change in hash algorithm so that it does not need empty slot

(tables can be 100% full)

9 files changed, 125 insertions, 108 deletions

diff --git a/lgc.c b/lgc.c
index f401505c..0d76d2b1 100644
--- a/lgc.c
+++ b/lgc.c
@@ -1,5 +1,5 @@
 /*
-** $Id: lgc.c,v 2.18 2004/12/06 17:53:42 roberto Exp roberto $
+** $Id: lgc.c,v 2.19 2004/12/13 12:15:11 roberto Exp roberto $
 ** Garbage Collector
 ** See Copyright Notice in lua.h
 */
@@ -63,7 +63,7 @@
 
 
 static void removeentry (Node *n) {
-  setnilvalue(gval(n));  /* remove corresponding value ... */
+  lua_assert(ttisnil(gval(n)));
   if (iscollectable(gkey(n)))
     setttype(gkey(n), LUA_TDEADKEY);  /* dead key; remove it */
 }
@@ -162,7 +162,6 @@ static int traversetable (global_State *g, Table *h) {
   const TValue *mode;
   if (h->metatable)
     markobject(g, h->metatable);
-  lua_assert(h->lsizenode || h->node == g->dummynode);
   mode = gfasttm(g, h->metatable, TM_MODE);
   if (mode && ttisstring(mode)) {  /* is there a weak mode? */
     weakkey = (strchr(svalue(mode), 'k') != NULL);
@@ -368,8 +367,10 @@ static void cleartable (GCObject *l) {
     while (i--) {
       Node *n = gnode(h, i);
       if (!ttisnil(gval(n)) &&  /* non-empty entry? */
-          (iscleared(key2tval(n), 1) || iscleared(gval(n), 0)))
+          (iscleared(key2tval(n), 1) || iscleared(gval(n), 0))) {
+        setnilvalue(gval(n));  /* remove value ... */
         removeentry(n);  /* remove entry from table */
+      }
     }
     l = h->gclist;
   }
diff --git a/lobject.h b/lobject.h
index 259bdac5..2c0a5d26 100644
--- a/lobject.h
+++ b/lobject.h
@@ -1,5 +1,5 @@
 /*
-** $Id: lobject.h,v 2.7 2004/11/01 15:06:50 roberto Exp roberto $
+** $Id: lobject.h,v 2.8 2004/12/04 18:10:22 roberto Exp roberto $
 ** Type definitions for Lua objects
 ** See Copyright Notice in lua.h
 */
@@ -330,7 +330,7 @@ typedef struct Table {
   struct Table *metatable;
   TValue *array;  /* array part */
   Node *node;
-  Node *firstfree;  /* this position is free; all positions after it are full */
+  Node *lastfree;  /* any free position is before this position */
   GCObject *gclist;
   int sizearray;  /* size of `array' array */
 } Table;
@@ -351,6 +351,8 @@ typedef struct Table {
 
 extern const TValue luaO_nilobject;
 
+#define ceillog2(x)     (luaO_log2((x)-1) + 1)
+
 int luaO_log2 (unsigned int x);
 int luaO_int2fb (unsigned int x);
 int luaO_fb2int (int x);
diff --git a/lparser.c b/lparser.c
index d131e5e0..cb348c40 100644
--- a/lparser.c
+++ b/lparser.c
@@ -1,5 +1,5 @@
 /*
-** $Id: lparser.c,v 2.11 2004/12/07 18:31:16 roberto Exp $
+** $Id: lparser.c,v 2.12 2005/01/04 15:55:12 roberto Exp roberto $
 ** Lua Parser
 ** See Copyright Notice in lua.h
 */
@@ -547,7 +547,7 @@ static void constructor (LexState *ls, expdesc *t) {
   check_match(ls, '}', '{', line);
   lastlistfield(fs, &cc);
   SETARG_B(fs->f->code[pc], luaO_int2fb(cc.na)); /* set initial array size */
-  SETARG_C(fs->f->code[pc], luaO_int2fb(cc.nh+1));  /* set initial table size */
+  SETARG_C(fs->f->code[pc], luaO_int2fb(cc.nh));  /* set initial table size */
 }
 
