Losh     }
2788e3e3a7aSWarner Losh   }
2798e3e3a7aSWarner Losh }
2808e3e3a7aSWarner Losh 
2818e3e3a7aSWarner Losh 
2828e3e3a7aSWarner Losh static int match_class (int c, int cl) {
2838e3e3a7aSWarner Losh   int res;
2848e3e3a7aSWarner Losh   switch (tolower(cl)) {
2858e3e3a7aSWarner Losh     case 'a' : res = isalpha(c); break;
2868e3e3a7aSWarner Losh     case 'c' : res = iscntrl(c); break;
2878e3e3a7aSWarner Losh     case 'd' : res = isdigit(c); break;
2888e3e3a7aSWarner Losh     case 'g' : res = isgraph(c); break;
2898e3e3a7aSWarner Losh     case 'l' : res = islower(c); break;
2908e3e3a7aSWarner Losh     case 'p' : res = ispunct(c); break;
2918e3e3a7aSWarner Losh     case 's' : res = isspace(c); break;
2928e3e3a7aSWarner Losh     case 'u' : res = isupper(c); break;
2938e3e3a7aSWarner Losh     case 'w' : res = isalnum(c); break;
2948e3e3a7aSWarner Losh     case 'x' : res = isxdigit(c); break;
2958e3e3a7aSWarner Losh     case 'z' : res = (c == 0); break;  /* deprecated option */
2968e3e3a7aSWarner Losh     default: return (cl == c);
2978e3e3a7aSWarner Losh   }
2988e3e3a7aSWarner Losh   return (islower(cl) ? res : !res);
2998e3e3a7aSWarner Losh }
3008e3e3a7aSWarner Losh 
3018e3e3a7aSWarner Losh 
3028e3e3a7aSWarner Losh static int matchbracketclass (int c, const char *p, const char *ec) {
3038e3e3a7aSWarner Losh   int sig = 1;
3048e3e3a7aSWarner Losh   if (*(p+1) == '^') {
3058e3e3a7aSWarner Losh     sig = 0;
3068e3e3a7aSWarner Losh     p++;  /* skip the '^' */
3078e3e3a7aSWarner Losh   }
3088e3e3a7aSWarner Losh   while (++p < ec) {
3098e3e3a7aSWarner Losh     if (*p == L_ESC) {
3108e3e3a7aSWarner Losh       p++;
3118e3e3a7aSWarner Losh       if (match_class(c, uchar(*p)))
3128e3e3a7aSWarner Losh         return sig;
3138e3e3a7aSWarner Losh     }
3148e3e3a7aSWarner Losh     else if ((*(p+1) == '-') && (p+2 < ec)) {
3158e3e3a7aSWarner Losh       p+=2;
3168e3e3a7aSWarner Losh       if (uchar(*(p-2)) <= c && c <= uchar(*p))
3178e3e3a7aSWarner Losh         return sig;
3188e3e3a7aSWarner Losh     }
3198e3e3a7aSWarner Losh     else if (uchar(*p) == c) return sig;
3208e3e3a7aSWarner Losh   }
3218e3e3a7aSWarner Losh   return !sig;
3228e3e3a7aSWarner Losh }
3238e3e3a7aSWarner Losh 
3248e3e3a7aSWarner Losh 
3258e3e3a7aSWarner Losh static int singlematch (MatchState *ms, const char *s, const char *p,
3268e3e3a7aSWarner Losh                         const char *ep) {
3278e3e3a7aSWarner Losh   if (s >= ms->src_end)
3288e3e3a7aSWarner Losh     return 0;
3298e3e3a7aSWarner Losh   else {
3308e3e3a7aSWarner Losh     int c = uchar(*s);
3318e3e3a7aSWarner Losh     switch (*p) {
3328e3e3a7aSWarner Losh       case '.': return 1;  /* matches any char */
3338e3e3a7aSWarner Losh       case L_ESC: return match_class(c, uchar(*(p+1)));
3348e3e3a7aSWarner Losh       case '[': return matchbracketclass(c, p, ep-1);
3358e3e3a7aSWarner Losh       default:  return (uchar(*p) == c);
3368e3e3a7aSWarner Losh     }
3378e3e3a7aSWarner Losh   }
3388e3e3a7aSWarner Losh }
3398e3e3a7aSWarner Losh 
3408e3e3a7aSWarner Losh 
3418e3e3a7aSWarner Losh static const char *matchbalance (MatchState *ms, const char *s,
3428e3e3a7aSWarner Losh                                    const char *p) {
3438e3e3a7aSWarner Losh   if (p >= ms->p_end - 1)
3448e3e3a7aSWarner Losh     luaL_error(ms->L, "malformed pattern (missing arguments to '%%b')");
3458e3e3a7aSWarner Losh   if (*s != *p) return NULL;
3468e3e3a7aSWarner Losh   else {
3478e3e3a7aSWarner Losh     int b = *p;
3488e3e3a7aSWarner Losh     int e = *(p+1);
3498e3e3a7aSWarner Losh     int cont = 1;
3508e3e3a7aSWarner Losh     while (++s < ms->src_end) {
3518e3e3a7aSWarner Losh       if (*s == e) {
3528e3e3a7aSWarner Losh         if (--cont == 0) return s+1;
3538e3e3a7aSWarner Losh       }
3548e3e3a7aSWarner Losh       else if (*s == b) cont++;
3558e3e3a7aSWarner Losh     }
3568e3e3a7aSWarner Losh   }
3578e3e3a7aSWarner Losh   return NULL;  /* string ends out of balance */
3588e3e3a7aSWarner Losh }
3598e3e3a7aSWarner Losh 
3608e3e3a7aSWarner Losh 
3618e3e3a7aSWarner Losh static const char *max_expand (MatchState *ms, const char *s,
3628e3e3a7aSWarner Losh                                  const char *p, const char *ep) {
3638e3e3a7aSWarner Losh   ptrdiff_t i = 0;  /* counts maximum expand for item */
3648e3e3a7aSWarner Losh   while (singlematch(ms, s + i, p, ep))
3658e3e3a7aSWarner Losh     i++;
3668e3e3a7aSWarner Losh   /* keeps trying to match with the maximum repetitions */
3678e3e3a7aSWarner Losh   while (i>=0) {
3688e3e3a7aSWarner Losh     const char *res = match(ms, (s+i), ep+1);
3698e3e3a7aSWarner Losh     if (res) return res;
3708e3e3a7aSWarner Losh     i--;  /* else didn't match; reduce 1 repetition to try again */
3718e3e3a7aSWarner Losh   }
3728e3e3a7aSWarner Losh   return NULL;
3738e3e3a7aSWarner Losh }
3748e3e3a7aSWarner Losh 
3758e3e3a7aSWarner Losh 
3768e3e3a7aSWarner Losh static const char *min_expand (MatchState *ms, const char *s,
3778e3e3a7aSWarner Losh                                  const char *p, const char *ep) {
3788e3e3a7aSWarner Losh   for (;;) {
3798e3e3a7aSWarner Losh     const char *res = match(ms, s, ep+1);
3808e3e3a7aSWarner Losh     if (res != NULL)
3818e3e3a7aSWarner Losh       return res;
3828e3e3a7aSWarner Losh     else if (singlematch(ms, s, p, ep))
3838e3e3a7aSWarner Losh       s++;  /* try with one more repetition */
3848e3e3a7aSWarner Losh     else return NULL;
3858e3e3a7aSWarner Losh   }
3868e3e3a7aSWarner Losh }
3878e3e3a7aSWarner Losh 
3888e3e3a7aSWarner Losh 
3898e3e3a7aSWarner Losh static const char *start_capture (MatchState *ms, const char *s,
3908e3e3a7aSWarner Losh                                     const char *p, int what) {
3918e3e3a7aSWarner Losh   const char *res;
3928e3e3a7aSWarner Losh   int level = ms->level;
3938e3e3a7aSWarner Losh   if (level >= LUA_MAXCAPTURES) luaL_error(ms->L, "too many captures");
3948e3e3a7aSWarner Losh   ms->capture[level].init = s;
3958e3e3a7aSWarner Losh   ms->capture[level].len = what;
3968e3e3a7aSWarner Losh   ms->level = level+1;
3978e3e3a7aSWarner Losh   if ((res=match(ms, s, p)) == NULL)  /* match failed? */
3988e3e3a7aSWarner Losh     ms->level--;  /* undo capture */
3998e3e3a7aSWarner Losh   return res;
4008e3e3a7aSWarner Losh }
4018e3e3a7aSWarner Losh 
4028e3e3a7aSWarner Losh 
4038e3e3a7aSWarner Losh static const char *end_capture (MatchState *ms, const char *s,
4048e3e3a7aSWarner Losh                                   const char *p) {
4058e3e3a7aSWarner Losh   int l = capture_to_close(ms);
4068e3e3a7aSWarner Losh   const char *res;
4078e3e3a7aSWarner Losh   ms->capture[l].len = s - ms->capture[l].init;  /* close capture */
4088e3e3a7aSWarner Losh   if ((res = match(ms, s, p)) == NULL)  /* match failed? */
4098e3e3a7aSWarner Losh     ms->capture[l].len = CAP_UNFINISHED;  /* undo capture */
4108e3e3a7aSWarner Losh   return res;
4118e3e3a7aSWarner Losh }
4128e3e3a7aSWarner Losh 
4138e3e3a7aSWarner Losh 
4148e3e3a7aSWarner Losh static const char *match_capture (MatchState *ms, const char *s, int l) {
4158e3e3a7aSWarner Losh   size_t len;
4168e3e3a7aSWarner Losh   l = check_capture(ms, l);
4178e3e3a7aSWarner Losh   len = ms->capture[l].len;
4188e3e3a7aSWarner Losh   if ((size_t)(ms->src_end-s) >= len &&
4198e3e3a7aSWarner Losh       memcmp(ms->capture[l].init, s, len) == 0)
4208e3e3a7aSWarner Losh     return s+len;
4218e3e3a7aSWarner Losh   else return NULL;
4228e3e3a7aSWarner Losh }
4238e3e3a7aSWarner Losh 
4248e3e3a7aSWarner Losh 
4258e3e3a7aSWarner Losh static const char *match (MatchState *ms, const char *s, const char *p) {
4268e3e3a7aSWarner Losh   if (ms->matchdepth-- == 0)
4278e3e3a7aSWarner Losh     luaL_error(ms->L, "pattern too complex");
4288e3e3a7aSWarner Losh   init: /* using goto's to optimize tail recursion */
4298e3e3a7aSWarner Losh   if (p != ms->p_end) {  /* end of pattern? */
4308e3e3a7aSWarner Losh     switch (*p) {
4318e3e3a7aSWarner Losh       case '(': {  /* start capture */
4328e3e3a7aSWarner Losh         if (*(p + 1) == ')')  /* position capture? */
4338e3e3a7aSWarner Losh           s = start_capture(ms, s, p + 2, CAP_POSITION);
4348e3e3a7aSWarner Losh         else
4358e3e3a7aSWarner Losh           s = start_capture(ms, s, p + 1, CAP_UNFINISHED);
4368e3e3a7aSWarner Losh         break;
4378e3e3a7aSWarner Losh       }
4388e3e3a7aSWarner Losh       case ')': {  /* end capture */
4398e3e3a7aSWarner Losh         s = end_capture(ms, s, p + 1);
4408e3e3a7aSWarner Losh         break;
4418e3e3a7aSWarner Losh       }
4428e3e3a7aSWarner Losh       case '$': {
4438e3e3a7aSWarner Losh         if ((p + 1) != ms->p_end)  /* is the '$' the last char in pattern? */
4448e3e3a7aSWarner Losh           goto dflt;  /* no; go to default */
4458e3e3a7aSWarner Losh         s = (s == ms->src_end) ? s : NULL;  /* check end of string */
4468e3e3a7aSWarner Losh         break;
4478e3e3a7aSWarner Losh       }
4488e3e3a7aSWarner Losh       case L_ESC: {  /* escaped sequences not in the format class[*+?-]? */
4498e3e3a7aSWarner Losh         switch (*(p + 1)) {
4508e3e3a7aSWarner Losh           case 'b': {  /* balanced string? */
4518e3e3a7aSWarner Losh             s = matchbalance(ms, s, p + 2);
4528e3e3a7aSWarner Losh             if (s != NULL) {
4538e3e3a7aSWarner Losh               p += 4; goto init;  /* return match(ms, s, p + 4); */
4548e3e3a7aSWarner Losh             }  /* else fail (s == NULL) */
4558e3e3a7aSWarner Losh             break;
4568e3e3a7aSWarner Losh           }
4578e3e3a7aSWarner Losh           case 'f': {  /* frontier? */
4588e3e3a7aSWarner Losh             const char *ep; char previous;
4598e3e3a7aSWarner Losh             p += 2;
4608e3e3a7aSWarner Losh             if (*p != '[')
4618e3e3a7aSWarner Losh               luaL_error(ms->L, "missing '[' after '%%f' in pattern");
4628e3e3a7aSWarner Losh             ep = classend(ms, p);  /* points to what is next */
