* src/LR0.c (state_list_t, state_list_append): New.

[bison.git] / src / LR0.c

diff --git a/src/LR0.c b/src/LR0.c
index 3f06fedff43954323552cf510b349f92b72a515b..27282bd1e88590337082ad773eab60b6a1022bc7 100644 (file)
--- a/src/LR0.c
+++ b/src/LR0.c
@@ -1,5 +1,5 @@
 /* Generate the nondeterministic finite state machine for bison,
 /* Generate the nondeterministic finite state machine for bison,

-   Copyright 1984, 1986, 1989, 2000, 2001  Free Software Foundation, Inc.
+   Copyright 1984, 1986, 1989, 2000, 2001, 2002  Free Software Foundation, Inc.

 
    This file is part of Bison, the GNU Compiler Compiler.
 
 
    This file is part of Bison, the GNU Compiler Compiler.
 


@@ -23,7 +23,10 @@
    The entry point is generate_states.  */
 
 #include "system.h"
    The entry point is generate_states.  */
 
 #include "system.h"

+#include "bitset.h"
+#include "quotearg.h"

 #include "symtab.h"
 #include "symtab.h"

+#include "gram.h"

 #include "getargs.h"
 #include "reader.h"
 #include "gram.h"
 #include "getargs.h"
 #include "reader.h"
 #include "gram.h"


@@ -34,40 +37,45 @@
 #include "lalr.h"
 #include "reduce.h"
 
 #include "lalr.h"
 #include "reduce.h"
 

-int nstates;
-/* Initialize the final state to -1, otherwise, it might be set to 0
-   by default, and since we don't compute the reductions of the final
-   state, we end up not computing the reductions of the initial state,
-   which is of course needed.
+typedef struct state_list_s
+{
+  struct state_list_s *next;
+  state_t *state;
+} state_list_t;

 
 

-   FINAL_STATE is properly set by new_state when it recognizes the
-   accessing symbol: EOF.  */
-int final_state = -1;
-static state_t *first_state = NULL;
+static state_list_t *first_state = NULL;
+static state_list_t *last_state = NULL;

 
 

-static state_t *this_state = NULL;
-static state_t *last_state = NULL;
+static void
+state_list_append (state_t *state)
+{
+  state_list_t *node = XMALLOC (state_list_t, 1);
+  node->next = NULL;
+  node->state = state;
+
+  if (!first_state)
+    first_state = node;
+  if (last_state)
+    last_state->next = node;
+  last_state = node;
+}

 
 static int nshifts;
 
 static int nshifts;

-static short *shift_symbol = NULL;
+static symbol_number_t *shift_symbol = NULL;

 
 static short *redset = NULL;
 
 static short *redset = NULL;

-static short *shiftset = NULL;
+static state_number_t *shiftset = NULL;

 
 

-static short **kernel_base = NULL;
+static item_number_t **kernel_base = NULL;

 static int *kernel_size = NULL;
 static int *kernel_size = NULL;

-static short *kernel_items = NULL;
-
-/* hash table for states, to recognize equivalent ones.  */
-
-#define        STATE_HASH_SIZE 1009
-static state_t **state_hash = NULL;
+static item_number_t *kernel_items = NULL;

 
 \f
 static void
 allocate_itemsets (void)
 {
 
 \f
 static void
 allocate_itemsets (void)
 {

-  int i;
+  int i, r;
+  item_number_t *rhsp;

 
   /* Count the number of occurrences of all the symbols in RITEMS.
      Note that useless productions (hence useless nonterminals) are
 
   /* Count the number of occurrences of all the symbols in RITEMS.
      Note that useless productions (hence useless nonterminals) are


@@ -76,11 +84,11 @@ allocate_itemsets (void)
   int count = 0;
   short *symbol_count = XCALLOC (short, nsyms + nuseless_nonterminals);
 
   int count = 0;
   short *symbol_count = XCALLOC (short, nsyms + nuseless_nonterminals);
 

-  for (i = 0; i < nritems; ++i)
-    if (ritem[i] >= 0)
+  for (r = 1; r < nrules + 1; ++r)
+    for (rhsp = rules[r].rhs; *rhsp >= 0; ++rhsp)

       {
        count++;
       {
        count++;

-       symbol_count[ritem[i]]++;
+       symbol_count[*rhsp]++;

