lua

A copy of the Lua development repository
Log | Files | Refs | README

lstring.c (10066B)


      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(INT_MAX, TString*))
     29 
     30 /*
     31 ** Initial size for the string table (must be power of 2).
     32 ** The Lua core alone registers ~50 strings (reserved words +
     33 ** metaevent keys + a few others). Libraries would typically add
     34 ** a few dozens more.
     35 */
     36 #if !defined(MINSTRTABSIZE)
     37 #define MINSTRTABSIZE   128
     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_VLNGSTR && b->tt == LUA_VLNGSTR);
     47   return (a == b) ||  /* same instance or... */
     48     ((len == b->u.lnglen) &&  /* equal length and ... */
     49      (memcmp(getlngstr(a), getlngstr(b), len) == 0));  /* equal contents */
     50 }
     51 
     52 
     53 unsigned luaS_hash (const char *str, size_t l, unsigned seed) {
     54   unsigned int h = seed ^ cast_uint(l);
     55   for (; l > 0; l--)
     56     h ^= ((h<<5) + (h>>2) + cast_byte(str[l - 1]));
     57   return h;
     58 }
     59 
     60 
     61 unsigned luaS_hashlongstr (TString *ts) {
     62   lua_assert(ts->tt == LUA_VLNGSTR);
     63   if (ts->extra == 0) {  /* no hash? */
     64     size_t len = ts->u.lnglen;
     65     ts->hash = luaS_hash(getlngstr(ts), len, ts->hash);
     66     ts->extra = 1;  /* now it has its hash */
     67   }
     68   return ts->hash;
     69 }
     70 
     71 
     72 static void tablerehash (TString **vect, int osize, int nsize) {
     73   int i;
     74   for (i = osize; i < nsize; i++)  /* clear new elements */
     75     vect[i] = NULL;
     76   for (i = 0; i < osize; i++) {  /* rehash old part of the array */
     77     TString *p = vect[i];
     78     vect[i] = NULL;
     79     while (p) {  /* for each string in the list */
     80       TString *hnext = p->u.hnext;  /* save next */
     81       unsigned int h = lmod(p->hash, nsize);  /* new position */
     82       p->u.hnext = vect[h];  /* chain it into array */
     83       vect[h] = p;
     84       p = hnext;
     85     }
     86   }
     87 }
     88 
     89 
     90 /*
     91 ** Resize the string table. If allocation fails, keep the current size.
     92 ** (This can degrade performance, but any non-zero size should work
     93 ** correctly.)
     94 */
     95 void luaS_resize (lua_State *L, int nsize) {
     96   stringtable *tb = &G(L)->strt;
     97   int osize = tb->size;
     98   TString **newvect;
     99   if (nsize < osize)  /* shrinking table? */
    100     tablerehash(tb->hash, osize, nsize);  /* depopulate shrinking part */
    101   newvect = luaM_reallocvector(L, tb->hash, osize, nsize, TString*);
    102   if (l_unlikely(newvect == NULL)) {  /* reallocation failed? */
    103     if (nsize < osize)  /* was it shrinking table? */
    104       tablerehash(tb->hash, nsize, osize);  /* restore to original size */
    105     /* leave table as it was */
    106   }
    107   else {  /* allocation succeeded */
    108     tb->hash = newvect;
    109     tb->size = nsize;
    110     if (nsize > osize)
    111       tablerehash(newvect, osize, nsize);  /* rehash for new size */
    112   }
    113 }
    114 
    115 
    116 /*
    117 ** Clear API string cache. (Entries cannot be empty, so fill them with
    118 ** a non-collectable string.)