4638e3e3a7aSWarner Losh             previous = (s == ms->src_init) ? '\0' : *(s - 1);
4648e3e3a7aSWarner Losh             if (!matchbracketclass(uchar(previous), p, ep - 1) &&
4658e3e3a7aSWarner Losh                matchbracketclass(uchar(*s), p, ep - 1)) {
4668e3e3a7aSWarner Losh               p = ep; goto init;  /* return match(ms, s, ep); */
4678e3e3a7aSWarner Losh             }
4688e3e3a7aSWarner Losh             s = NULL;  /* match failed */
4698e3e3a7aSWarner Losh             break;
4708e3e3a7aSWarner Losh           }
4718e3e3a7aSWarner Losh           case '0': case '1': case '2': case '3':
4728e3e3a7aSWarner Losh           case '4': case '5': case '6': case '7':
4738e3e3a7aSWarner Losh           case '8': case '9': {  /* capture results (%0-%9)? */
4748e3e3a7aSWarner Losh             s = match_capture(ms, s, uchar(*(p + 1)));
4758e3e3a7aSWarner Losh             if (s != NULL) {
4768e3e3a7aSWarner Losh               p += 2; goto init;  /* return match(ms, s, p + 2) */
4778e3e3a7aSWarner Losh             }
4788e3e3a7aSWarner Losh             break;
4798e3e3a7aSWarner Losh           }
4808e3e3a7aSWarner Losh           default: goto dflt;
4818e3e3a7aSWarner Losh         }
4828e3e3a7aSWarner Losh         break;
4838e3e3a7aSWarner Losh       }
4848e3e3a7aSWarner Losh       default: dflt: {  /* pattern class plus optional suffix */
4858e3e3a7aSWarner Losh         const char *ep = classend(ms, p);  /* points to optional suffix */
4868e3e3a7aSWarner Losh         /* does not match at least once? */
4878e3e3a7aSWarner Losh         if (!singlematch(ms, s, p, ep)) {
4888e3e3a7aSWarner Losh           if (*ep == '*' || *ep == '?' || *ep == '-') {  /* accept empty? */
4898e3e3a7aSWarner Losh             p = ep + 1; goto init;  /* return match(ms, s, ep + 1); */
4908e3e3a7aSWarner Losh           }
4918e3e3a7aSWarner Losh           else  /* '+' or no suffix */
4928e3e3a7aSWarner Losh             s = NULL;  /* fail */
4938e3e3a7aSWarner Losh         }
4948e3e3a7aSWarner Losh         else {  /* matched once */
4958e3e3a7aSWarner Losh           switch (*ep) {  /* handle optional suffix */
4968e3e3a7aSWarner Losh             case '?': {  /* optional */
4978e3e3a7aSWarner Losh               const char *res;
4988e3e3a7aSWarner Losh               if ((res = match(ms, s + 1, ep + 1)) != NULL)
4998e3e3a7aSWarner Losh                 s = res;
5008e3e3a7aSWarner Losh               else {
5018e3e3a7aSWarner Losh                 p = ep + 1; goto init;  /* else return match(ms, s, ep + 1); */
5028e3e3a7aSWarner Losh               }
5038e3e3a7aSWarner Losh               break;
5048e3e3a7aSWarner Losh             }
5058e3e3a7aSWarner Losh             case '+':  /* 1 or more repetitions */
5068e3e3a7aSWarner Losh               s++;  /* 1 match already done */
5078e3e3a7aSWarner Losh               /* FALLTHROUGH */
5088e3e3a7aSWarner Losh             case '*':  /* 0 or more repetitions */
5098e3e3a7aSWarner Losh               s = max_expand(ms, s, p, ep);
5108e3e3a7aSWarner Losh               break;
5118e3e3a7aSWarner Losh             case '-':  /* 0 or more repetitions (minimum) */
5128e3e3a7aSWarner Losh               s = min_expand(ms, s, p, ep);
5138e3e3a7aSWarner Losh               break;
5148e3e3a7aSWarner Losh             default:  /* no suffix */
5158e3e3a7aSWarner Losh               s++; p = ep; goto init;  /* return match(ms, s + 1, ep); */
5168e3e3a7aSWarner Losh           }
5178e3e3a7aSWarner Losh         }
5188e3e3a7aSWarner Losh         break;
5198e3e3a7aSWarner Losh       }
5208e3e3a7aSWarner Losh     }
5218e3e3a7aSWarner Losh   }
5228e3e3a7aSWarner Losh   ms->matchdepth++;
5238e3e3a7aSWarner Losh   return s;
5248e3e3a7aSWarner Losh }
5258e3e3a7aSWarner Losh 
5268e3e3a7aSWarner Losh 
5278e3e3a7aSWarner Losh 
5288e3e3a7aSWarner Losh static const char *lmemfind (const char *s1, size_t l1,
5298e3e3a7aSWarner Losh                                const char *s2, size_t l2) {
5308e3e3a7aSWarner Losh   if (l2 == 0) return s1;  /* empty strings are everywhere */
5318e3e3a7aSWarner Losh   else if (l2 > l1) return NULL;  /* avoids a negative 'l1' */
5328e3e3a7aSWarner Losh   else {
5338e3e3a7aSWarner Losh     const char *init;  /* to search for a '*s2' inside 's1' */
5348e3e3a7aSWarner Losh     l2--;  /* 1st char will be checked by 'memchr' */
5358e3e3a7aSWarner Losh     l1 = l1-l2;  /* 's2' cannot be found after that */
5368e3e3a7aSWarner Losh     while (l1 > 0 && (init = (const char *)memchr(s1, *s2, l1)) != NULL) {
5378e3e3a7aSWarner Losh       init++;   /* 1st char is already checked */
5388e3e3a7aSWarner Losh       if (memcmp(init, s2+1, l2) == 0)
5398e3e3a7aSWarner Losh         return init-1;
5408e3e3a7aSWarner Losh       else {  /* correct 'l1' and 's1' to try again */
5418e3e3a7aSWarner Losh         l1 -= init-s1;
5428e3e3a7aSWarner Losh         s1 = init;
5438e3e3a7aSWarner Losh       }
5448e3e3a7aSWarner Losh     }
5458e3e3a7aSWarner Losh     return NULL;  /* not found */
5468e3e3a7aSWarner Losh   }
5478e3e3a7aSWarner Losh }
5488e3e3a7aSWarner Losh 
5498e3e3a7aSWarner Losh 
5508e3e3a7aSWarner Losh static void push_onecapture (MatchState *ms, int i, const char *s,
5518e3e3a7aSWarner Losh                                                     const char *e) {
5528e3e3a7aSWarner Losh   if (i >= ms->level) {
5538e3e3a7aSWarner Losh     if (i == 0)  /* ms->level == 0, too */
5548e3e3a7aSWarner Losh       lua_pushlstring(ms->L, s, e - s);  /* add whole match */
5558e3e3a7aSWarner Losh     else
5568e3e3a7aSWarner Losh       luaL_error(ms->L, "invalid capture index %%%d", i + 1);
5578e3e3a7aSWarner Losh   }
5588e3e3a7aSWarner Losh   else {
5598e3e3a7aSWarner Losh     ptrdiff_t l = ms->capture[i].len;
5608e3e3a7aSWarner Losh     if (l == CAP_UNFINISHED) luaL_error(ms->L, "unfinished capture");
5618e3e3a7aSWarner Losh     if (l == CAP_POSITION)
5628e3e3a7aSWarner Losh       lua_pushinteger(ms->L, (ms->capture[i].init - ms->src_init) + 1);
5638e3e3a7aSWarner Losh     else
5648e3e3a7aSWarner Losh       lua_pushlstring(ms->L, ms->capture[i].init, l);
5658e3e3a7aSWarner Losh   }
5668e3e3a7aSWarner Losh }
5678e3e3a7aSWarner Losh 
5688e3e3a7aSWarner Losh 
5698e3e3a7aSWarner Losh static int push_captures (MatchState *ms, const char *s, const char *e) {
5708e3e3a7aSWarner Losh   int i;
5718e3e3a7aSWarner Losh   int nlevels = (ms->level == 0 && s) ? 1 : ms->level;
5728e3e3a7aSWarner Losh   luaL_checkstack(ms->L, nlevels, "too many captures");
5738e3e3a7aSWarner Losh   for (i = 0; i < nlevels; i++)
5748e3e3a7aSWarner Losh     push_onecapture(ms, i, s, e);
5758e3e3a7aSWarner Losh   return nlevels;  /* number of strings pushed */
5768e3e3a7aSWarner Losh }
5778e3e3a7aSWarner Losh 
5788e3e3a7aSWarner Losh 
5798e3e3a7aSWarner Losh /* check whether pattern has no special characters */
5808e3e3a7aSWarner Losh static int nospecials (const char *p, size_t l) {
5818e3e3a7aSWarner Losh   size_t upto = 0;
5828e3e3a7aSWarner Losh   do {
5838e3e3a7aSWarner Losh     if (strpbrk(p + upto, SPECIALS))
5848e3e3a7aSWarner Losh       return 0;  /* pattern has a special character */
5858e3e3a7aSWarner Losh     upto += strlen(p + upto) + 1;  /* may have more after \0 */
5868e3e3a7aSWarner Losh   } while (upto <= l);
5878e3e3a7aSWarner Losh   return 1;  /* no special chars found */
5888e3e3a7aSWarner Losh }
5898e3e3a7aSWarner Losh 
5908e3e3a7aSWarner Losh 
5918e3e3a7aSWarner Losh static void prepstate (MatchState *ms, lua_State *L,
5928e3e3a7aSWarner Losh                        const char *s, size_t ls, const char *p, size_t lp) {
5938e3e3a7aSWarner Losh   ms->L = L;
5948e3e3a7aSWarner Losh   ms->matchdepth = MAXCCALLS;
5958e3e3a7aSWarner Losh   ms->src_init = s;
5968e3e3a7aSWarner Losh   ms->src_end = s + ls;
5978e3e3a7aSWarner Losh   ms->p_end = p + lp;
5988e3e3a7aSWarner Losh }
5998e3e3a7aSWarner Losh 
6008e3e3a7aSWarner Losh 
6018e3e3a7aSWarner Losh static void reprepstate (MatchState *ms) {
6028e3e3a7aSWarner Losh   ms->level = 0;
6038e3e3a7aSWarner Losh   lua_assert(ms->matchdepth == MAXCCALLS);
6048e3e3a7aSWarner Losh }
6058e3e3a7aSWarner Losh 
6068e3e3a7aSWarner Losh 
6078e3e3a7aSWarner Losh static int str_find_aux (lua_State *L, int find) {
6088e3e3a7aSWarner Losh   size_t ls, lp;
6098e3e3a7aSWarner Losh   const char *s = luaL_checklstring(L, 1, &ls);
6108e3e3a7aSWarner Losh   const char *p = luaL_checklstring(L, 2, &lp);
6118e3e3a7aSWarner Losh   lua_Integer init = posrelat(luaL_optinteger(L, 3, 1), ls);
6128e3e3a7aSWarner Losh   if (init < 1) init = 1;
6138e3e3a7aSWarner Losh   else if (init > (lua_Integer)ls + 1) {  /* start after string's end? */
6148e3e3a7aSWarner Losh     lua_pushnil(L);  /* cannot find anything */
6158e3e3a7aSWarner Losh     return 1;
6168e3e3a7aSWarner Losh   }
6178e3e3a7aSWarner Losh   /* explicit request or no special characters? */
6188e3e3a7aSWarner Losh   if (find && (lua_toboolean(L, 4) || nospecials(p, lp))) {
6198e3e3a7aSWarner Losh     /* do a plain search */
6208e3e3a7aSWarner Losh     const char *s2 = lmemfind(s + init - 1, ls - (size_t)init + 1, p, lp);
6218e3e3a7aSWarner Losh     if (s2) {
6228e3e3a7aSWarner Losh       lua_pushinteger(L, (s2 - s) + 1);
6238e3e3a7aSWarner Losh       lua_pushinteger(L, (s2 - s) + lp);
6248e3e3a7aSWarner Losh       return 2;
6258e3e3a7aSWarner Losh     }
6268e3e3a7aSWarner Losh   }
6278e3e3a7aSWarner Losh   else {
6288e3e3a7aSWarner Losh     MatchState ms;
6298e3e3a7aSWarner Losh     const char *s1 = s + init - 1;
6308e3e3a7aSWarner Losh     int anchor = (*p == '^');
6318e3e3a7aSWarner Losh     if (anchor) {
6328e3e3a7aSWarner Losh       p++; lp--;  /* skip anchor character */
6338e3e3a7aSWarner Losh     }
6348e3e3a7aSWarner Losh     prepstate(&ms, L, s, ls, p, lp);
6358e3e3a7aSWarner Losh     do {
6368e3e3a7aSWarner Losh       const char *res;
6378e3e3a7aSWarner Losh       reprepstate(&ms);
6388e3e3a7aSWarner Losh       if ((res=match(&ms, s1, p)) != NULL) {