       }
 
   /* See comments before new_itemsets.  All the vectors of items
       }
 
   /* See comments before new_itemsets.  All the vectors of items


@@ -89,9 +97,9 @@ allocate_itemsets (void)
      appears as an item, which is symbol_count[symbol].
      We allocate that much space for each symbol.  */
 
      appears as an item, which is symbol_count[symbol].
      We allocate that much space for each symbol.  */
 

-  kernel_base = XCALLOC (short *, nsyms);
+  kernel_base = XCALLOC (item_number_t *, nsyms);

   if (count)
   if (count)

-    kernel_items = XCALLOC (short, count);
+    kernel_items = XCALLOC (item_number_t, count);

 
   count = 0;
   for (i = 0; i < nsyms; i++)
 
   count = 0;
   for (i = 0; i < nsyms; i++)


@@ -110,9 +118,10 @@ allocate_storage (void)
 {
   allocate_itemsets ();
 
 {
   allocate_itemsets ();
 

-  shiftset = XCALLOC (short, nsyms);
+  shiftset = XCALLOC (state_number_t, nsyms);

   redset = XCALLOC (short, nrules + 1);
   redset = XCALLOC (short, nrules + 1);

-  state_hash = XCALLOC (state_t *, STATE_HASH_SIZE);
+  state_hash_new ();
+  shift_symbol = XCALLOC (symbol_number_t, nsyms);

 }
 
 
 }
 
 


@@ -125,53 +134,51 @@ free_storage (void)
   free (kernel_base);
   free (kernel_size);
   XFREE (kernel_items);
   free (kernel_base);
   free (kernel_size);
   XFREE (kernel_items);

-  free (state_hash);
+  state_hash_free ();

 }
 
 
 
 
 }
 
 
 
 

-/*----------------------------------------------------------------.
-| Find which symbols can be shifted in the current state, and for |
-| each one record which items would be active after that shift.   |
-| Uses the contents of itemset.                                   |
-|                                                                 |
-| shift_symbol is set to a vector of the symbols that can be      |
-| shifted.  For each symbol in the grammar, kernel_base[symbol]   |
-| points to a vector of item numbers activated if that symbol is  |
-| shifted, and kernel_size[symbol] is their numbers.              |
-`----------------------------------------------------------------*/
+/*---------------------------------------------------------------.
+| Find which symbols can be shifted in STATE, and for each one   |
+| record which items would be active after that shift.  Uses the |
+| contents of itemset.                                           |
+|                                                                |
+| shift_symbol is set to a vector of the symbols that can be     |
+| shifted.  For each symbol in the grammar, kernel_base[symbol]  |
+| points to a vector of item numbers activated if that symbol is |
+| shifted, and kernel_size[symbol] is their numbers.             |
+`---------------------------------------------------------------*/

 
 static void
 
 static void

-new_itemsets (void)
+new_itemsets (state_t *state)

 {
   int i;
 
   if (trace_flag)
     fprintf (stderr, "Entering new_itemsets, state = %d\n",
 {
   int i;
 
   if (trace_flag)
     fprintf (stderr, "Entering new_itemsets, state = %d\n",

-            this_state->number);
+            state->number);

 
   for (i = 0; i < nsyms; i++)
     kernel_size[i] = 0;
 
 
   for (i = 0; i < nsyms; i++)
     kernel_size[i] = 0;
 

-  shift_symbol = XCALLOC (short, nsyms);

   nshifts = 0;
 
   nshifts = 0;
 

-  for (i = 0; i < nitemset; ++i)
-    {
-      int symbol = ritem[itemset[i]];
-      if (symbol >= 0)
-       {
-         if (!kernel_size[symbol])
-           {
-             shift_symbol[nshifts] = symbol;
-             nshifts++;
-           }
-
-         kernel_base[symbol][kernel_size[symbol]] = itemset[i] + 1;
-         kernel_size[symbol]++;
-       }
-    }
+  for (i = 0; i < nritemset; ++i)
+    if (ritem[itemset[i]] >= 0)
+      {
+       symbol_number_t symbol
+         = item_number_as_symbol_number (ritem[itemset[i]]);
+       if (!kernel_size[symbol])
+         {
+           shift_symbol[nshifts] = symbol;
+           nshifts++;
+         }
+
+       kernel_base[symbol][kernel_size[symbol]] = itemset[i] + 1;
+       kernel_size[symbol]++;
+      }

 }
 
 
 }
 
 