 /* }====================================================================== */
diff --git a/lstate.c b/lstate.c
index 1645eccf..54917649 100644
--- a/lstate.c
+++ b/lstate.c
@@ -1,5 +1,5 @@
 /*
-** $Id: lstate.c,v 2.19 2004/12/13 12:15:11 roberto Exp $
+** $Id: lstate.c,v 2.20 2005/01/04 15:55:12 roberto Exp roberto $
 ** Global State
 ** See Copyright Notice in lua.h
 */
@@ -189,9 +189,6 @@ LUA_API lua_State *lua_newstate (lua_Alloc f, void *ud) {
   g->grayagain = NULL;
   g->weak = NULL;
   g->tmudata = NULL;
-  setnilvalue(gkey(g->dummynode));
-  setnilvalue(gval(g->dummynode));
-  gnext(g->dummynode) = NULL;
   g->totalbytes = sizeof(LG);
   g->gcpace = GCDIV;
   g->incgc = 1;
diff --git a/lstate.h b/lstate.h
index 06a5c5dc..26a94082 100644
--- a/lstate.h
+++ b/lstate.h
@@ -1,5 +1,5 @@
 /*
-** $Id: lstate.h,v 2.9 2004/12/06 17:53:42 roberto Exp roberto $
+** $Id: lstate.h,v 2.10 2004/12/13 12:15:11 roberto Exp roberto $
 ** Global State
 ** See Copyright Notice in lua.h
 */
@@ -90,7 +90,6 @@ typedef struct global_State {
   lua_CFunction panic;  /* to be called in unprotected errors */
   TValue _registry;
   struct lua_State *mainthread;
-  Node dummynode[1];  /* common node array for all empty tables */
   TString *tmname[TM_N];  /* array with tag-method names */
 } global_State;
 
diff --git a/ltable.c b/ltable.c
index d9a1ee20..a3a169b7 100644
--- a/ltable.c
+++ b/ltable.c
@@ -1,5 +1,5 @@
 /*
-** $Id: ltable.c,v 2.12 2004/12/04 18:10:22 roberto Exp $
+** $Id: ltable.c,v 2.13 2005/01/04 15:55:12 roberto Exp roberto $
 ** Lua tables (hash)
 ** See Copyright Notice in lua.h
 */
@@ -68,6 +68,13 @@
 #define numints         cast(int, sizeof(lua_Number)/sizeof(int))
 
 
+
+const Node luaH_dummynode = {
+  {{NULL}, LUA_TNIL},  /* value */
+  {{NULL}, LUA_TNIL, NULL}  /* key */
+};
+
+
 /*
 ** hash for lua_Numbers
 */
@@ -176,31 +183,32 @@ int luaH_next (lua_State *L, Table *t, StkId key) {
 */
 
 
-static void computesizes  (int nums[], int ntotal, int *narray, int *nhash) {
+static int computesizes (int nums[], int *narray) {
   int i;
-  int a = nums[0];  /* number of elements smaller than 2^i */
-  int na = a;  /* number of elements to go to array part */
-  int n = (na == 0) ? -1 : 0;  /* (log of) optimal size for array part */
-  for (i = 1; a < *narray && *narray >= twoto(i-1); i++) {
+  int twotoi;  /* 2^i */
+  int a = 0;  /* number of elements smaller than 2^i */
+  int na = 0;  /* number of elements to go to array part */
+  int n = 0;  /* optimal size for array part */
+  for (i = 0, twotoi = 1; twotoi/2 < *narray; i++, twotoi *= 2) {
     if (nums[i] > 0) {
       a += nums[i];
-      if (a >= twoto(i-1)) {  /* more than half elements in use? */
-        n = i;
-        na = a;
+      if (a > twotoi/2) {  /* more than half elements present? */
+        n = twotoi;  /* optimal size (till now) */
+        na = a;  /* all elements smaller than n will go to array part */
       }
     }
+    if (a == *narray) break;  /* all elements already counted */
   }
-  lua_assert(na <= *narray && *narray <= ntotal);
-  *nhash = ntotal - na;
-  *narray = (n == -1) ? 0 : twoto(n);
-  lua_assert(na <= *narray && na >= *narray/2);
+  *narray = n;
+  lua_assert(*narray/2 <= na && na <= *narray);
+  return na;
 }
 
 
 static int countint (const TValue *key, int *nums) {
   int k = arrayindex(key);
   if (0 < k && k <= MAXASIZE) {  /* is `key' an appropriate array index? */
-    nums[luaO_log2(k-1)+1]++;  /* count as such */
+    nums[ceillog2(k)]++;  /* count as such */
     return 1;
   }
   else
@@ -208,40 +216,44 @@ static int countint (const TValue *key, int *nums) {
 }
 