6398e3e3a7aSWarner Losh         if (find) {
6408e3e3a7aSWarner Losh           lua_pushinteger(L, (s1 - s) + 1);  /* start */
6418e3e3a7aSWarner Losh           lua_pushinteger(L, res - s);   /* end */
6428e3e3a7aSWarner Losh           return push_captures(&ms, NULL, 0) + 2;
6438e3e3a7aSWarner Losh         }
6448e3e3a7aSWarner Losh         else
6458e3e3a7aSWarner Losh           return push_captures(&ms, s1, res);
6468e3e3a7aSWarner Losh       }
6478e3e3a7aSWarner Losh     } while (s1++ < ms.src_end && !anchor);
6488e3e3a7aSWarner Losh   }
6498e3e3a7aSWarner Losh   lua_pushnil(L);  /* not found */
6508e3e3a7aSWarner Losh   return 1;
6518e3e3a7aSWarner Losh }
6528e3e3a7aSWarner Losh 
6538e3e3a7aSWarner Losh 
6548e3e3a7aSWarner Losh static int str_find (lua_State *L) {
6558e3e3a7aSWarner Losh   return str_find_aux(L, 1);
6568e3e3a7aSWarner Losh }
6578e3e3a7aSWarner Losh 
6588e3e3a7aSWarner Losh 
6598e3e3a7aSWarner Losh static int str_match (lua_State *L) {
6608e3e3a7aSWarner Losh   return str_find_aux(L, 0);
6618e3e3a7aSWarner Losh }
6628e3e3a7aSWarner Losh 
6638e3e3a7aSWarner Losh 
6648e3e3a7aSWarner Losh /* state for 'gmatch' */
6658e3e3a7aSWarner Losh typedef struct GMatchState {
6668e3e3a7aSWarner Losh   const char *src;  /* current position */
6678e3e3a7aSWarner Losh   const char *p;  /* pattern */
6688e3e3a7aSWarner Losh   const char *lastmatch;  /* end of last match */
6698e3e3a7aSWarner Losh   MatchState ms;  /* match state */
6708e3e3a7aSWarner Losh } GMatchState;
6718e3e3a7aSWarner Losh 
6728e3e3a7aSWarner Losh 
6738e3e3a7aSWarner Losh static int gmatch_aux (lua_State *L) {
6748e3e3a7aSWarner Losh   GMatchState *gm = (GMatchState *)lua_touserdata(L, lua_upvalueindex(3));
6758e3e3a7aSWarner Losh   const char *src;
6768e3e3a7aSWarner Losh   gm->ms.L = L;
6778e3e3a7aSWarner Losh   for (src = gm->src; src <= gm->ms.src_end; src++) {
6788e3e3a7aSWarner Losh     const char *e;
6798e3e3a7aSWarner Losh     reprepstate(&gm->ms);
6808e3e3a7aSWarner Losh     if ((e = match(&gm->ms, src, gm->p)) != NULL && e != gm->lastmatch) {
6818e3e3a7aSWarner Losh       gm->src = gm->lastmatch = e;
6828e3e3a7aSWarner Losh       return push_captures(&gm->ms, src, e);
6838e3e3a7aSWarner Losh     }
6848e3e3a7aSWarner Losh   }
6858e3e3a7aSWarner Losh   return 0;  /* not found */
6868e3e3a7aSWarner Losh }
6878e3e3a7aSWarner Losh 
6888e3e3a7aSWarner Losh 
6898e3e3a7aSWarner Losh static int gmatch (lua_State *L) {
6908e3e3a7aSWarner Losh   size_t ls, lp;
6918e3e3a7aSWarner Losh   const char *s = luaL_checklstring(L, 1, &ls);
6928e3e3a7aSWarner Losh   const char *p = luaL_checklstring(L, 2, &lp);
6938e3e3a7aSWarner Losh   GMatchState *gm;
6948e3e3a7aSWarner Losh   lua_settop(L, 2);  /* keep them on closure to avoid being collected */
6958e3e3a7aSWarner Losh   gm = (GMatchState *)lua_newuserdata(L, sizeof(GMatchState));
6968e3e3a7aSWarner Losh   prepstate(&gm->ms, L, s, ls, p, lp);
6978e3e3a7aSWarner Losh   gm->src = s; gm->p = p; gm->lastmatch = NULL;
6988e3e3a7aSWarner Losh   lua_pushcclosure(L, gmatch_aux, 3);
6998e3e3a7aSWarner Losh   return 1;
7008e3e3a7aSWarner Losh }
7018e3e3a7aSWarner Losh 
7028e3e3a7aSWarner Losh 
7038e3e3a7aSWarner Losh static void add_s (MatchState *ms, luaL_Buffer *b, const char *s,
7048e3e3a7aSWarner Losh                                                    const char *e) {
7058e3e3a7aSWarner Losh   size_t l, i;
7068e3e3a7aSWarner Losh   lua_State *L = ms->L;
7078e3e3a7aSWarner Losh   const char *news = lua_tolstring(L, 3, &l);
7088e3e3a7aSWarner Losh   for (i = 0; i < l; i++) {
7098e3e3a7aSWarner Losh     if (news[i] != L_ESC)
7108e3e3a7aSWarner Losh       luaL_addchar(b, news[i]);
7118e3e3a7aSWarner Losh     else {
7128e3e3a7aSWarner Losh       i++;  /* skip ESC */
7138e3e3a7aSWarner Losh       if (!isdigit(uchar(news[i]))) {
7148e3e3a7aSWarner Losh         if (news[i] != L_ESC)
7158e3e3a7aSWarner Losh           luaL_error(L, "invalid use of '%c' in replacement string", L_ESC);
7168e3e3a7aSWarner Losh         luaL_addchar(b, news[i]);
7178e3e3a7aSWarner Losh       }
7188e3e3a7aSWarner Losh       else if (news[i] == '0')
7198e3e3a7aSWarner Losh           luaL_addlstring(b, s, e - s);
7208e3e3a7aSWarner Losh       else {
7218e3e3a7aSWarner Losh         push_onecapture(ms, news[i] - '1', s, e);
7228e3e3a7aSWarner Losh         luaL_tolstring(L, -1, NULL);  /* if number, convert it to string */
7238e3e3a7aSWarner Losh         lua_remove(L, -2);  /* remove original value */
7248e3e3a7aSWarner Losh         luaL_addvalue(b);  /* add capture to accumulated result */
7258e3e3a7aSWarner Losh       }
7268e3e3a7aSWarner Losh     }
7278e3e3a7aSWarner Losh   }
7288e3e3a7aSWarner Losh }
7298e3e3a7aSWarner Losh 
7308e3e3a7aSWarner Losh 
7318e3e3a7aSWarner Losh static void add_value (MatchState *ms, luaL_Buffer *b, const char *s,
7328e3e3a7aSWarner Losh                                        const char *e, int tr) {
7338e3e3a7aSWarner Losh   lua_State *L = ms->L;
7348e3e3a7aSWarner Losh   switch (tr) {
7358e3e3a7aSWarner Losh     case LUA_TFUNCTION: {
7368e3e3a7aSWarner Losh       int n;
7378e3e3a7aSWarner Losh       lua_pushvalue(L, 3);
7388e3e3a7aSWarner Losh       n = push_captures(ms, s, e);
7398e3e3a7aSWarner Losh       lua_call(L, n, 1);
7408e3e3a7aSWarner Losh       break;
7418e3e3a7aSWarner Losh     }
7428e3e3a7aSWarner Losh     case LUA_TTABLE: {
7438e3e3a7aSWarner Losh       push_onecapture(ms, 0, s, e);
7448e3e3a7aSWarner Losh       lua_gettable(L, 3);
7458e3e3a7aSWarner Losh       break;
7468e3e3a7aSWarner Losh     }
7478e3e3a7aSWarner Losh     default: {  /* LUA_TNUMBER or LUA_TSTRING */
7488e3e3a7aSWarner Losh       add_s(ms, b, s, e);
7498e3e3a7aSWarner Losh       return;
7508e3e3a7aSWarner Losh     }
7518e3e3a7aSWarner Losh   }
7528e3e3a7aSWarner Losh   if (!lua_toboolean(L, -1)) {  /* nil or false? */
7538e3e3a7aSWarner Losh     lua_pop(L, 1);
7548e3e3a7aSWarner Losh     lua_pushlstring(L, s, e - s);  /* keep original text */
7558e3e3a7aSWarner Losh   }
7568e3e3a7aSWarner Losh   else if (!lua_isstring(L, -1))
7578e3e3a7aSWarner Losh     luaL_error(L, "invalid replacement value (a %s)", luaL_typename(L, -1));
7588e3e3a7aSWarner Losh   luaL_addvalue(b);  /* add result to accumulator */
7598e3e3a7aSWarner Losh }
7608e3e3a7aSWarner Losh 
7618e3e3a7aSWarner Losh 
7628e3e3a7aSWarner Losh static int str_gsub (lua_State *L) {
7638e3e3a7aSWarner Losh   size_t srcl, lp;
7648e3e3a7aSWarner Losh   const char *src = luaL_checklstring(L, 1, &srcl);  /* subject */
7658e3e3a7aSWarner Losh   const char *p = luaL_checklstring(L, 2, &lp);  /* pattern */
7668e3e3a7aSWarner Losh   const char *lastmatch = NULL;  /* end of last match */
7678e3e3a7aSWarner Losh   int tr = lua_type(L, 3);  /* replacement type */
7688e3e3a7aSWarner Losh   lua_Integer max_s = luaL_optinteger(L, 4, srcl + 1);  /* max replacements */
7698e3e3a7aSWarner Losh   int anchor = (*p == '^');
7708e3e3a7aSWarner Losh   lua_Integer n = 0;  /* replacement count */
7718e3e3a7aSWarner Losh   MatchState ms;
7728e3e3a7aSWarner Losh   luaL_Buffer b;
7738e3e3a7aSWarner Losh   luaL_argcheck(L, tr == LUA_TNUMBER || tr == LUA_TSTRING ||
7748e3e3a7aSWarner Losh                    tr == LUA_TFUNCTION || tr == LUA_TTABLE, 3,
7758e3e3a7aSWarner Losh                       "string/function/table expected");
7768e3e3a7aSWarner Losh   luaL_buffinit(L, &b);
7778e3e3a7aSWarner Losh   if (anchor) {
7788e3e3a7aSWarner Losh     p++; lp--;  /* skip anchor character */
7798e3e3a7aSWarner Losh   }
7808e3e3a7aSWarner Losh   prepstate(&ms, L, src, srcl, p, lp);
7818e3e3a7aSWarner Losh   while (n < max_s) {
7828e3e3a7aSWarner Losh     const char *e;
7838e3e3a7aSWarner Losh     reprepstate(&ms);  /* (re)prepare state for new match */
7848e3e3a7aSWarner Losh     if ((e = match(&ms, src, p)) != NULL && e != lastmatch) {  /* match? */
7858e3e3a7aSWarner Losh       n++;
7868e3e3a7aSWarner Losh       add_value(&ms, &b, src, e, tr);  /* add replacement to buffer */
7878e3e3a7aSWarner Losh       src = lastmatch = e;
7888e3e3a7aSWarner Losh     }
7898e3e3a7aSWarner Losh     else if (src < ms.src_end)  /* otherwise, skip one character */
7908e3e3a7aSWarner Losh       luaL_addchar(&b, *src++);
7918e3e3a7aSWarner Losh     else break;  /* end of subject */
7928e3e3a7aSWarner Losh     if (anchor) break;
7938e3e3a7aSWarner Losh   }
7948e3e3a7aSWarner Losh   luaL_addlstring(&b, src, ms.src_end-src);
7958e3e3a7aSWarner Losh   luaL_pushresult(&b);
7968e3e3a7aSWarner Losh   lua_pushinteger(L, n);  /* number of substitutions */
7978e3e3a7aSWarner Losh   return 2;
7988e3e3a7aSWarner Losh }
7998e3e3a7aSWarner Losh 
8008e3e3a7aSWarner Losh /* }====================================================== */
8018e3e3a7aSWarner Losh 
8028e3e3a7aSWarner Losh 
8038e3e3a7aSWarner Losh 
8048e3e3a7aSWarner Losh /*
8058e3e3a7aSWarner Losh ** {======================================================
8068e3e3a7aSWarner Losh ** STRING FORMAT
8078e3e3a7aSWarner Losh ** =======================================================
8088e3e3a7aSWarner Losh */
8098e3e3a7aSWarner Losh 
8108e3e3a7aSWarner Losh #if !defined(lua_number2strx)	/* { */
8118e3e3a7aSWarner Losh 
8128e3e3a7aSWarner Losh /*
8138e3e3a7aSWarner Losh ** Hexadecimal floating-point formatter
8148e3e3a7aSWarner Losh */
8158e3e3a7aSWarner Losh 
8168e3e3a7aSWarner Losh #include <math.h>
8178e3e3a7aSWarner Losh 
8188e3e3a7aSWarner Losh #define SIZELENMOD	(sizeof(LUA_NUMBER_FRMLEN)/sizeof(char))
8198e3e3a7aSWarner Losh 
8208e3e3a7aSWarner Losh 
8218e3e3a7aSWarner Losh /*
8228e3e3a7aSWarner Losh ** Number of bits that goes into the first digit. It can be any value
8238e3e3a7aSWarner Losh ** between 1 and 4; the following definition tries to align the number
8248e3e3a7aSWarner Losh ** to nibble boundaries by making what is left after that first digit a
8258e3e3a7aSWarner Losh ** multiple of 4.