@@ -182,33 +189,24 @@ new_itemsets (void)
 `-----------------------------------------------------------------*/
 
 static state_t *
 `-----------------------------------------------------------------*/
 
 static state_t *

-new_state (int symbol)
+new_state (symbol_number_t symbol, size_t core_size, item_number_t *core)

 {
 {

-  state_t *p;
+  state_t *res;

 
   if (trace_flag)
     fprintf (stderr, "Entering new_state, state = %d, symbol = %d (%s)\n",
 
   if (trace_flag)
     fprintf (stderr, "Entering new_state, state = %d, symbol = %d (%s)\n",

-            this_state->number, symbol, symbols[symbol]->tag);
-
-  if (nstates >= MAXSHORT)
-    fatal (_("too many states (max %d)"), MAXSHORT);
-
-  p = STATE_ALLOC (kernel_size[symbol]);
-  p->accessing_symbol = symbol;
-  p->number = nstates;
-  p->nitems = kernel_size[symbol];
-
-  shortcpy (p->items, kernel_base[symbol], kernel_size[symbol]);
+            nstates, symbol, symbol_tag_get (symbols[symbol]));

 
 

-  last_state->next = p;
-  last_state = p;
-  nstates++;
+  res = state_new (symbol, core_size, core);
+  state_hash_insert (res);

 
 

-  /* If this is the eoftoken, then this is the final state. */
-  if (symbol == 0)
-    final_state = p->number;
+  /* If this is the eoftoken, and this is not the initial state, then
+     this is the final state.  */
+  if (symbol == 0 && first_state)
+    final_state = res;

 
 

-  return p;
+  state_list_append (res);
+  return res;

 }
 
 
 }
 
 


@@ -218,56 +216,18 @@ new_state (int symbol)
 | equivalent one exists already.  Used by append_states.        |
 `--------------------------------------------------------------*/
 
 | equivalent one exists already.  Used by append_states.        |
 `--------------------------------------------------------------*/
 

-static int
-get_state (int symbol)
+static state_number_t
+get_state (symbol_number_t symbol, size_t core_size, item_number_t *core)

 {
 {

-  int key;
-  int i;

   state_t *sp;
 
   if (trace_flag)
   state_t *sp;
 
   if (trace_flag)

-    fprintf (stderr, "Entering get_state, state = %d, symbol = %d (%s)\n",
-            this_state->number, symbol, symbols[symbol]->tag);
-
-  /* Add up the target state's active item numbers to get a hash key.
-     */
-  key = 0;
-  for (i = 0; i < kernel_size[symbol]; ++i)
-    key += kernel_base[symbol][i];
-  key = key % STATE_HASH_SIZE;
-  sp = state_hash[key];
-
-  if (sp)
-    {
-      int found = 0;
-      while (!found)
-       {
-         if (sp->nitems == kernel_size[symbol])
-           {
-             found = 1;
-             for (i = 0; i < kernel_size[symbol]; ++i)
-               if (kernel_base[symbol][i] != sp->items[i])
-                 found = 0;
-           }
-
-         if (!found)
-           {
-             if (sp->link)
-               {
-                 sp = sp->link;
-               }
-             else              /* bucket exhausted and no match */
-               {
-                 sp = sp->link = new_state (symbol);
-                 found = 1;
-               }
-           }
-       }
-    }
-  else                         /* bucket is empty */
-    {
-      state_hash[key] = sp = new_state (symbol);
-    }
+    fprintf (stderr, "Entering get_state, symbol = %d (%s)\n",
+            symbol, symbol_tag_get (symbols[symbol]));
+
+  sp = state_hash_lookup (core_size, core);
+  if (!sp)
+    sp = new_state (symbol, core_size, core);

 
   if (trace_flag)
     fprintf (stderr, "Exiting get_state => %d\n", sp->number);
 
   if (trace_flag)
     fprintf (stderr, "Exiting get_state => %d\n", sp->number);


@@ -277,21 +237,21 @@ get_state (int symbol)
 
 /*------------------------------------------------------------------.
 | Use the information computed by new_itemsets to find the state    |
 
 /*------------------------------------------------------------------.
 | Use the information computed by new_itemsets to find the state    |

-| numbers reached by each shift transition from the current state.  |
+| numbers reached by each shift transition from STATE.              |

 |                                                                   |
 |                                                                   |

-| shiftset is set up as a vector of state numbers of those states.  |
+| SHIFTSET is set up as a vector of state numbers of those states.  |

 `------------------------------------------------------------------*/
 
 static void
 `------------------------------------------------------------------*/
 
 static void

-append_states (void)
+append_states (state_t *state)

