1 /* $NetBSD: lstring.c,v 1.9 2018/08/04 17:30:01 alnsn Exp $ */ 2 3 /* 4 ** Id: lstring.c,v 2.56.1.1 2017/04/19 17:20:42 roberto Exp 5 ** String table (keeps all strings handled by Lua) 6 ** See Copyright Notice in lua.h 7 */ 8 9 #define lstring_c 10 #define LUA_CORE 11 12 #include "lprefix.h" 13 14 15 #ifndef _KERNEL 16 #include <string.h> 17 #endif /* _KERNEL */ 18 19 #include "lua.h" 20 21 #include "ldebug.h" 22 #include "ldo.h" 23 #include "lmem.h" 24 #include "lobject.h" 25 #include "lstate.h" 26 #include "lstring.h" 27 28 29 #define MEMERRMSG "not enough memory" 30 31 32 /* 33 ** Lua will use at most ~(2^LUAI_HASHLIMIT) bytes from a string to 34 ** compute its hash 35 */ 36 #if !defined(LUAI_HASHLIMIT) 37 #define LUAI_HASHLIMIT 5 38 #endif 39 40 41 /* 42 ** equality for long strings 43 */ 44 int luaS_eqlngstr (TString *a, TString *b) { 45 size_t len = a->u.lnglen; 46 lua_assert(a->tt == LUA_TLNGSTR && b->tt == LUA_TLNGSTR); 47 return (a == b) || /* same instance or... */ 48 ((len == b->u.lnglen) && /* equal length and ... */ 49 (memcmp(getstr(a), getstr(b), len) == 0)); /* equal contents */ 50 } 51 52 53 unsigned int luaS_hash (const char *str, size_t l, unsigned int seed) { 54 unsigned int h = seed ^ cast(unsigned int, l); 55 size_t step = (l >> LUAI_HASHLIMIT) + 1; 56 for (; l >= step; l -= step) 57 h ^= ((h<<5) + (h>>2) + cast_byte(str[l - 1])); 58 return h; 59 } 60 61 62 unsigned int luaS_hashlongstr (TString *ts) { 63 lua_assert(ts->tt == LUA_TLNGSTR); 64 if (ts->extra == 0) { /* no hash? */ 65 ts->hash = luaS_hash(getstr(ts), ts->u.lnglen, ts->hash); 66 ts->extra = 1; /* now it has its hash */ 67 } 68 return ts->hash; 69 } 70 71 72 /* 73 ** resizes the string table 74 */ 75 void luaS_resize (lua_State *L, int newsize) { 76 int i; 77 stringtable *tb = &G(L)->strt; 78 if (newsize > tb->size) { /* grow table if needed */ 79 luaM_reallocvector(L, tb->hash, tb->size, newsize, TString *); 80 for (i = tb->size; i < newsize; i++) 81 tb->hash[i] = NULL; 82 } 83 for (i = 0; i < tb->size; i++) { /* rehash */ 84 TString *p = tb->hash[i]; 85 tb->hash[i] = NULL; 86 while (p) { /* for each node in the list */ 87 TString *hnext = p->u.hnext; /* save next */ 88 unsigned int h = lmod(p->hash, newsize); /* new position */ 89 p->u.hnext = tb->hash[h]; /* chain it */ 90 tb->hash[h] = p; 91 p = hnext; 92 } 93 } 94 if (newsize < tb->size) { /* shrink table if needed */ 95 /* vanishing slice should be empty */ 96 lua_assert(tb->hash[newsize] == NULL && tb->hash[tb->size - 1] == NULL); 97 luaM_reallocvector(L, tb->hash, tb->size, newsize, TString *); 98 } 99 tb->size = newsize; 100 } 101 102 103 /* 104 ** Clear API string cache. (Entries cannot be empty, so fill them with 105 ** a non-collectable string.) 106 */ 107 void luaS_clearcache (global_State *g) { 108 int i, j; 109 for (i = 0; i < STRCACHE_N; i++) 110 for (j = 0; j < STRCACHE_M; j++) { 111 if (iswhite(g->strcache[i][j])) /* will entry be collected? */ 112 g->strcache[i][j] = g->memerrmsg; /* replace it with something fixed */ 113 } 114 } 115 116 117 /* 118 ** Initialize the string table and the string cache 119 */ 120 void luaS_init (lua_State *L) { 121 global_State *g = G(L); 122 int i, j; 123 luaS_resize(L, MINSTRTABSIZE); /* initial size of string table */ 124 /* pre-create memory-error message */ 125 g->memerrmsg = luaS_newliteral(L, MEMERRMSG); 126 luaC_fix(L, obj2gco(g->memerrmsg)); /* it should never be collected */ 127 for (i = 0; i < STRCACHE_N; i++) /* fill cache with valid strings */ 128 for (j = 0; j < STRCACHE_M; j++) 129 