8268e3e3a7aSWarner Losh */
8278e3e3a7aSWarner Losh #define L_NBFD		((l_mathlim(MANT_DIG) - 1)%4 + 1)
8288e3e3a7aSWarner Losh 
8298e3e3a7aSWarner Losh 
8308e3e3a7aSWarner Losh /*
8318e3e3a7aSWarner Losh ** Add integer part of 'x' to buffer and return new 'x'
8328e3e3a7aSWarner Losh */
8338e3e3a7aSWarner Losh static lua_Number adddigit (char *buff, int n, lua_Number x) {
8348e3e3a7aSWarner Losh   lua_Number dd = l_mathop(floor)(x);  /* get integer part from 'x' */
8358e3e3a7aSWarner Losh   int d = (int)dd;
8368e3e3a7aSWarner Losh   buff[n] = (d < 10 ? d + '0' : d - 10 + 'a');  /* add to buffer */
8378e3e3a7aSWarner Losh   return x - dd;  /* return what is left */
8388e3e3a7aSWarner Losh }
8398e3e3a7aSWarner Losh 
8408e3e3a7aSWarner Losh 
8418e3e3a7aSWarner Losh static int num2straux (char *buff, int sz, lua_Number x) {
8428e3e3a7aSWarner Losh   /* if 'inf' or 'NaN', format it like '%g' */
8438e3e3a7aSWarner Losh   if (x != x || x == (lua_Number)HUGE_VAL || x == -(lua_Number)HUGE_VAL)
8448e3e3a7aSWarner Losh     return l_sprintf(buff, sz, LUA_NUMBER_FMT, (LUAI_UACNUMBER)x);
8458e3e3a7aSWarner Losh   else if (x == 0) {  /* can be -0... */
8468e3e3a7aSWarner Losh     /* create "0" or "-0" followed by exponent */
8478e3e3a7aSWarner Losh     return l_sprintf(buff, sz, LUA_NUMBER_FMT "x0p+0", (LUAI_UACNUMBER)x);
8488e3e3a7aSWarner Losh   }
8498e3e3a7aSWarner Losh   else {
8508e3e3a7aSWarner Losh     int e;
8518e3e3a7aSWarner Losh     lua_Number m = l_mathop(frexp)(x, &e);  /* 'x' fraction and exponent */
8528e3e3a7aSWarner Losh     int n = 0;  /* character count */
8538e3e3a7aSWarner Losh     if (m < 0) {  /* is number negative? */
8548e3e3a7aSWarner Losh       buff[n++] = '-';  /* add signal */
8558e3e3a7aSWarner Losh       m = -m;  /* make it positive */
8568e3e3a7aSWarner Losh     }
8578e3e3a7aSWarner Losh     buff[n++] = '0'; buff[n++] = 'x';  /* add "0x" */
8588e3e3a7aSWarner Losh     m = adddigit(buff, n++, m * (1 << L_NBFD));  /* add first digit */
8598e3e3a7aSWarner Losh     e -= L_NBFD;  /* this digit goes before the radix point */
8608e3e3a7aSWarner Losh     if (m > 0) {  /* more digits? */
8618e3e3a7aSWarner Losh       buff[n++] = lua_getlocaledecpoint();  /* add radix point */
8628e3e3a7aSWarner Losh       do {  /* add as many digits as needed */
8638e3e3a7aSWarner Losh         m = adddigit(buff, n++, m * 16);
8648e3e3a7aSWarner Losh       } while (m > 0);
8658e3e3a7aSWarner Losh     }
8668e3e3a7aSWarner Losh     n += l_sprintf(buff + n, sz - n, "p%+d", e);  /* add exponent */
8678e3e3a7aSWarner Losh     lua_assert(n < sz);
8688e3e3a7aSWarner Losh     return n;
8698e3e3a7aSWarner Losh   }
8708e3e3a7aSWarner Losh }
8718e3e3a7aSWarner Losh 
8728e3e3a7aSWarner Losh 
8738e3e3a7aSWarner Losh static int lua_number2strx (lua_State *L, char *buff, int sz,
8748e3e3a7aSWarner Losh                             const char *fmt, lua_Number x) {
8758e3e3a7aSWarner Losh   int n = num2straux(buff, sz, x);
8768e3e3a7aSWarner Losh   if (fmt[SIZELENMOD] == 'A') {
8778e3e3a7aSWarner Losh     int i;
8788e3e3a7aSWarner Losh     for (i = 0; i < n; i++)
8798e3e3a7aSWarner Losh       buff[i] = toupper(uchar(buff[i]));
8808e3e3a7aSWarner Losh   }
8818e3e3a7aSWarner Losh   else if (fmt[SIZELENMOD] != 'a')
8828e3e3a7aSWarner Losh     luaL_error(L, "modifiers for format '%%a'/'%%A' not implemented");
8838e3e3a7aSWarner Losh   return n;
8848e3e3a7aSWarner Losh }
8858e3e3a7aSWarner Losh 
8868e3e3a7aSWarner Losh #endif				/* } */
8878e3e3a7aSWarner Losh 
8888e3e3a7aSWarner Losh 
8898e3e3a7aSWarner Losh /*
8908e3e3a7aSWarner Losh ** Maximum size of each formatted item. This maximum size is produced
8918e3e3a7aSWarner Losh ** by format('%.99f', -maxfloat), and is equal to 99 + 3 ('-', '.',
8928e3e3a7aSWarner Losh ** and '\0') + number of decimal digits to represent maxfloat (which
8938e3e3a7aSWarner Losh ** is maximum exponent + 1). (99+3+1 then rounded to 120 for "extra
8948e3e3a7aSWarner Losh ** expenses", such as locale-dependent stuff)
8958e3e3a7aSWarner Losh */
8968e3e3a7aSWarner Losh #define MAX_ITEM        (120 + l_mathlim(MAX_10_EXP))
8978e3e3a7aSWarner Losh 
8988e3e3a7aSWarner Losh 
8998e3e3a7aSWarner Losh /* valid flags in a format specification */
9008e3e3a7aSWarner Losh #define FLAGS	"-+ #0"
9018e3e3a7aSWarner Losh 
9028e3e3a7aSWarner Losh /*
9038e3e3a7aSWarner Losh ** maximum size of each format specification (such as "%-099.99d")
9048e3e3a7aSWarner Losh */
9058e3e3a7aSWarner Losh #define MAX_FORMAT	32
9068e3e3a7aSWarner Losh 
9078e3e3a7aSWarner Losh 
9088e3e3a7aSWarner Losh static void addquoted (luaL_Buffer *b, const char *s, size_t len) {
9098e3e3a7aSWarner Losh   luaL_addchar(b, '"');
9108e3e3a7aSWarner Losh   while (len--) {
9118e3e3a7aSWarner Losh     if (*s == '"' || *s == '\\' || *s == '\n') {
9128e3e3a7aSWarner Losh       luaL_addchar(b, '\\');
9138e3e3a7aSWarner Losh       luaL_addchar(b, *s);
9148e3e3a7aSWarner Losh     }
9158e3e3a7aSWarner Losh     else if (iscntrl(uchar(*s))) {
9168e3e3a7aSWarner Losh       char buff[10];
9178e3e3a7aSWarner Losh       if (!isdigit(uchar(*(s+1))))
9188e3e3a7aSWarner Losh         l_sprintf(buff, sizeof(buff), "\\%d", (int)uchar(*s));
9198e3e3a7aSWarner Losh       else
9208e3e3a7aSWarner Losh         l_sprintf(buff, sizeof(buff), "\\%03d", (int)uchar(*s));
9218e3e3a7aSWarner Losh       luaL_addstring(b, buff);
9228e3e3a7aSWarner Losh     }
9238e3e3a7aSWarner Losh     else
9248e3e3a7aSWarner Losh       luaL_addchar(b, *s);
9258e3e3a7aSWarner Losh     s++;
9268e3e3a7aSWarner Losh   }
9278e3e3a7aSWarner Losh   luaL_addchar(b, '"');
9288e3e3a7aSWarner Losh }
9298e3e3a7aSWarner Losh 
9308e3e3a7aSWarner Losh 
931*07d397d7SWarner Losh #if LUA_FLOAT_TYPE != LUA_FLOAT_INT64
9328e3e3a7aSWarner Losh /*
9338e3e3a7aSWarner Losh ** Ensures the 'buff' string uses a dot as the radix character.