 
-static void numuse (const Table *t, int *narray, int *nhash, const TValue *ek) {
-  int nums[MAXBITS+1];
-  int i, lg;
-  int totaluse = 0;
-  /* count elements in array part */
-  for (i=0, lg=0; lg<=MAXBITS; lg++) {  /* for each slice [2^(lg-1) to 2^lg) */
-    int ttlg = twoto(lg);  /* 2^lg */
-    if (ttlg > t->sizearray) {
-      ttlg = t->sizearray;
-      if (i >= ttlg) break;
+static int numusearray (const Table *t, int *nums) {
+  int lg;
+  int ttlg;  /* 2^lg */
+  int ause = 0;  /* summation of `nums' */
+  int i = 1;  /* count to traverse all array keys */
+  for (lg=0, ttlg=1; lg<=MAXBITS; lg++, ttlg*=2) {  /* for each slice */
+    int lc = 0;  /* counter */
+    int lim = ttlg;
+    if (lim > t->sizearray) {
+      lim = t->sizearray;  /* adjust upper limit */
+      if (i > lim)
+        break;  /* no more elements to count */
     }
-    nums[lg] = 0;
-    for (; i<ttlg; i++) {
-      if (!ttisnil(&t->array[i])) {
-        nums[lg]++;
-        totaluse++;
-      }
+    /* count elements in range (2^(lg-1), 2^lg] */
+    for (; i <= lim; i++) {
+      if (!ttisnil(&t->array[i-1]))
+        lc++;
     }
+    nums[lg] += lc;
+    ause += lc;
   }
-  for (; lg<=MAXBITS; lg++) nums[lg] = 0;  /* reset other counts */
-  *narray = totaluse;  /* all previous uses were in array part */
-  /* count elements in hash part */
-  i = sizenode(t);
+  return ause;
+}
+
+
+static int numusehash (const Table *t, int *nums, int *pnasize) {
+  int totaluse = 0;  /* total number of elements */
+  int ause = 0;  /* summation of `nums' */
+  int i = sizenode(t);
   while (i--) {
     Node *n = &t->node[i];
     if (!ttisnil(gval(n))) {
-      *narray += countint(key2tval(n), nums);
+      ause += countint(key2tval(n), nums);
       totaluse++;
     }
   }
-  /* count extra key */
-  *narray += countint(ek, nums);
-  totaluse++;
-  computesizes(nums, totaluse, narray, nhash);
+  *pnasize += ause;
+  return totaluse;
 }
 
 
@@ -254,18 +266,18 @@ static void setarrayvector (lua_State *L, Table *t, int size) {
 }
 
 
-static void setnodevector (lua_State *L, Table *t, int lsize) {
-  int i;
-  int size = twoto(lsize);
-  if (lsize > MAXBITS)
-    luaG_runerror(L, "table overflow");
-  if (lsize == 0) {  /* no elements to hash part? */
-    t->node = G(L)->dummynode;  /* use common `dummynode' */
-    lua_assert(ttisnil(gkey(t->node)));  /* assert invariants: */
-    lua_assert(ttisnil(gval(t->node)));
-    lua_assert(gnext(t->node) == NULL);  /* (`dummynode' must be empty) */
+static void setnodevector (lua_State *L, Table *t, int size) {
+  int lsize;
+  if (size == 0) {  /* no elements to hash part? */
+    t->node = cast(Node *, &luaH_dummynode);  /* use common `dummynode' */
+    lsize = 0;
   }
   else {
+    int i;
+    lsize = ceillog2(size);
+    if (lsize > MAXBITS)
+      luaG_runerror(L, "table overflow");
+    size = twoto(lsize);
     t->node = luaM_newvector(L, size, Node);
     for (i=0; i<size; i++) {
       gnext(&t->node[i]) = NULL;
@@ -274,31 +286,19 @@ static void setnodevector (lua_State *L, Table *t, int lsize) {
     }
   }
   t->lsizenode = cast(lu_byte, lsize);
-  t->firstfree = gnode(t, size-1);  /* first free position to be used */
+  t->lastfree = gnode(t, size);  /* all positions are free */
 }
 