    119 */
    120 void luaS_clearcache (global_State *g) {
    121   int i, j;
    122   for (i = 0; i < STRCACHE_N; i++)
    123     for (j = 0; j < STRCACHE_M; j++) {
    124       if (iswhite(g->strcache[i][j]))  /* will entry be collected? */
    125         g->strcache[i][j] = g->memerrmsg;  /* replace it with something fixed */
    126     }
    127 }
    128 
    129 
    130 /*
    131 ** Initialize the string table and the string cache
    132 */
    133 void luaS_init (lua_State *L) {
    134   global_State *g = G(L);
    135   int i, j;
    136   stringtable *tb = &G(L)->strt;
    137   tb->hash = luaM_newvector(L, MINSTRTABSIZE, TString*);
    138   tablerehash(tb->hash, 0, MINSTRTABSIZE);  /* clear array */
    139   tb->size = MINSTRTABSIZE;
    140   /* pre-create memory-error message */
    141   g->memerrmsg = luaS_newliteral(L, MEMERRMSG);
    142   luaC_fix(L, obj2gco(g->memerrmsg));  /* it should never be collected */
    143   for (i = 0; i < STRCACHE_N; i++)  /* fill cache with valid strings */
    144     for (j = 0; j < STRCACHE_M; j++)
    145       g->strcache[i][j] = g->memerrmsg;
    146 }
    147 
    148 
    149 size_t luaS_sizelngstr (size_t len, int kind) {
    150   switch (kind) {
    151     case LSTRREG:  /* regular long string */
    152       /* don't need 'falloc'/'ud', but need space for content */
    153       return offsetof(TString, falloc) + (len + 1) * sizeof(char);
    154     case LSTRFIX:  /* fixed external long string */
    155       /* don't need 'falloc'/'ud' */
    156       return offsetof(TString, falloc);
    157     default:  /* external long string with deallocation */
    158       lua_assert(kind == LSTRMEM);
    159       return sizeof(TString);
    160   }
    161 }
    162 
    163 
    164 /*
    165 ** creates a new string object
    166 */
    167 static TString *createstrobj (lua_State *L, size_t totalsize, lu_byte tag,
    168                               unsigned h) {
    169   TString *ts;
    170   GCObject *o;
    171   o = luaC_newobj(L, tag, totalsize);
    172   ts = gco2ts(o);
    173   ts->hash = h;
    174   ts->extra = 0;
    175   return ts;
    176 }
    177 
    178 
    179 TString *luaS_createlngstrobj (lua_State *L, size_t l) {
    180   size_t totalsize = luaS_sizelngstr(l, LSTRREG);
    181   TString *ts = createstrobj(L, totalsize, LUA_VLNGSTR, G(L)->seed);
    182   ts->u.lnglen = l;
    183   ts->shrlen = LSTRREG;  /* signals that it is a regular long string */
    184   ts->contents = cast_charp(ts) + offsetof(TString, falloc);
    185   ts->contents[l] = '\0';  /* ending 0 */
    186   return ts;
    187 }
    188 
    189 
    190 void luaS_remove (lua_State *L, TString *ts) {
    191   stringtable *tb = &G(L)->strt;
    192   TString **p = &tb->hash[lmod(ts->hash, tb->size)];
    193   while (*p != ts)  /* find previous element */
    194     p = &(*p)->u.hnext;
    195   *p = (*p)->u.hnext;  /* remove element from its list */
    196   tb->nuse--;
    197 }
    198 
    199 
    200 static void growstrtab (lua_State *L, stringtable *tb) {
    201   if (l_unlikely(tb->nuse == INT_MAX)) {  /* too many strings? */
    202     luaC_fullgc(L, 1);  /* try to free some... */
    203     if (tb->nuse == INT_MAX)  /* still too many? */
    204       luaM_error(L);  /* cannot even create a message... */
    205   }
    206   if (tb->size <= MAXSTRTB / 2)  /* can grow string table? */
    207     luaS_resize(L, tb->size * 2);
    208 }
    209 
    210 
    211 /*
    212 ** Checks whether short string exists and reuses it or creates a new one.
    213 */
    214 static TString *internshrstr (lua_State *L, const char *str, size_t l) {
    215   TString *ts;
    216   global_State *g = G(L);
    217   stringtable *tb = &g->strt;
    218   unsigned int h = luaS_hash(str, l, g->seed);
    219   TString **list = &tb->hash[lmod(h, tb->size)];
    220   lua_assert(str != NULL);  /* otherwise 'memcmp'/'memcpy' are undefined */
    221   for (ts = *list; ts != NULL; ts = ts->u.hnext) {
    222     if (l == cast_uint(ts->shrlen) &&
    223         (memcmp(str, getshrstr(ts), l * sizeof(char)) == 0)) {
    224       /* found! */
    225       if (isdead(g, ts))  /* dead (but not collected yet)? */
    226         changewhite(ts);  /* resurrect it */
    227       return ts;
    228     }
    229   }
    230   /* else must create a new string */
    231   if (tb->nuse >= tb->size) {  /* need to grow string table? */
    232     growstrtab(L, tb);
    233     list = &tb->hash[lmod(h, tb->size)];  /* rehash with new size */
    234   }
    235   ts = createstrobj(L, sizestrshr(l), LUA_VSHRSTR, h);
    236   ts->shrlen = cast(ls_byte, l);
    237   getshrstr(ts)[l] = '\0';  /* ending 0 */
    238   memcpy(getshrstr(ts), str, l * sizeof(char));
    239   ts->u.hnext = *list;
    240   *list = ts;
    241   tb->nuse++;
    242   return ts;
    243 }
    244 
    245 
    246 /*
    247 ** new string (with explicit length)
    248 */
    249 TString *luaS_newlstr (lua_State *L, const char *str, size_t l) {
    250   if (l <= LUAI_MAXSHORTLEN)  /* short string? */
    251     return internshrstr(L, str, l);
    252   else {
    253     TString *ts;
    254     if (l_unlikely(l * sizeof(char) >= (MAX_SIZE - sizeof(TString))))
    255       luaM_toobig(L);
    256     ts = luaS_createlngstrobj(L, l);
    257     memcpy(getlngstr(ts), str, l * sizeof(char));
    258     return ts;
    259   }
    260 }
    261 
    262 
    263 /*
    264 ** Create or reuse a zero-terminated string, first checking in the
    265 ** cache (using the string address as a key). The cache can contain
    266 ** only zero-terminated strings, so it is safe to use 'strcmp' to
    267 ** check hits.