9348e3e3a7aSWarner Losh */
9358e3e3a7aSWarner Losh static void checkdp (char *buff, int nb) {
9368e3e3a7aSWarner Losh   if (memchr(buff, '.', nb) == NULL) {  /* no dot? */
9378e3e3a7aSWarner Losh     char point = lua_getlocaledecpoint();  /* try locale point */
9388e3e3a7aSWarner Losh     char *ppoint = (char *)memchr(buff, point, nb);
9398e3e3a7aSWarner Losh     if (ppoint) *ppoint = '.';  /* change it to a dot */
9408e3e3a7aSWarner Losh   }
9418e3e3a7aSWarner Losh }
942*07d397d7SWarner Losh #endif
9438e3e3a7aSWarner Losh 
9448e3e3a7aSWarner Losh static void addliteral (lua_State *L, luaL_Buffer *b, int arg) {
9458e3e3a7aSWarner Losh   switch (lua_type(L, arg)) {
9468e3e3a7aSWarner Losh     case LUA_TSTRING: {
9478e3e3a7aSWarner Losh       size_t len;
9488e3e3a7aSWarner Losh       const char *s = lua_tolstring(L, arg, &len);
9498e3e3a7aSWarner Losh       addquoted(b, s, len);
9508e3e3a7aSWarner Losh       break;
9518e3e3a7aSWarner Losh     }
9528e3e3a7aSWarner Losh     case LUA_TNUMBER: {
9538e3e3a7aSWarner Losh       char *buff = luaL_prepbuffsize(b, MAX_ITEM);
9548e3e3a7aSWarner Losh       int nb;
955919cf86cSWarner Losh #if LUA_FLOAT_TYPE != LUA_FLOAT_INT64
9568e3e3a7aSWarner Losh       if (!lua_isinteger(L, arg)) {  /* float? */
9578e3e3a7aSWarner Losh         lua_Number n = lua_tonumber(L, arg);  /* write as hexa ('%a') */
9588e3e3a7aSWarner Losh         nb = lua_number2strx(L, buff, MAX_ITEM, "%" LUA_NUMBER_FRMLEN "a", n);
9598e3e3a7aSWarner Losh         checkdp(buff, nb);  /* ensure it uses a dot */
9608e3e3a7aSWarner Losh       }
9618e3e3a7aSWarner Losh       else {  /* integers */
962919cf86cSWarner Losh #else
963919cf86cSWarner Losh       {
964919cf86cSWarner Losh #endif
9658e3e3a7aSWarner Losh         lua_Integer n = lua_tointeger(L, arg);
9668e3e3a7aSWarner Losh         const char *format = (n == LUA_MININTEGER)  /* corner case? */
9678e3e3a7aSWarner Losh                            ? "0x%" LUA_INTEGER_FRMLEN "x"  /* use hexa */
9688e3e3a7aSWarner Losh                            : LUA_INTEGER_FMT;  /* else use default format */
9698e3e3a7aSWarner Losh         nb = l_sprintf(buff, MAX_ITEM, format, (LUAI_UACINT)n);
9708e3e3a7aSWarner Losh       }
9718e3e3a7aSWarner Losh       luaL_addsize(b, nb);
9728e3e3a7aSWarner Losh       break;
9738e3e3a7aSWarner Losh     }
9748e3e3a7aSWarner Losh     case LUA_TNIL: case LUA_TBOOLEAN: {
9758e3e3a7aSWarner Losh       luaL_tolstring(L, arg, NULL);
9768e3e3a7aSWarner Losh       luaL_addvalue(b);
9778e3e3a7aSWarner Losh       break;
9788e3e3a7aSWarner Losh     }
9798e3e3a7aSWarner Losh     default: {
9808e3e3a7aSWarner Losh       luaL_argerror(L, arg, "value has no literal form");
9818e3e3a7aSWarner Losh     }
9828e3e3a7aSWarner Losh   }
9838e3e3a7aSWarner Losh }
9848e3e3a7aSWarner Losh 
9858e3e3a7aSWarner Losh 
9868e3e3a7aSWarner Losh static const char *scanformat (lua_State *L, const char *strfrmt, char *form) {
9878e3e3a7aSWarner Losh   const char *p = strfrmt;
9888e3e3a7aSWarner Losh   while (*p != '\0' && strchr(FLAGS, *p) != NULL) p++;  /* skip flags */
9898e3e3a7aSWarner Losh   if ((size_t)(p - strfrmt) >= sizeof(FLAGS)/sizeof(char))
9908e3e3a7aSWarner Losh     luaL_error(L, "invalid format (repeated flags)");
9918e3e3a7aSWarner Losh   if (isdigit(uchar(*p))) p++;  /* skip width */
9928e3e3a7aSWarner Losh   if (isdigit(uchar(*p))) p++;  /* (2 digits at most) */
9938e3e3a7aSWarner Losh   if (*p == '.') {
9948e3e3a7aSWarner Losh     p++;
9958e3e3a7aSWarner Losh     if (isdigit(uchar(*p))) p++;  /* skip precision */
9968e3e3a7aSWarner Losh     if (isdigit(uchar(*p))) p++;  /* (2 digits at most) */
9978e3e3a7aSWarner Losh   }
9988e3e3a7aSWarner Losh   if (isdigit(uchar(*p)))
9998e3e3a7aSWarner Losh     luaL_error(L, "invalid format (width or precision too long)");
10008e3e3a7aSWarner Losh   *(form++) = '%';
10018e3e3a7aSWarner Losh   memcpy(form, strfrmt, ((p - strfrmt) + 1) * sizeof(char));
10028e3e3a7aSWarner Losh   form += (p - strfrmt) + 1;
10038e3e3a7aSWarner Losh   *form = '\0';
10048e3e3a7aSWarner Losh   return p;
10058e3e3a7aSWarner Losh }
10068e3e3a7aSWarner Losh 
10078e3e3a7aSWarner Losh 
10088e3e3a7aSWarner Losh /*
10098e3e3a7aSWarner Losh ** add length modifier into formats
10108e3e3a7aSWarner Losh */
10118e3e3a7aSWarner Losh static void addlenmod (char *form, const char *lenmod) {
10128e3e3a7aSWarner Losh   size_t l = strlen(form);
10138e3e3a7aSWarner Losh   size_t lm = strlen(lenmod);
10148e3e3a7aSWarner Losh   char spec = form[l - 1];
10158e3e3a7aSWarner Losh   strcpy(form + l - 1, lenmod);
10168e3e3a7aSWarner Losh   form[l + lm - 1] = spec;
10178e3e3a7aSWarner Losh   form[l + lm] = '\0';
10188e3e3a7aSWarner Losh }
10198e3e3a7aSWarner Losh 
10208e3e3a7aSWarner Losh 
10218e3e3a7aSWarner Losh static int str_format (lua_State *L) {
10228e3e3a7aSWarner Losh   int top = lua_gettop(L);
10238e3e3a7aSWarner Losh   int arg = 1;
10248e3e3a7aSWarner Losh   size_t sfl;
10258e3e3a7aSWarner Losh   const char *strfrmt = luaL_checklstring(L, arg, &sfl);
10268e3e3a7aSWarner Losh   const char *strfrmt_end = strfrmt+sfl;
10278e3e3a7aSWarner Losh   luaL_Buffer b;
10288e3e3a7aSWarner Losh   luaL_buffinit(L, &b);
10298e3e3a7aSWarner Losh   while (strfrmt < strfrmt_end) {
10308e3e3a7aSWarner Losh     if (*strfrmt != L_ESC)
10318e3e3a7aSWarner Losh       luaL_addchar(&b, *strfrmt++);
10328e3e3a7aSWarner Losh     else if (*++strfrmt == L_ESC)
10338e3e3a7aSWarner Losh       luaL_addchar(&b, *strfrmt++);  /* %% */
10348e3e3a7aSWarner Losh     else { /* format item */
10358e3e3a7aSWarner Losh       char form[MAX_FORMAT];  /* to store the format ('%...') */
10368e3e3a7aSWarner Losh       char *buff = luaL_prepbuffsize(&b, MAX_ITEM);  /* to put formatted item */
10378e3e3a7aSWarner Losh       int nb = 0;  /* number of bytes in added item */
10388e3e3a7aSWarner Losh       if (++arg > top)
10398e3e3a7aSWarner Losh         luaL_argerror(L, arg, "no value");
10408e3e3a7aSWarner Losh       strfrmt = scanformat(L, strfrmt, form);
10418e3e3a7aSWarner Losh       switch (*strfrmt++) {
10428e3e3a7aSWarner Losh         case 'c': {
10438e3e3a7aSWarner Losh           nb = l_sprintf(buff, MAX_ITEM, form, (int)luaL_checkinteger(L, arg));
10448e3e3a7aSWarner Losh           break;
10458e3e3a7aSWarner Losh         }
10468e3e3a7aSWarner Losh         case 'd': case 'i':
10478e3e3a7aSWarner Losh         case 'o': case 'u': case 'x': case 'X': {
10488e3e3a7aSWarner Losh           lua_Integer n = luaL_checkinteger(L, arg);
10498e3e3a7aSWarner Losh           addlenmod(form, LUA_INTEGER_FRMLEN);
10508e3e3a7aSWarner Losh           nb = l_sprintf(buff, MAX_ITEM, form, (LUAI_UACINT)n);
10518e3e3a7aSWarner Losh           break;
10528e3e3a7aSWarner Losh         }
10538e3e3a7aSWarner Losh         case 'a': case 'A':
10548e3e3a7aSWarner Losh           addlenmod(form, LUA_NUMBER_FRMLEN);
10558e3e3a7aSWarner Losh           nb = lua_number2strx(L, buff, MAX_ITEM, form,
10568e3e3a7aSWarner Losh                                   luaL_checknumber(L, arg));
10578e3e3a7aSWarner Losh           break;
10588e3e3a7aSWarner Losh         case 'e': case 'E': case 'f':
10598e3e3a7aSWarner Losh         case 'g': case 'G': {
10608e3e3a7aSWarner Losh           lua_Number n = luaL_checknumber(L, arg);
10618e3e3a7aSWarner Losh           addlenmod(form, LUA_NUMBER_FRMLEN);
10628e3e3a7aSWarner Losh           nb = l_sprintf(buff, MAX_ITEM, form, (LUAI_UACNUMBER)n);
10638e3e3a7aSWarner Losh           break;
10648e3e3a7aSWarner Losh         }
10658e3e3a7aSWarner Losh         case 'q': {
10668e3e3a7aSWarner Losh           addliteral(L, &b, arg);
10678e3e3a7aSWarner Losh           break;
10688e3e3a7aSWarner Losh         }
10698e3e3a7aSWarner Losh         case 's': {
10708e3e3a7aSWarner Losh           size_t l;
10718e3e3a7aSWarner Losh           const char *s = luaL_tolstring(L, arg, &l);
10728e3e3a7aSWarner Losh           if (form[2] == '\0')  /* no modifiers? */
10738e3e3a7aSWarner Losh             luaL_addvalue(&b);  /* keep entire string */
10748e3e3a7aSWarner Losh           else {
10758e3e3a7aSWarner Losh             luaL_argcheck(L, l == strlen(s), arg, "string contains zeros");
10768e3e3a7aSWarner Losh             if (!strchr(form, '.') && l >= 100) {
10778e3e3a7aSWarner Losh               /* no precision and string is too long to be formatted */
10788e3e3a7aSWarner Losh               luaL_addvalue(&b);  /* keep entire string */
10798e3e3a7aSWarner Losh             }
10808e3e3a7aSWarner Losh             else {  /* format the string into 'buff' */
10818e3e3a7aSWarner Losh               nb = l_sprintf(buff, MAX_ITEM, form, s);
10828e3e3a7aSWarner Losh               lua_pop(L, 1);  /* remove result from 'luaL_tolstring' */
10838e3e3a7aSWarner Losh             }
10848e3e3a7aSWarner Losh           }
10858e3e3a7aSWarner Losh           break;
10868e3e3a7aSWarner Losh         }
10878e3e3a7aSWarner Losh         default: {  /* also treat cases 'pnLlh' */
10888e3e3a7aSWarner Losh           return luaL_error(L, "invalid option '%%%c' to 'format'",
10898e3e3a7aSWarner Losh                                *(strfrmt - 1));
10908e3e3a7aSWarner Losh         }
10918e3e3a7aSWarner Losh       }
10928e3e3a7aSWarner Losh       lua_assert(nb < MAX_ITEM);
10938e3e3a7aSWarner Losh       luaL_addsize(&b, nb);
10948e3e3a7aSWarner Losh     }
10958e3e3a7aSWarner Losh   }
10968e3e3a7aSWarner Losh   luaL_pushresult(&b);