 
-static void luaH_resize (lua_State *L, Table *t, int nasize, int nhsize) {
+static void resize (lua_State *L, Table *t, int nasize, int nhsize) {
   int i;
   int oldasize = t->sizearray;
   int oldhsize = t->lsizenode;
-  Node *nold;
-  Node temp[1];
-  if (oldhsize)
-    nold = t->node;  /* save old hash ... */
-  else {  /* old hash is `dummynode' */
-    lua_assert(t->node == G(L)->dummynode);
-    temp[0] = t->node[0];  /* copy it to `temp' */
-    nold = temp;
-    setnilvalue(gkey(G(L)->dummynode));  /* restate invariant */
-    setnilvalue(gval(G(L)->dummynode));
-    lua_assert(gnext(G(L)->dummynode) == NULL);
-  }
+  Node *nold = t->node;  /* save old hash ... */
   if (nasize > oldasize)  /* array part must grow? */
     setarrayvector(L, t, nasize);
   /* create new hash part with appropriate size */
   setnodevector(L, t, nhsize);  
-  /* re-insert elements */
   if (nasize < oldasize) {  /* array part must shrink? */
     t->sizearray = nasize;
     /* re-insert elements from vanishing slice */
@@ -309,28 +309,39 @@ static void luaH_resize (lua_State *L, Table *t, int nasize, int nhsize) {
     /* shrink array */
     luaM_reallocvector(L, t->array, oldasize, nasize, TValue);
   }
-  /* re-insert elements in hash part */
+  /* re-insert elements from hash part */
   for (i = twoto(oldhsize) - 1; i >= 0; i--) {
     Node *old = nold+i;
     if (!ttisnil(gval(old)))
       setobjt2t(L, luaH_set(L, t, key2tval(old)), gval(old));
   }
-  if (oldhsize)
+  if (nold != &luaH_dummynode)
     luaM_freearray(L, nold, twoto(oldhsize), Node);  /* free old array */
 }
 
 
 void luaH_resizearray (lua_State *L, Table *t, int nasize) {
-  luaH_resize(L, t, nasize, t->lsizenode);
+  int nsize = (t->node == &luaH_dummynode) ? 0 : sizenode(t);
+  resize(L, t, nasize, nsize);
 }
 
 
 static void rehash (lua_State *L, Table *t, const TValue *ek) {
-  int nasize, nhsize;
-  /* compute new sizes for array and hash parts */
-  numuse(t, &nasize, &nhsize, ek);
+  int nasize, na;
+  int nums[MAXBITS+1];  /* nums[i] = number of keys between 2^(i-1) and 2^i */
+  int i;
+  int totaluse;
+  for (i=0; i<=MAXBITS; i++) nums[i] = 0;  /* reset counts */
+  nasize = numusearray(t, nums);  /* count keys in array part */
+  totaluse = nasize;  /* all those keys are integer keys */
+  totaluse += numusehash(t, nums, &nasize);  /* count keys in hash part */
+  /* count extra key */
+  nasize += countint(ek, nums);
+  totaluse++;
+  /* compute new size for array part */
+  na = computesizes(nums, &nasize);
   /* resize the table to new computed sizes */
-  luaH_resize(L, t, nasize, luaO_log2(nhsize)+1);
+  resize(L, t, nasize, totaluse - na);
 }
 
 
@@ -349,21 +360,30 @@ Table *luaH_new (lua_State *L, int narray, int nhash) {
   t->array = NULL;
   t->sizearray = 0;
   t->lsizenode = 0;
-  t->node = NULL;
+  t->node = cast(Node *, &luaH_dummynode);
   setarrayvector(L, t, narray);
-  setnodevector(L, t, luaO_log2(nhash)+1);
+  setnodevector(L, t, nhash);
   return t;
 }
 
 
 void luaH_free (lua_State *L, Table *t) {
-  if (t->lsizenode)
+  if (t->node != &luaH_dummynode)
     luaM_freearray(L, t->node, sizenode(t), Node);
   luaM_freearray(L, t->array, t->sizearray, TValue);
   luaM_free(L, t);
 }
 
 
+static Node *getfreepos (lua_State *L, Table *t) {
+  while (t->lastfree-- > t->node) {
+    if (ttisnil(gkey(t->lastfree)))
+      return t->lastfree;
+  }
+  return NULL;  /* could not find a free place */
+}
+
+
 