g->strcache[i][j] = g->memerrmsg; 130 } 131 132 133 134 /* 135 ** creates a new string object 136 */ 137 static TString *createstrobj (lua_State *L, size_t l, int tag, unsigned int h) { 138 TString *ts; 139 GCObject *o; 140 size_t totalsize; /* total size of TString object */ 141 totalsize = sizelstring(l); 142 o = luaC_newobj(L, tag, totalsize); 143 ts = gco2ts(o); 144 ts->hash = h; 145 ts->extra = 0; 146 getstr(ts)[l] = '\0'; /* ending 0 */ 147 return ts; 148 } 149 150 151 TString *luaS_createlngstrobj (lua_State *L, size_t l) { 152 TString *ts = createstrobj(L, l, LUA_TLNGSTR, G(L)->seed); 153 ts->u.lnglen = l; 154 return ts; 155 } 156 157 158 void luaS_remove (lua_State *L, TString *ts) { 159 stringtable *tb = &G(L)->strt; 160 TString **p = &tb->hash[lmod(ts->hash, tb->size)]; 161 while (*p != ts) /* find previous element */ 162 p = &(*p)->u.hnext; 163 *p = (*p)->u.hnext; /* remove element from its list */ 164 tb->nuse--; 165 } 166 167 168 /* 169 ** checks whether short string exists and reuses it or creates a new one 170 */ 171 static TString *internshrstr (lua_State *L, const char *str, size_t l) { 172 TString *ts; 173 global_State *g = G(L); 174 unsigned int h = luaS_hash(str, l, g->seed); 175 TString **list = &g->strt.hash[lmod(h, g->strt.size)]; 176 lua_assert(str != NULL); /* otherwise 'memcmp'/'memcpy' are undefined */ 177 for (ts = *list; ts != NULL; ts = ts->u.hnext) { 178 if (l == ts->shrlen && 179 (memcmp(str, getstr(ts), l * sizeof(char)) == 0)) { 180 /* found! */ 181 if (isdead(g, ts)) /* dead (but not collected yet)? */ 182 changewhite(ts); /* resurrect it */ 183 return ts; 184 } 185 } 186 if (g->strt.nuse >= g->strt.size && g->strt.size <= MAX_INT/2) { 187 luaS_resize(L, g->strt.size * 2); 188 list = &g->strt.hash[lmod(h, g->strt.size)]; /* recompute with new size */ 189 } 190 ts = createstrobj(L, l, LUA_TSHRSTR, h); 191 memcpy(getstr(ts), str, l * sizeof(char)); 192 ts->shrlen = cast_byte(l); 193 ts->u.hnext = *list; 194 *list = ts; 195 g->strt.nuse++; 196 return ts; 197 } 198 199 200 /* 201 ** new string (with explicit length) 202 */ 203 TString *luaS_newlstr (lua_State *L, const char *str, size_t l) { 204 if (l <= LUAI_MAXSHORTLEN) /* short string? */ 205 return internshrstr(L, str, l); 206 else { 207 TString *ts; 208 if (l >= (MAX_SIZE - sizeof(TString))/sizeof(char)) 209 luaM_toobig(L); 210 ts = luaS_createlngstrobj(L, l); 211 memcpy(getstr(ts), str, l * sizeof(char)); 212 return ts; 213 } 214 } 215 216 217 /* 218 ** Create or reuse a zero-terminated string, first checking in the 219 ** cache (using the string address as a key). The cache can contain 220 ** only zero-terminated strings, so it is safe to use 'strcmp' to 221 ** check hits. 222 */ 223 TString *luaS_new (lua_State *L, const char *str) { 224 unsigned int i = point2uint(str) % STRCACHE_N; /* hash */ 225 int j; 226 TString **p = G(L)->strcache[i]; 227 for (j = 0; j < STRCACHE_M; j++) { 228 if (strcmp(str, getstr(p[j])) == 0) /* hit? */ 229 return p[j]; /* that is it */ 230 } 231 /* normal route */ 232 for (j = STRCACHE_M - 1; j > 0; j--) 233 p[j] = p[j - 1]; /* move out last element */ 234 /* new element is first in the list */ 235 p[0] = luaS_newlstr(L, str, strlen(str)); 236 return p[0]; 237 } 238 239 240 Udata *luaS_newudata (lua_State *L, size_t s) { 241 Udata *u; 242 GCObject *o; 243 if (s > MAX_SIZE - sizeof(Udata)) 244 luaM_toobig(L); 245 o = luaC_newobj(L, LUA_TUSERDATA, sizeludata(s)); 246 u = gco2u(o); 247 u->len = s; 248 u->metatable = NULL; 249 setuservalue(L, u, luaO_nilobject); 250 return u; 251 } 252 253