10978e3e3a7aSWarner Losh   return 1;
10988e3e3a7aSWarner Losh }
10998e3e3a7aSWarner Losh 
11008e3e3a7aSWarner Losh /* }====================================================== */
11018e3e3a7aSWarner Losh 
11028e3e3a7aSWarner Losh 
11038e3e3a7aSWarner Losh /*
11048e3e3a7aSWarner Losh ** {======================================================
11058e3e3a7aSWarner Losh ** PACK/UNPACK
11068e3e3a7aSWarner Losh ** =======================================================
11078e3e3a7aSWarner Losh */
11088e3e3a7aSWarner Losh 
11098e3e3a7aSWarner Losh 
11108e3e3a7aSWarner Losh /* value used for padding */
11118e3e3a7aSWarner Losh #if !defined(LUAL_PACKPADBYTE)
11128e3e3a7aSWarner Losh #define LUAL_PACKPADBYTE		0x00
11138e3e3a7aSWarner Losh #endif
11148e3e3a7aSWarner Losh 
11158e3e3a7aSWarner Losh /* maximum size for the binary representation of an integer */
11168e3e3a7aSWarner Losh #define MAXINTSIZE	16
11178e3e3a7aSWarner Losh 
11188e3e3a7aSWarner Losh /* number of bits in a character */
11198e3e3a7aSWarner Losh #define NB	CHAR_BIT
11208e3e3a7aSWarner Losh 
11218e3e3a7aSWarner Losh /* mask for one character (NB 1's) */
11228e3e3a7aSWarner Losh #define MC	((1 << NB) - 1)
11238e3e3a7aSWarner Losh 
11248e3e3a7aSWarner Losh /* size of a lua_Integer */
11258e3e3a7aSWarner Losh #define SZINT	((int)sizeof(lua_Integer))
11268e3e3a7aSWarner Losh 
11278e3e3a7aSWarner Losh 
11288e3e3a7aSWarner Losh /* dummy union to get native endianness */
11298e3e3a7aSWarner Losh static const union {
11308e3e3a7aSWarner Losh   int dummy;
11318e3e3a7aSWarner Losh   char little;  /* true iff machine is little endian */
11328e3e3a7aSWarner Losh } nativeendian = {1};
11338e3e3a7aSWarner Losh 
11348e3e3a7aSWarner Losh 
11358e3e3a7aSWarner Losh /* dummy structure to get native alignment requirements */
11368e3e3a7aSWarner Losh struct cD {
11378e3e3a7aSWarner Losh   char c;
113807c17b2bSWarner Losh   union {
113907c17b2bSWarner Losh #if LUA_FLOAT_TYPE != LUA_FLOAT_INT64
114007c17b2bSWarner Losh 	  double d;
114107c17b2bSWarner Losh #endif
114207c17b2bSWarner Losh 	  void *p; lua_Integer i; lua_Number n; } u;
11438e3e3a7aSWarner Losh };
11448e3e3a7aSWarner Losh 
11458e3e3a7aSWarner Losh #define MAXALIGN	(offsetof(struct cD, u))
11468e3e3a7aSWarner Losh 
11478e3e3a7aSWarner Losh 
11488e3e3a7aSWarner Losh /*
11498e3e3a7aSWarner Losh ** Union for serializing floats
11508e3e3a7aSWarner Losh */
11518e3e3a7aSWarner Losh typedef union Ftypes {
115207c17b2bSWarner Losh #if LUA_FLOAT_TYPE != LUA_FLOAT_INT64
11538e3e3a7aSWarner Losh   float f;
11548e3e3a7aSWarner Losh   double d;
115507c17b2bSWarner Losh #endif
11568e3e3a7aSWarner Losh   lua_Number n;
11578e3e3a7aSWarner Losh   char buff[5 * sizeof(lua_Number)];  /* enough for any float type */
11588e3e3a7aSWarner Losh } Ftypes;
11598e3e3a7aSWarner Losh 
11608e3e3a7aSWarner Losh 
11618e3e3a7aSWarner Losh /*
11628e3e3a7aSWarner Losh ** information to pack/unpack stuff
11638e3e3a7aSWarner Losh */
11648e3e3a7aSWarner Losh typedef struct Header {
11658e3e3a7aSWarner Losh   lua_State *L;
11668e3e3a7aSWarner Losh   int islittle;
11678e3e3a7aSWarner Losh   int maxalign;
11688e3e3a7aSWarner Losh } Header;
11698e3e3a7aSWarner Losh 
11708e3e3a7aSWarner Losh 
11718e3e3a7aSWarner Losh /*
11728e3e3a7aSWarner Losh ** options for pack/unpack
11738e3e3a7aSWarner Losh */
11748e3e3a7aSWarner Losh typedef enum KOption {
11758e3e3a7aSWarner Losh   Kint,		/* signed integers */
11768e3e3a7aSWarner Losh   Kuint,	/* unsigned integers */
11778e3e3a7aSWarner Losh   Kfloat,	/* floating-point numbers */
11788e3e3a7aSWarner Losh   Kchar,	/* fixed-length strings */
11798e3e3a7aSWarner Losh   Kstring,	/* strings with prefixed length */
11808e3e3a7aSWarner Losh   Kzstr,	/* zero-terminated strings */
11818e3e3a7aSWarner Losh   Kpadding,	/* padding */
11828e3e3a7aSWarner Losh   Kpaddalign,	/* padding for alignment */
11838e3e3a7aSWarner Losh   Knop		/* no-op (configuration or spaces) */
11848e3e3a7aSWarner Losh } KOption;
11858e3e3a7aSWarner Losh 
11868e3e3a7aSWarner Losh 
11878e3e3a7aSWarner Losh /*
11888e3e3a7aSWarner Losh ** Read an integer numeral from string 'fmt' or return 'df' if
11898e3e3a7aSWarner Losh ** there is no numeral
11908e3e3a7aSWarner Losh */
11918e3e3a7aSWarner Losh static int digit (int c) { return '0' <= c && c <= '9'; }
11928e3e3a7aSWarner Losh 
11938e3e3a7aSWarner Losh static int getnum (const char **fmt, int df) {
11948e3e3a7aSWarner Losh   if (!digit(**fmt))  /* no number? */
11958e3e3a7aSWarner Losh     return df;  /* return default value */
11968e3e3a7aSWarner Losh   else {
11978e3e3a7aSWarner Losh     int a = 0;
11988e3e3a7aSWarner Losh     do {
11998e3e3a7aSWarner Losh       a = a*10 + (*((*fmt)++) - '0');
12008e3e3a7aSWarner Losh     } while (digit(**fmt) && a <= ((int)MAXSIZE - 9)/10);
12018e3e3a7aSWarner Losh     return a;
12028e3e3a7aSWarner Losh   }
12038e3e3a7aSWarner Losh }
12048e3e3a7aSWarner Losh 
12058e3e3a7aSWarner Losh 
12068e3e3a7aSWarner Losh /*
12078e3e3a7aSWarner Losh ** Read an integer numeral and raises an error if it is larger
12088e3e3a7aSWarner Losh ** than the maximum size for integers.
12098e3e3a7aSWarner Losh */
12108e3e3a7aSWarner Losh static int getnumlimit (Header *h, const char **fmt, int df) {
12118e3e3a7aSWarner Losh   int sz = getnum(fmt, df);
12128e3e3a7aSWarner Losh   if (sz > MAXINTSIZE || sz <= 0)
12138e3e3a7aSWarner Losh     luaL_error(h->L, "integral size (%d) out of limits [1,%d]",
12148e3e3a7aSWarner Losh                      sz, MAXINTSIZE);
12158e3e3a7aSWarner Losh   return sz;
12168e3e3a7aSWarner Losh }
12178e3e3a7aSWarner Losh 
12188e3e3a7aSWarner Losh 
12198e3e3a7aSWarner Losh /*
12208e3e3a7aSWarner Losh ** Initialize Header
12218e3e3a7aSWarner Losh */
12228e3e3a7aSWarner Losh static void initheader (lua_State *L, Header *h) {
12238e3e3a7aSWarner Losh   h->L = L;
12248e3e3a7aSWarner Losh   h->islittle = nativeendian.little;
12258e3e3a7aSWarner Losh   h->maxalign = 1;
12268e3e3a7aSWarner Losh }
12278e3e3a7aSWarner Losh 
12288e3e3a7aSWarner Losh 
12298e3e3a7aSWarner Losh /*
12308e3e3a7aSWarner Losh ** Read and classify next option. 'size' is filled with option's size.
12318e3e3a7aSWarner Losh */
12328e3e3a7aSWarner Losh static KOption getoption (Header *h, const char **fmt, int *size) {
12338e3e3a7aSWarner Losh   int opt = *((*fmt)++);
12348e3e3a7aSWarner Losh   *size = 0;  /* default */
12358e3e3a7aSWarner Losh   switch (opt) {
12368e3e3a7aSWarner Losh     case 'b': *size = sizeof(char); return Kint;
12378e3e3a7aSWarner Losh     case 'B': *size = sizeof(char); return Kuint;
12388e3e3a7aSWarner Losh     case 'h': *size = sizeof(short); return Kint;
12398e3e3a7aSWarner Losh     case 'H': *size = sizeof(short); return Kuint;
12408e3e3a7aSWarner Losh     case 'l': *size = sizeof(long); return Kint;
12418e3e3a7aSWarner Losh     case 'L': *size = sizeof(long); return Kuint;
12428e3e3a7aSWarner Losh     case 'j': *size = sizeof(lua_Integer); return Kint;
12438e3e3a7aSWarner Losh     case 'J': *size = sizeof(lua_Integer); return Kuint;
12448e3e3a7aSWarner Losh     case 'T': *size = sizeof(size_t); return Kuint;
124507c17b2bSWarner Losh #if LUA_FLOAT_TYPE != LUA_FLOAT_INT64
12468e3e3a7aSWarner Losh     case 'f': *size = sizeof(float); return Kfloat;
12478e3e3a7aSWarner Losh     case 'd': *size = sizeof(double); return Kfloat;
124807c17b2bSWarner Losh #endif
12498e3e3a7aSWarner Losh     case 'n': *size = sizeof(lua_Number); return Kfloat;
12508e3e3a7aSWarner Losh     case 'i': *size = getnumlimit(h, fmt, sizeof(int)); return Kint;
12518e3e3a7aSWarner Losh     case 'I': *size = getnumlimit(h, fmt, sizeof(int)); return Kuint;
12528e3e3a7aSWarner Losh     case 's': *size = getnumlimit(h, fmt, sizeof(size_t)); return Kstring;
12538e3e3a7aSWarner Losh     case 'c':
12548e3e3a7aSWarner Losh       *size = getnum(fmt, -1);
12558e3e3a7aSWarner Losh       if (*size == -1)
12568e3e3a7aSWarner Losh         luaL_error(h->L, "missing size for format option 'c'");
12578e3e3a7aSWarner Losh       return Kchar;
12588e3e3a7aSWarner Losh     case 'z': return Kzstr;
12598e3e3a7aSWarner Losh     case 'x': *size = 1; return Kpadding;
12608e3e3a7aSWarner Losh     case 'X': return Kpaddalign;
12618e3e3a7aSWarner Losh     case ' ': break;
12628e3e3a7aSWarner Losh     case '<': h->islittle = 1; break;
12638e3e3a7aSWarner Losh     case '>': h->islittle = 0; break;
12648e3e3a7aSWarner Losh     case '=': h->islittle = nativeendian.little; break;
12658e3e3a7aSWarner Losh     case '!': h->maxalign = getnumlimit(h, fmt, MAXALIGN); break;
12668e3e3a7aSWarner Losh     default: luaL_error(h->L, "invalid format option '%c'", opt);
12678e3e3a7aSWarner Losh   }
12688e3e3a7aSWarner Losh   return Knop;
12698e3e3a7aSWarner Losh }
12708e3e3a7aSWarner Losh 
12718e3e3a7aSWarner Losh 
12728e3e3a7aSWarner Losh /*
12738e3e3a7aSWarner Losh ** Read, classify, and fill other details about the next option.
12748e3e3a7aSWarner Losh ** 'psize' is filled with option's size, 'notoalign' with its
12758e3e3a7aSWarner Losh ** alignment requirements.
12768e3e3a7aSWarner Losh ** Local variable 'size' gets the size to be aligned. (Kpadal option
12778e3e3a7aSWarner Losh ** always gets its full alignment, other options are limited by
12788e3e3a7aSWarner Losh ** the maximum alignment ('maxalign'). Kchar option needs no alignment
12798e3e3a7aSWarner Losh ** despite its size.
