| 1 | /* |
| 2 | ** $Id: lstring.c $ |
| 3 | ** String table (keeps all strings handled by Lua) |
| 4 | ** See Copyright Notice in lua.h |
| 5 | */ |
| 6 | |
| 7 | #define lstring_c |
| 8 | #define LUA_CORE |
| 9 | |
| 10 | #include "lprefix.h" |
| 11 | |
| 12 | |
| 13 | #include <string.h> |
| 14 | |
| 15 | #include "lua.h" |
| 16 | |
| 17 | #include "ldebug.h" |
| 18 | #include "ldo.h" |
| 19 | #include "lmem.h" |
| 20 | #include "lobject.h" |
| 21 | #include "lstate.h" |
| 22 | #include "lstring.h" |
| 23 | |
| 24 | |
| 25 | /* |
| 26 | ** Maximum size for string table. |
| 27 | */ |
| 28 | #define MAXSTRTB cast_int(luaM_limitN(MAX_INT, TString*)) |
| 29 | |
| 30 | |
| 31 | /* |
| 32 | ** equality for long strings |
| 33 | */ |
| 34 | int luaS_eqlngstr (TString *a, TString *b) { |
| 35 | size_t len = a->u.lnglen; |
| 36 | lua_assert(a->tt == LUA_VLNGSTR && b->tt == LUA_VLNGSTR); |
| 37 | return (a == b) || /* same instance or... */ |
| 38 | ((len == b->u.lnglen) && /* equal length and ... */ |
| 39 | (memcmp(getstr(a), getstr(b), len) == 0)); /* equal contents */ |
| 40 | } |
| 41 | |
| 42 | |
| 43 | unsigned int luaS_hash (const char *str, size_t l, unsigned int seed) { |
| 44 | unsigned int h = seed ^ cast_uint(l); |
| 45 | for (; l > 0; l--) |
| 46 | h ^= ((h<<5) + (h>>2) + cast_byte(str[l - 1])); |
| 47 | return h; |
| 48 | } |
| 49 | |
| 50 | |
| 51 | unsigned int luaS_hashlongstr (TString *ts) { |
| 52 | lua_assert(ts->tt == LUA_VLNGSTR); |
| 53 | if (ts->extra == 0) { /* no hash? */ |
| 54 | size_t len = ts->u.lnglen; |
| 55 | ts->hash = luaS_hash(getstr(ts), len, ts->hash); |
| 56 | ts->extra = 1; /* now it has its hash */ |
| 57 | } |
| 58 | return ts->hash; |
| 59 | } |
| 60 | |
| 61 | |
| 62 | static void tablerehash (TString **vect, int osize, int nsize) { |
| 63 | int i; |
| 64 | for (i = osize; i < nsize; i++) /* clear new elements */ |
| 65 | vect[i] = NULL; |
| 66 | for (i = 0; i < osize; i++) { /* rehash old part of the array */ |
| 67 | TString *p = vect[i]; |
| 68 | vect[i] = NULL; |
| 69 | while (p) { /* for each string in the list */ |
| 70 | TString *hnext = p->u.hnext; /* save next */ |
| 71 | unsigned int h = lmod(p->hash, nsize); /* new position */ |
| 72 | p->u.hnext = vect[h]; /* chain it into array */ |
| 73 | vect[h] = p; |
| 74 | p = hnext; |
| 75 | } |
| 76 | } |
| 77 | } |
| 78 | |
| 79 | |
| 80 | /* |
| 81 | ** Resize the string table. If allocation fails, keep the current size. |
| 82 | ** (This can degrade performance, but any non-zero size should work |
| 83 | ** correctly.) |
| 84 | */ |
| 85 | void luaS_resize (lua_State *L, int nsize) { |
| 86 | stringtable *tb = &G(L)->strt; |
| 87 | int osize = tb->size; |
| 88 | TString **newvect; |
| 89 | if (nsize < osize) /* shrinking table? */ |
| 90 | tablerehash(tb->hash, osize, nsize); /* depopulate shrinking part */ |
| 91 | newvect = luaM_reallocvector(L, tb->hash, osize, nsize, TString*); |
| 92 | if (l_unlikely(newvect == NULL)) { /* reallocation failed? */ |
| 93 | if (nsize < osize) /* was it shrinking table? */ |
| 94 | tablerehash(tb->hash, nsize, osize); /* restore to original size */ |
| 95 | /* leave table as it was */ |
| 96 | } |