    268 */
    269 TString *luaS_new (lua_State *L, const char *str) {
    270   unsigned int i = point2uint(str) % STRCACHE_N;  /* hash */
    271   int j;
    272   TString **p = G(L)->strcache[i];
    273   for (j = 0; j < STRCACHE_M; j++) {
    274     if (strcmp(str, getstr(p[j])) == 0)  /* hit? */
    275       return p[j];  /* that is it */
    276   }
    277   /* normal route */
    278   for (j = STRCACHE_M - 1; j > 0; j--)
    279     p[j] = p[j - 1];  /* move out last element */
    280   /* new element is first in the list */
    281   p[0] = luaS_newlstr(L, str, strlen(str));
    282   return p[0];
    283 }
    284 
    285 
    286 Udata *luaS_newudata (lua_State *L, size_t s, unsigned short nuvalue) {
    287   Udata *u;
    288   int i;
    289   GCObject *o;
    290   if (l_unlikely(s > MAX_SIZE - udatamemoffset(nuvalue)))
    291     luaM_toobig(L);
    292   o = luaC_newobj(L, LUA_VUSERDATA, sizeudata(nuvalue, s));
    293   u = gco2u(o);
    294   u->len = s;
    295   u->nuvalue = nuvalue;
    296   u->metatable = NULL;
    297   for (i = 0; i < nuvalue; i++)
    298     setnilvalue(&u->uv[i].uv);
    299   return u;
    300 }
    301 
    302 
    303 struct NewExt {
    304   ls_byte kind;
    305   const char *s;
    306    size_t len;
    307   TString *ts;  /* output */
    308 };
    309 
    310 
    311 static void f_newext (lua_State *L, void *ud) {
    312   struct NewExt *ne = cast(struct NewExt *, ud);
    313   size_t size = luaS_sizelngstr(0, ne->kind);
    314   ne->ts = createstrobj(L, size, LUA_VLNGSTR, G(L)->seed);
    315 }
    316 
    317 
    318 static void f_pintern (lua_State *L, void *ud) {
    319   struct NewExt *ne = cast(struct NewExt *, ud);
    320   ne->ts = internshrstr(L, ne->s, ne->len);
    321 }
    322 
    323 
    324 TString *luaS_newextlstr (lua_State *L,
    325 	          const char *s, size_t len, lua_Alloc falloc, void *ud) {
    326   struct NewExt ne;
    327   if (len <= LUAI_MAXSHORTLEN) {  /* short string? */
    328     ne.s = s; ne.len = len;
    329     if (!falloc)
    330       f_pintern(L, &ne);  /* just internalize string */
    331     else {
    332       TStatus status = luaD_rawrunprotected(L, f_pintern, &ne);
    333       (*falloc)(ud, cast_voidp(s), len + 1, 0);  /* free external string */
    334       if (status != LUA_OK)  /* memory error? */
    335         luaM_error(L);  /* re-raise memory error */
    336     }
    337     return ne.ts;
    338   }
    339   /* "normal" case: long strings */
    340   if (!falloc) {
    341     ne.kind = LSTRFIX;
    342     f_newext(L, &ne);  /* just create header */
    343   }
    344   else {
    345     ne.kind = LSTRMEM;
    346     if (luaD_rawrunprotected(L, f_newext, &ne) != LUA_OK) {  /* mem. error? */
    347       (*falloc)(ud, cast_voidp(s), len + 1, 0);  /* free external string */
    348       luaM_error(L);  /* re-raise memory error */
    349     }
    350     ne.ts->falloc = falloc;
    351     ne.ts->ud = ud;
    352   }
    353   ne.ts->shrlen = ne.kind;
    354   ne.ts->u.lnglen = len;
    355   ne.ts->contents = cast_charp(s);
    356   return ne.ts;
    357 }
    358 
    359