12808e3e3a7aSWarner Losh */
12818e3e3a7aSWarner Losh static KOption getdetails (Header *h, size_t totalsize,
12828e3e3a7aSWarner Losh                            const char **fmt, int *psize, int *ntoalign) {
12838e3e3a7aSWarner Losh   KOption opt = getoption(h, fmt, psize);
12848e3e3a7aSWarner Losh   int align = *psize;  /* usually, alignment follows size */
12858e3e3a7aSWarner Losh   if (opt == Kpaddalign) {  /* 'X' gets alignment from following option */
12868e3e3a7aSWarner Losh     if (**fmt == '\0' || getoption(h, fmt, &align) == Kchar || align == 0)
12878e3e3a7aSWarner Losh       luaL_argerror(h->L, 1, "invalid next option for option 'X'");
12888e3e3a7aSWarner Losh   }
12898e3e3a7aSWarner Losh   if (align <= 1 || opt == Kchar)  /* need no alignment? */
12908e3e3a7aSWarner Losh     *ntoalign = 0;
12918e3e3a7aSWarner Losh   else {
12928e3e3a7aSWarner Losh     if (align > h->maxalign)  /* enforce maximum alignment */
12938e3e3a7aSWarner Losh       align = h->maxalign;
12948e3e3a7aSWarner Losh     if ((align & (align - 1)) != 0)  /* is 'align' not a power of 2? */
12958e3e3a7aSWarner Losh       luaL_argerror(h->L, 1, "format asks for alignment not power of 2");
12968e3e3a7aSWarner Losh     *ntoalign = (align - (int)(totalsize & (align - 1))) & (align - 1);
12978e3e3a7aSWarner Losh   }
12988e3e3a7aSWarner Losh   return opt;
12998e3e3a7aSWarner Losh }
13008e3e3a7aSWarner Losh 
13018e3e3a7aSWarner Losh 
13028e3e3a7aSWarner Losh /*
13038e3e3a7aSWarner Losh ** Pack integer 'n' with 'size' bytes and 'islittle' endianness.
13048e3e3a7aSWarner Losh ** The final 'if' handles the case when 'size' is larger than
13058e3e3a7aSWarner Losh ** the size of a Lua integer, correcting the extra sign-extension
13068e3e3a7aSWarner Losh ** bytes if necessary (by default they would be zeros).
13078e3e3a7aSWarner Losh */
13088e3e3a7aSWarner Losh static void packint (luaL_Buffer *b, lua_Unsigned n,
13098e3e3a7aSWarner Losh                      int islittle, int size, int neg) {
13108e3e3a7aSWarner Losh   char *buff = luaL_prepbuffsize(b, size);
13118e3e3a7aSWarner Losh   int i;
13128e3e3a7aSWarner Losh   buff[islittle ? 0 : size - 1] = (char)(n & MC);  /* first byte */
13138e3e3a7aSWarner Losh   for (i = 1; i < size; i++) {
13148e3e3a7aSWarner Losh     n >>= NB;
13158e3e3a7aSWarner Losh     buff[islittle ? i : size - 1 - i] = (char)(n & MC);
13168e3e3a7aSWarner Losh   }
13178e3e3a7aSWarner Losh   if (neg && size > SZINT) {  /* negative number need sign extension? */
13188e3e3a7aSWarner Losh     for (i = SZINT; i < size; i++)  /* correct extra bytes */
13198e3e3a7aSWarner Losh       buff[islittle ? i : size - 1 - i] = (char)MC;
13208e3e3a7aSWarner Losh   }
13218e3e3a7aSWarner Losh   luaL_addsize(b, size);  /* add result to buffer */
13228e3e3a7aSWarner Losh }
13238e3e3a7aSWarner Losh 
13248e3e3a7aSWarner Losh 
13258e3e3a7aSWarner Losh /*
13268e3e3a7aSWarner Losh ** Copy 'size' bytes from 'src' to 'dest', correcting endianness if
13278e3e3a7aSWarner Losh ** given 'islittle' is different from native endianness.
13288e3e3a7aSWarner Losh */
13298e3e3a7aSWarner Losh static void copywithendian (volatile char *dest, volatile const char *src,
13308e3e3a7aSWarner Losh                             int size, int islittle) {
13318e3e3a7aSWarner Losh   if (islittle == nativeendian.little) {
13328e3e3a7aSWarner Losh     while (size-- != 0)
13338e3e3a7aSWarner Losh       *(dest++) = *(src++);
13348e3e3a7aSWarner Losh   }
13358e3e3a7aSWarner Losh   else {
13368e3e3a7aSWarner Losh     dest += size - 1;
13378e3e3a7aSWarner Losh     while (size-- != 0)
13388e3e3a7aSWarner Losh       *(dest--) = *(src++);
13398e3e3a7aSWarner Losh   }
13408e3e3a7aSWarner Losh }
13418e3e3a7aSWarner Losh 
13428e3e3a7aSWarner Losh 
13438e3e3a7aSWarner Losh static int str_pack (lua_State *L) {
13448e3e3a7aSWarner Losh   luaL_Buffer b;
13458e3e3a7aSWarner Losh   Header h;
13468e3e3a7aSWarner Losh   const char *fmt = luaL_checkstring(L, 1);  /* format string */
13478e3e3a7aSWarner Losh   int arg = 1;  /* current argument to pack */
13488e3e3a7aSWarner Losh   size_t totalsize = 0;  /* accumulate total size of result */
13498e3e3a7aSWarner Losh   initheader(L, &h);
13508e3e3a7aSWarner Losh   lua_pushnil(L);  /* mark to separate arguments from string buffer */
13518e3e3a7aSWarner Losh   luaL_buffinit(L, &b);
13528e3e3a7aSWarner Losh   while (*fmt != '\0') {
13538e3e3a7aSWarner Losh     int size, ntoalign;
13548e3e3a7aSWarner Losh     KOption opt = getdetails(&h, totalsize, &fmt, &size, &ntoalign);
13558e3e3a7aSWarner Losh     totalsize += ntoalign + size;
13568e3e3a7aSWarner Losh     while (ntoalign-- > 0)
13578e3e3a7aSWarner Losh      luaL_addchar(&b, LUAL_PACKPADBYTE);  /* fill alignment */
13588e3e3a7aSWarner Losh     arg++;
13598e3e3a7aSWarner Losh     switch (opt) {
13608e3e3a7aSWarner Losh       case Kint: {  /* signed integers */
13618e3e3a7aSWarner Losh         lua_Integer n = luaL_checkinteger(L, arg);
13628e3e3a7aSWarner Losh         if (size < SZINT) {  /* need overflow check? */
13638e3e3a7aSWarner Losh           lua_Integer lim = (lua_Integer)1 << ((size * NB) - 1);
13648e3e3a7aSWarner Losh           luaL_argcheck(L, -lim <= n && n < lim, arg, "integer overflow");
13658e3e3a7aSWarner Losh         }
13668e3e3a7aSWarner Losh         packint(&b, (lua_Unsigned)n, h.islittle, size, (n < 0));
13678e3e3a7aSWarner Losh         break;
13688e3e3a7aSWarner Losh       }
13698e3e3a7aSWarner Losh       case Kuint: {  /* unsigned integers */
13708e3e3a7aSWarner Losh         lua_Integer n = luaL_checkinteger(L, arg);
13718e3e3a7aSWarner Losh         if (size < SZINT)  /* need overflow check? */
13728e3e3a7aSWarner Losh           luaL_argcheck(L, (lua_Unsigned)n < ((lua_Unsigned)1 << (size * NB)),
13738e3e3a7aSWarner Losh                            arg, "unsigned overflow");
13748e3e3a7aSWarner Losh         packint(&b, (lua_Unsigned)n, h.islittle, size, 0);
13758e3e3a7aSWarner Losh         break;
13768e3e3a7aSWarner Losh       }
13778e3e3a7aSWarner Losh       case Kfloat: {  /* floating-point options */
13788e3e3a7aSWarner Losh         volatile Ftypes u;
13798e3e3a7aSWarner Losh         char *buff = luaL_prepbuffsize(&b, size);
13808e3e3a7aSWarner Losh         lua_Number n = luaL_checknumber(L, arg);  /* get argument */
138107c17b2bSWarner Losh #if LUA_FLOAT_TYPE != LUA_FLOAT_INT64
13828e3e3a7aSWarner Losh         if (size == sizeof(u.f)) u.f = (float)n;  /* copy it into 'u' */
13838e3e3a7aSWarner Losh         else if (size == sizeof(u.d)) u.d = (double)n;
13848e3e3a7aSWarner Losh         else u.n = n;
138507c17b2bSWarner Losh #else
138607c17b2bSWarner Losh 	u.n = n;
138707c17b2bSWarner Losh #endif
13888e3e3a7aSWarner Losh         /* move 'u' to final result, correcting endianness if needed */
13898e3e3a7aSWarner Losh         copywithendian(buff, u.buff, size, h.islittle);
13908e3e3a7aSWarner Losh         luaL_addsize(&b, size);
13918e3e3a7aSWarner Losh         break;
13928e3e3a7aSWarner Losh       }
13938e3e3a7aSWarner Losh       case Kchar: {  /* fixed-size string */
13948e3e3a7aSWarner Losh         size_t len;
13958e3e3a7aSWarner Losh         const char *s = luaL_checklstring(L, arg, &len);
13968e3e3a7aSWarner Losh         luaL_argcheck(L, len <= (size_t)size, arg,
13978e3e3a7aSWarner Losh                          "string longer than given size");
13988e3e3a7aSWarner Losh         luaL_addlstring(&b, s, len);  /* add string */
13998e3e3a7aSWarner Losh         while (len++ < (size_t)size)  /* pad extra space */
14008e3e3a7aSWarner Losh           luaL_addchar(&b, LUAL_PACKPADBYTE);
14018e3e3a7aSWarner Losh         break;
14028e3e3a7aSWarner Losh       }
14038e3e3a7aSWarner Losh       case Kstring: {  /* strings with length count */
14048e3e3a7aSWarner Losh         size_t len;
14058e3e3a7aSWarner Losh         const char *s = luaL_checklstring(L, arg, &len);
14068e3e3a7aSWarner Losh         luaL_argcheck(L, size >= (int)sizeof(size_t) ||
14078e3e3a7aSWarner Losh                          len < ((size_t)1 << (size * NB)),
14088e3e3a7aSWarner Losh                          arg, "string length does not fit in given size");
14098e3e3a7aSWarner Losh         packint(&b, (lua_Unsigned)len, h.islittle, size, 0);  /* pack length */
14108e3e3a7aSWarner Losh         luaL_addlstring(&b, s, len);
14118e3e3a7aSWarner Losh         totalsize += len;
14128e3e3a7aSWarner Losh         break;
14138e3e3a7aSWarner Losh       }
14148e3e3a7aSWarner Losh       case Kzstr: {  /* zero-terminated string */
14158e3e3a7aSWarner Losh         size_t len;
14168e3e3a7aSWarner Losh         const char *s = luaL_checklstring(L, arg, &len);
14178e3e3a7aSWarner Losh         luaL_argcheck(L, strlen(s) == len, arg, "string contains zeros");
14188e3e3a7aSWarner Losh         luaL_addlstring(&b, s, len);
14198e3e3a7aSWarner Losh         luaL_addchar(&b, '\0');  /* add zero at the end */
14208e3e3a7aSWarner Losh         totalsize += len + 1;
14218e3e3a7aSWarner Losh         break;
14228e3e3a7aSWarner Losh       }
14238e3e3a7aSWarner Losh       case Kpadding: luaL_addchar(&b, LUAL_PACKPADBYTE);  /* FALLTHROUGH */
14248e3e3a7aSWarner Losh       case Kpaddalign: case Knop:
14258e3e3a7aSWarner Losh         arg--;  /* undo increment */
14268e3e3a7aSWarner Losh         break;
14278e3e3a7aSWarner Losh     }
14288e3e3a7aSWarner Losh   }
14298e3e3a7aSWarner Losh   luaL_pushresult(&b);
14308e3e3a7aSWarner Losh   return 1;
14318e3e3a7aSWarner Losh }
14328e3e3a7aSWarner Losh 
14338e3e3a7aSWarner Losh 
14348e3e3a7aSWarner Losh static int str_packsize (lua_State *L) {
14358e3e3a7aSWarner Losh   Header h;
14368e3e3a7aSWarner Losh   const char *fmt = luaL_checkstring(L, 1);  /* format string */
14378e3e3a7aSWarner Losh   size_t totalsize = 0;  /* accumulate total size of result */
14388e3e3a7aSWarner Losh   initheader(L, &h);
14398e3e3a7aSWarner Losh   while (*fmt != '\0') {
14408e3e3a7aSWarner Losh     int size, ntoalign;
14418e3e3a7aSWarner Losh     KOption opt = getdetails(&h, totalsize, &fmt, &size, &ntoalign);
14428e3e3a7aSWarner Losh     size += ntoalign;  /* total space used by option */
14438e3e3a7aSWarner Losh     luaL_argcheck(L, totalsize <= MAXSIZE - size, 1,
14448e3e3a7aSWarner Losh                      "format result too large");
14458e3e3a7aSWarner Losh     totalsize += size;
14468e3e3a7aSWarner Losh     switch (opt) {
14478e3e3a7aSWarner Losh       case Kstring:  /* strings with length count */
14488e3e3a7aSWarner Losh       case Kzstr:    /* zero-terminated string */
14498e3e3a7aSWarner Losh         luaL_argerror(L, 1, "variable-length format");
14508e3e3a7aSWarner Losh         /* call never return, but to avoid warnings: *//* FALLTHROUGH */
14518e3e3a7aSWarner Losh       default:  break;
14528e3e3a7aSWarner Losh     }
14538e3e3a7aSWarner Losh   }
14548e3e3a7aSWarner Losh   lua_pushinteger(L, (lua_Integer)totalsize);
14558e3e3a7aSWarner Losh   return 1;
14568e3e3a7aSWarner Losh }
14578e3e3a7aSWarner Losh 
14588e3e3a7aSWarner Losh 
14598e3e3a7aSWarner Losh /*
14608e3e3a7aSWarner Losh ** Unpack an integer with 'size' bytes and 'islittle' endianness.