| 97 | else { /* allocation succeeded */ |
| 98 | tb->hash = newvect; |
| 99 | tb->size = nsize; |
| 100 | if (nsize > osize) |
| 101 | tablerehash(newvect, osize, nsize); /* rehash for new size */ |
| 102 | } |
| 103 | } |
| 104 | |
| 105 | |
| 106 | /* |
| 107 | ** Clear API string cache. (Entries cannot be empty, so fill them with |
| 108 | ** a non-collectable string.) |
| 109 | */ |
| 110 | void luaS_clearcache (global_State *g) { |
| 111 | int i, j; |
| 112 | for (i = 0; i < STRCACHE_N; i++) |
| 113 | for (j = 0; j < STRCACHE_M; j++) { |
| 114 | if (iswhite(g->strcache[i][j])) /* will entry be collected? */ |
| 115 | g->strcache[i][j] = g->memerrmsg; /* replace it with something fixed */ |
| 116 | } |
| 117 | } |
| 118 | |
| 119 | |
| 120 | /* |
| 121 | ** Initialize the string table and the string cache |
| 122 | */ |
| 123 | void luaS_init (lua_State *L) { |
| 124 | global_State *g = G(L); |
| 125 | int i, j; |
| 126 | stringtable *tb = &G(L)->strt; |
| 127 | tb->hash = luaM_newvector(L, MINSTRTABSIZE, TString*); |
| 128 | tablerehash(tb->hash, 0, MINSTRTABSIZE); /* clear array */ |
| 129 | tb->size = MINSTRTABSIZE; |
| 130 | /* pre-create memory-error message */ |
| 131 | g->memerrmsg = luaS_newliteral(L, MEMERRMSG); |
| 132 | luaC_fix(L, obj2gco(g->memerrmsg)); /* it should never be collected */ |
| 133 | for (i = 0; i < STRCACHE_N; i++) /* fill cache with valid strings */ |
| 134 | for (j = 0; j < STRCACHE_M; j++) |
| 135 | g->strcache[i][j] = g->memerrmsg; |
| 136 | } |
| 137 | |
| 138 | |
| 139 | |
| 140 | /* |
| 141 | ** creates a new string object |
| 142 | */ |
| 143 | static TString *createstrobj (lua_State *L, size_t l, int tag, unsigned int h) { |
| 144 | TString *ts; |
| 145 | GCObject *o; |
| 146 | size_t totalsize; /* total size of TString object */ |
| 147 | totalsize = sizelstring(l); |
| 148 | o = luaC_newobj(L, tag, totalsize); |
| 149 | ts = gco2ts(o); |
| 150 | ts->hash = h; |
| 151 | ts->extra = 0; |
| 152 | getstr(ts)[l] = '\0'; /* ending 0 */ |
| 153 | return ts; |
| 154 | } |
| 155 | |
| 156 | |
| 157 | TString *luaS_createlngstrobj (lua_State *L, size_t l) { |
| 158 | TString *ts = createstrobj(L, l, LUA_VLNGSTR, G(L)->seed); |
| 159 | ts->u.lnglen = l; |
| 160 | return ts; |
| 161 | } |
| 162 | |
| 163 | |
| 164 | void luaS_remove (lua_State *L, TString *ts) { |
| 165 | stringtable *tb = &G(L)->strt; |
| 166 | TString **p = &tb->hash[lmod(ts->hash, tb->size)]; |
| 167 | while (*p != ts) /* find previous element */ |
| 168 | p = &(*p)->u.hnext; |
| 169 | *p = (*p)->u.hnext; /* remove element from its list */ |
| 170 | tb->nuse--; |
| 171 | } |
| 172 | |
| 173 | |
| 174 | static void growstrtab (lua_State *L, stringtable *tb) { |
| 175 | if (l_unlikely(tb->nuse == MAX_INT)) { /* too many strings? */ |
| 176 | luaC_fullgc(L, 1); /* try to free some... */ |
| 177 | if (tb->nuse == MAX_INT) /* still too many? */ |
| 178 | luaM_error(L); /* cannot even create a message... */ |
| 179 | } |
| 180 | if (tb->size <= MAXSTRTB / 2) /* can grow string table? */ |
| 181 | luaS_resize(L, tb->size * 2); |
| 182 | } |
| 183 | |
| 184 | |
| 185 | /* |
| 186 | ** Checks whether short string exists and reuses it or creates a new one. |