 /*
 ** inserts a new key into a hash table; first, check whether key's main 
@@ -374,10 +394,15 @@ void luaH_free (lua_State *L, Table *t) {
 */
 static TValue *newkey (lua_State *L, Table *t, const TValue *key) {
   Node *mp = luaH_mainposition(t, key);
-  if (!ttisnil(gval(mp))) {  /* main position is not free? */
-    /* `mp' of colliding node */
-    Node *othern = luaH_mainposition(t, key2tval(mp));
-    Node *n = t->firstfree;  /* get a free place */
+  if (!ttisnil(gval(mp)) || mp == &luaH_dummynode) {
+    Node *othern;
+    Node *n = getfreepos(L, t);  /* get a free place */
+    if (n == NULL) {  /* cannot find a free place? */
+      rehash(L, t, key);  /* grow table */
+      return luaH_set(L, t, key);  /* re-insert key into grown table */
+    }
+    lua_assert(n != &luaH_dummynode);
+    othern = luaH_mainposition(t, key2tval(mp));
     if (othern != mp) {  /* is colliding node out of its main position? */
       /* yes; move colliding node into free position */
       while (gnext(othern) != mp) othern = gnext(othern);  /* find previous */
@@ -396,15 +421,7 @@ static TValue *newkey (lua_State *L, Table *t, const TValue *key) {
   gkey(mp)->value = key->value; gkey(mp)->tt = key->tt;
   luaC_barriert(L, t, key);
   lua_assert(ttisnil(gval(mp)));
-  for (;;) {  /* correct `firstfree' */
-    if (ttisnil(gkey(t->firstfree)))
-      return gval(mp);  /* OK; table still has a free place */
-    else if (t->firstfree == t->node) break;  /* cannot decrement from here */
-    else (t->firstfree)--;
-  }
-  /* no more free places; must create one */
-  rehash(L, t, key);  /* grow table */
-  return luaH_set(L, t, key);  /* re-insert in new table */
+  return gval(mp);
 }
 
 
diff --git a/ltable.h b/ltable.h
index 3927f088..d8ba8c8a 100644
--- a/ltable.h
+++ b/ltable.h
@@ -1,5 +1,5 @@
 /*
-** $Id: ltable.h,v 2.3 2004/10/06 18:34:16 roberto Exp roberto $
+** $Id: ltable.h,v 2.4 2005/01/04 15:55:12 roberto Exp roberto $
 ** Lua tables (hash)
 ** See Copyright Notice in lua.h
 */
@@ -18,6 +18,8 @@
 #define key2tval(n)     (cast(const TValue *, gkey(n)))
 
 
+extern const Node luaH_dummynode;
+
 const TValue *luaH_getnum (Table *t, int key);
 TValue *luaH_setnum (lua_State *L, Table *t, int key);
 const TValue *luaH_getstr (Table *t, TString *key);
diff --git a/ltests.c b/ltests.c
index 37720a26..6139106c 100644
--- a/ltests.c
+++ b/ltests.c
@@ -1,5 +1,5 @@
 /*
-** $Id: ltests.c,v 2.14 2004/10/06 18:34:16 roberto Exp $
+** $Id: ltests.c,v 2.15 2004/11/01 15:06:50 roberto Exp roberto $
 ** Internal Module for Debugging of the Lua Implementation
 ** See Copyright Notice in lua.h
 */
@@ -200,7 +200,6 @@ static void checktable (global_State *g, Table *h) {
   GCObject *hgc = obj2gco(h);
   if (h->metatable)
     checkobjref(g, hgc, h->metatable);
-  lua_assert(h->lsizenode || h->node == g->dummynode);
   mode = gfasttm(g, h->metatable, TM_MODE);
   if (mode && ttisstring(mode)) {  /* is there a weak mode? */
     weakkey = (strchr(svalue(mode), 'k') != NULL);
@@ -542,8 +541,8 @@ static int table_query (lua_State *L) {
   t = hvalue(obj_at(L, 1));
   if (i == -1) {
     lua_pushinteger(L, t->sizearray);
-    lua_pushinteger(L, sizenode(t));
-    lua_pushinteger(L, t->firstfree - t->node);
+    lua_pushinteger(L, t->node == &luaH_dummynode ? 0 : sizenode(t));
+    lua_pushinteger(L, t->lastfree - t->node);
   }
   else if (i < t->sizearray) {
     lua_pushinteger(L, i);
diff --git a/lvm.c b/lvm.c
index 52c5b4c1..2f9612c4 100644
--- a/lvm.c
+++ b/lvm.c
@@ -1,5 +1,5 @@
 /*
-** $Id: lvm.c,v 2.18 2004/12/03 20:35:33 roberto Exp $
+** $Id: lvm.c,v 2.19 2005/01/04 15:55:12 roberto Exp roberto $
 ** Lua virtual machine
 ** See Copyright Notice in lua.h
 */
@@ -462,7 +462,7 @@ StkId luaV_execute (lua_State *L, int nexeccalls) {
       case OP_NEWTABLE: {
         int b = GETARG_B(i);
         int c = GETARG_C(i);
-        sethvalue(L, ra, luaH_new(L, luaO_fb2int(b), luaO_fb2int(c) - 1));
+        sethvalue(L, ra, luaH_new(L, luaO_fb2int(b), luaO_fb2int(c)));
         L->ci->savedpc = pc;
         luaC_checkGC(L);  /***/
         base = L->base;