14618e3e3a7aSWarner Losh ** If size is smaller than the size of a Lua integer and integer
14628e3e3a7aSWarner Losh ** is signed, must do sign extension (propagating the sign to the
14638e3e3a7aSWarner Losh ** higher bits); if size is larger than the size of a Lua integer,
14648e3e3a7aSWarner Losh ** it must check the unread bytes to see whether they do not cause an
14658e3e3a7aSWarner Losh ** overflow.
14668e3e3a7aSWarner Losh */
14678e3e3a7aSWarner Losh static lua_Integer unpackint (lua_State *L, const char *str,
14688e3e3a7aSWarner Losh                               int islittle, int size, int issigned) {
14698e3e3a7aSWarner Losh   lua_Unsigned res = 0;
14708e3e3a7aSWarner Losh   int i;
14718e3e3a7aSWarner Losh   int limit = (size  <= SZINT) ? size : SZINT;
14728e3e3a7aSWarner Losh   for (i = limit - 1; i >= 0; i--) {
14738e3e3a7aSWarner Losh     res <<= NB;
14748e3e3a7aSWarner Losh     res |= (lua_Unsigned)(unsigned char)str[islittle ? i : size - 1 - i];
14758e3e3a7aSWarner Losh   }
14768e3e3a7aSWarner Losh   if (size < SZINT) {  /* real size smaller than lua_Integer? */
14778e3e3a7aSWarner Losh     if (issigned) {  /* needs sign extension? */
14788e3e3a7aSWarner Losh       lua_Unsigned mask = (lua_Unsigned)1 << (size*NB - 1);
14798e3e3a7aSWarner Losh       res = ((res ^ mask) - mask);  /* do sign extension */
14808e3e3a7aSWarner Losh     }
14818e3e3a7aSWarner Losh   }
14828e3e3a7aSWarner Losh   else if (size > SZINT) {  /* must check unread bytes */
14838e3e3a7aSWarner Losh     int mask = (!issigned || (lua_Integer)res >= 0) ? 0 : MC;
14848e3e3a7aSWarner Losh     for (i = limit; i < size; i++) {
14858e3e3a7aSWarner Losh       if ((unsigned char)str[islittle ? i : size - 1 - i] != mask)
14868e3e3a7aSWarner Losh         luaL_error(L, "%d-byte integer does not fit into Lua Integer", size);
14878e3e3a7aSWarner Losh     }
14888e3e3a7aSWarner Losh   }
14898e3e3a7aSWarner Losh   return (lua_Integer)res;
14908e3e3a7aSWarner Losh }
14918e3e3a7aSWarner Losh 
14928e3e3a7aSWarner Losh 
14938e3e3a7aSWarner Losh static int str_unpack (lua_State *L) {
14948e3e3a7aSWarner Losh   Header h;
14958e3e3a7aSWarner Losh   const char *fmt = luaL_checkstring(L, 1);
14968e3e3a7aSWarner Losh   size_t ld;
14978e3e3a7aSWarner Losh   const char *data = luaL_checklstring(L, 2, &ld);
14988e3e3a7aSWarner Losh   size_t pos = (size_t)posrelat(luaL_optinteger(L, 3, 1), ld) - 1;
14998e3e3a7aSWarner Losh   int n = 0;  /* number of results */
15008e3e3a7aSWarner Losh   luaL_argcheck(L, pos <= ld, 3, "initial position out of string");
15018e3e3a7aSWarner Losh   initheader(L, &h);
15028e3e3a7aSWarner Losh   while (*fmt != '\0') {
15038e3e3a7aSWarner Losh     int size, ntoalign;
15048e3e3a7aSWarner Losh     KOption opt = getdetails(&h, pos, &fmt, &size, &ntoalign);
15058e3e3a7aSWarner Losh     if ((size_t)ntoalign + size > ~pos || pos + ntoalign + size > ld)
15068e3e3a7aSWarner Losh       luaL_argerror(L, 2, "data string too short");
15078e3e3a7aSWarner Losh     pos += ntoalign;  /* skip alignment */
15088e3e3a7aSWarner Losh     /* stack space for item + next position */
15098e3e3a7aSWarner Losh     luaL_checkstack(L, 2, "too many results");
15108e3e3a7aSWarner Losh     n++;
15118e3e3a7aSWarner Losh     switch (opt) {
15128e3e3a7aSWarner Losh       case Kint:
15138e3e3a7aSWarner Losh       case Kuint: {
15148e3e3a7aSWarner Losh         lua_Integer res = unpackint(L, data + pos, h.islittle, size,
15158e3e3a7aSWarner Losh                                        (opt == Kint));
15168e3e3a7aSWarner Losh         lua_pushinteger(L, res);
15178e3e3a7aSWarner Losh         break;
15188e3e3a7aSWarner Losh       }
15198e3e3a7aSWarner Losh       case Kfloat: {
15208e3e3a7aSWarner Losh         volatile Ftypes u;
15218e3e3a7aSWarner Losh         lua_Number num;
15228e3e3a7aSWarner Losh         copywithendian(u.buff, data + pos, size, h.islittle);
152307c17b2bSWarner Losh #if LUA_FLOAT_TYPE != LUA_FLOAT_INT64
15248e3e3a7aSWarner Losh         if (size == sizeof(u.f)) num = (lua_Number)u.f;
15258e3e3a7aSWarner Losh         else if (size == sizeof(u.d)) num = (lua_Number)u.d;
15268e3e3a7aSWarner Losh         else num = u.n;
152707c17b2bSWarner Losh #else
152807c17b2bSWarner Losh 	num = u.n;
152907c17b2bSWarner Losh #endif
15308e3e3a7aSWarner Losh         lua_pushnumber(L, num);
15318e3e3a7aSWarner Losh         break;
15328e3e3a7aSWarner Losh       }
15338e3e3a7aSWarner Losh       case Kchar: {
15348e3e3a7aSWarner Losh         lua_pushlstring(L, data + pos, size);
15358e3e3a7aSWarner Losh         break;
15368e3e3a7aSWarner Losh       }
15378e3e3a7aSWarner Losh       case Kstring: {
15388e3e3a7aSWarner Losh         size_t len = (size_t)unpackint(L, data + pos, h.islittle, size, 0);
15398e3e3a7aSWarner Losh         luaL_argcheck(L, pos + len + size <= ld, 2, "data string too short");
15408e3e3a7aSWarner Losh         lua_pushlstring(L, data + pos + size, len);
15418e3e3a7aSWarner Losh         pos += len;  /* skip string */
15428e3e3a7aSWarner Losh         break;
15438e3e3a7aSWarner Losh       }
15448e3e3a7aSWarner Losh       case Kzstr: {
15458e3e3a7aSWarner Losh         size_t len = (int)strlen(data + pos);
15468e3e3a7aSWarner Losh         lua_pushlstring(L, data + pos, len);
15478e3e3a7aSWarner Losh         pos += len + 1;  /* skip string plus final '\0' */
15488e3e3a7aSWarner Losh         break;
15498e3e3a7aSWarner Losh       }
15508e3e3a7aSWarner Losh       case Kpaddalign: case Kpadding: case Knop:
15518e3e3a7aSWarner Losh         n--;  /* undo increment */
15528e3e3a7aSWarner Losh         break;
15538e3e3a7aSWarner Losh     }
15548e3e3a7aSWarner Losh     pos += size;
15558e3e3a7aSWarner Losh   }
15568e3e3a7aSWarner Losh   lua_pushinteger(L, pos + 1);  /* next position */
15578e3e3a7aSWarner Losh   return n + 1;
15588e3e3a7aSWarner Losh }
15598e3e3a7aSWarner Losh 
15608e3e3a7aSWarner Losh /* }====================================================== */
15618e3e3a7aSWarner Losh 
15628e3e3a7aSWarner Losh 
15638e3e3a7aSWarner Losh static const luaL_Reg strlib[] = {
15648e3e3a7aSWarner Losh   {"byte", str_byte},
15658e3e3a7aSWarner Losh   {"char", str_char},
15668e3e3a7aSWarner Losh   {"dump", str_dump},
15678e3e3a7aSWarner Losh   {"find", str_find},
15688e3e3a7aSWarner Losh   {"format", str_format},
15698e3e3a7aSWarner Losh   {"gmatch", gmatch},
15708e3e3a7aSWarner Losh   {"gsub", str_gsub},
15718e3e3a7aSWarner Losh   {"len", str_len},
15728e3e3a7aSWarner Losh   {"lower", str_lower},
15738e3e3a7aSWarner Losh   {"match", str_match},
15748e3e3a7aSWarner Losh   {"rep", str_rep},
15758e3e3a7aSWarner Losh   {"reverse", str_reverse},
15768e3e3a7aSWarner Losh   {"sub", str_sub},
15778e3e3a7aSWarner Losh   {"upper", str_upper},
15788e3e3a7aSWarner Losh   {"pack", str_pack},
15798e3e3a7aSWarner Losh   {"packsize", str_packsize},
15808e3e3a7aSWarner Losh   {"unpack", str_unpack},
15818e3e3a7aSWarner Losh   {NULL, NULL}
15828e3e3a7aSWarner Losh };
15838e3e3a7aSWarner Losh 
15848e3e3a7aSWarner Losh 
15858e3e3a7aSWarner Losh static void createmetatable (lua_State *L) {
15868e3e3a7aSWarner Losh   lua_createtable(L, 0, 1);  /* table to be metatable for strings */
15878e3e3a7aSWarner Losh   lua_pushliteral(L, "");  /* dummy string */
15888e3e3a7aSWarner Losh   lua_pushvalue(L, -2);  /* copy table */
15898e3e3a7aSWarner Losh   lua_setmetatable(L, -2);  /* set table as metatable for strings */
15908e3e3a7aSWarner Losh   lua_pop(L, 1);  /* pop dummy string */
15918e3e3a7aSWarner Losh   lua_pushvalue(L, -2);  /* get string library */
15928e3e3a7aSWarner Losh   lua_setfield(L, -2, "__index");  /* metatable.__index = string */
15938e3e3a7aSWarner Losh   lua_pop(L, 1);  /* pop metatable */
15948e3e3a7aSWarner Losh }
15958e3e3a7aSWarner Losh 
15968e3e3a7aSWarner Losh 
15978e3e3a7aSWarner Losh /*
15988e3e3a7aSWarner Losh ** Open string library
15998e3e3a7aSWarner Losh */
16008e3e3a7aSWarner Losh LUAMOD_API int luaopen_string (lua_State *L) {
16018e3e3a7aSWarner Losh   luaL_newlib(L, strlib);
16028e3e3a7aSWarner Losh   createmetatable(L);
16038e3e3a7aSWarner Losh   return 1;
16048e3e3a7aSWarner Losh }
16058e3e3a7aSWarner Losh 
1606