| 187 | */ |
| 188 | static TString *internshrstr (lua_State *L, const char *str, size_t l) { |
| 189 | TString *ts; |
| 190 | global_State *g = G(L); |
| 191 | stringtable *tb = &g->strt; |
| 192 | unsigned int h = luaS_hash(str, l, g->seed); |
| 193 | TString **list = &tb->hash[lmod(h, tb->size)]; |
| 194 | lua_assert(str != NULL); /* otherwise 'memcmp'/'memcpy' are undefined */ |
| 195 | for (ts = *list; ts != NULL; ts = ts->u.hnext) { |
| 196 | if (l == ts->shrlen && (memcmp(str, getstr(ts), l * sizeof(char)) == 0)) { |
| 197 | /* found! */ |
| 198 | if (isdead(g, ts)) /* dead (but not collected yet)? */ |
| 199 | changewhite(ts); /* resurrect it */ |
| 200 | return ts; |
| 201 | } |
| 202 | } |
| 203 | /* else must create a new string */ |
| 204 | if (tb->nuse >= tb->size) { /* need to grow string table? */ |
| 205 | growstrtab(L, tb); |
| 206 | list = &tb->hash[lmod(h, tb->size)]; /* rehash with new size */ |
| 207 | } |
| 208 | ts = createstrobj(L, l, LUA_VSHRSTR, h); |
| 209 | memcpy(getstr(ts), str, l * sizeof(char)); |
| 210 | ts->shrlen = cast_byte(l); |
| 211 | ts->u.hnext = *list; |
| 212 | *list = ts; |
| 213 | tb->nuse++; |
| 214 | return ts; |
| 215 | } |
| 216 | |
| 217 | |
| 218 | /* |
| 219 | ** new string (with explicit length) |
| 220 | */ |
| 221 | TString *luaS_newlstr (lua_State *L, const char *str, size_t l) { |
| 222 | if (l <= LUAI_MAXSHORTLEN) /* short string? */ |
| 223 | return internshrstr(L, str, l); |
| 224 | else { |
| 225 | TString *ts; |
| 226 | if (l_unlikely(l >= (MAX_SIZE - sizeof(TString))/sizeof(char))) |
| 227 | luaM_toobig(L); |
| 228 | ts = luaS_createlngstrobj(L, l); |
| 229 | memcpy(getstr(ts), str, l * sizeof(char)); |
| 230 | return ts; |
| 231 | } |
| 232 | } |
| 233 | |
| 234 | |
| 235 | /* |
| 236 | ** Create or reuse a zero-terminated string, first checking in the |
| 237 | ** cache (using the string address as a key). The cache can contain |
| 238 | ** only zero-terminated strings, so it is safe to use 'strcmp' to |
| 239 | ** check hits. |
| 240 | */ |
| 241 | TString *luaS_new (lua_State *L, const char *str) { |
| 242 | unsigned int i = point2uint(str) % STRCACHE_N; /* hash */ |
| 243 | int j; |
| 244 | TString **p = G(L)->strcache[i]; |
| 245 | for (j = 0; j < STRCACHE_M; j++) { |
| 246 | if (strcmp(str, getstr(p[j])) == 0) /* hit? */ |
| 247 | return p[j]; /* that is it */ |
| 248 | } |
| 249 | /* normal route */ |
| 250 | for (j = STRCACHE_M - 1; j > 0; j--) |
| 251 | p[j] = p[j - 1]; /* move out last element */ |
| 252 | /* new element is first in the list */ |
| 253 | p[0] = luaS_newlstr(L, str, strlen(str)); |
| 254 | return p[0]; |
| 255 | } |
| 256 | |
| 257 | |
| 258 | Udata *luaS_newudata (lua_State *L, size_t s, int nuvalue) { |
| 259 | Udata *u; |
| 260 | int i; |
| 261 | GCObject *o; |
| 262 | if (l_unlikely(s > MAX_SIZE - udatamemoffset(nuvalue))) |
| 263 | luaM_toobig(L); |
| 264 | o = luaC_newobj(L, LUA_VUSERDATA, sizeudata(nuvalue, s)); |
| 265 | u = gco2u(o); |
| 266 | u->len = s; |
| 267 | u->nuvalue = nuvalue; |
| 268 | u->metatable = NULL; |
| 269 | for (i = 0; i < nuvalue; i++) |
| 270 | setnilvalue(&u->uv[i].uv); |
| 271 | return u; |
| 272 | } |
| 273 | |
| 274 | |