 {
   int i;
   int j;
 {
   int i;
   int j;

-  int symbol;
+  symbol_number_t symbol;

 
   if (trace_flag)
     fprintf (stderr, "Entering append_states, state = %d\n",
 
   if (trace_flag)
     fprintf (stderr, "Entering append_states, state = %d\n",

-            this_state->number);
+            state->number);

 
   /* first sort shift_symbol into increasing order */
 
 
   /* first sort shift_symbol into increasing order */
 


@@ -308,30 +268,24 @@ append_states (void)
     }
 
   for (i = 0; i < nshifts; i++)
     }
 
   for (i = 0; i < nshifts; i++)

-    shiftset[i] = get_state (shift_symbol[i]);
+    {
+      symbol = shift_symbol[i];
+      shiftset[i] = get_state (symbol,
+                              kernel_size[symbol], kernel_base[symbol]);
+    }

 }
 
 
 static void
 new_states (void)
 {
 }
 
 
 static void
 new_states (void)
 {

-  first_state = last_state = this_state = STATE_ALLOC (0);
-  nstates = 1;
+  /* The 0 at the lhs is the index of the item of this initial rule.  */
+  kernel_base[0][0] = 0;
+  kernel_size[0] = 1;
+  state_list_append (new_state (0, kernel_size[0], kernel_base[0]));

 }
 
 
 }
 
 

-/*------------------------------------------------------------.
-| Save the NSHIFTS of SHIFTSET into the current linked list.  |
-`------------------------------------------------------------*/
-
-static void
-save_shifts (void)
-{
-  shifts *p = shifts_new (nshifts);
-  shortcpy (p->shifts, shiftset, nshifts);
-  this_state->shifts = p;
-}
-

 
 /*----------------------------------------------------------------.
 | Find which rules can be used for reduction transitions from the |
 
 /*----------------------------------------------------------------.
 | Find which rules can be used for reduction transitions from the |


@@ -340,18 +294,18 @@ save_shifts (void)
 `----------------------------------------------------------------*/
 
 static void
 `----------------------------------------------------------------*/
 
 static void

-save_reductions (void)
+save_reductions (state_t *state)

 {
   int count = 0;
   int i;
 
   /* If this is the final state, we want it to have no reductions at
      all, although it has one for `START_SYMBOL EOF .'.  */
 {
   int count = 0;
   int i;
 
   /* If this is the final state, we want it to have no reductions at
      all, although it has one for `START_SYMBOL EOF .'.  */

-  if (this_state->number == final_state)
+  if (final_state && state->number == final_state->number)

     return;
 
   /* Find and count the active items that represent ends of rules. */
     return;
 
   /* Find and count the active items that represent ends of rules. */

-  for (i = 0; i < nitemset; ++i)
+  for (i = 0; i < nritemset; ++i)

     {
       int item = ritem[itemset[i]];
       if (item < 0)
     {
       int item = ritem[itemset[i]];
       if (item < 0)


@@ -359,37 +313,45 @@ save_reductions (void)
     }
 
   /* Make a reductions structure and copy the data into it.  */
     }
 
   /* Make a reductions structure and copy the data into it.  */

-  this_state->reductions = reductions_new (count);
-  shortcpy (this_state->reductions->rules, redset, count);
+  state->reductions = reductions_new (count);
+  memcpy (state->reductions->rules, redset, count * sizeof (redset[0]));

 }
 
 \f
 }
 
 \f

-/*--------------------.
+/*---------------.

 | Build STATES.  |
 | Build STATES.  |

-`--------------------*/
+`---------------*/

 
 static void
 set_states (void)
 {
 
 static void
 set_states (void)
 {

-  state_t *sp;

   states = XCALLOC (state_t *, nstates);
 
   states = XCALLOC (state_t *, nstates);
 

-  for (sp = first_state; sp; sp = sp->next)
+  while (first_state)

     {
     {

+      state_list_t *this = first_state;
+

       /* Pessimization, but simplification of the code: make sure all
         the states have a shifts, errs, and reductions, even if
         reduced to 0.  */
       /* Pessimization, but simplification of the code: make sure all
         the states have a shifts, errs, and reductions, even if
         reduced to 0.  */

-      if (!sp->shifts)
-       sp->shifts = shifts_new (0);
-      if (!sp->errs)
-       sp->errs = errs_new (0);
-      if (!sp->reductions)
-       sp->reductions = reductions_new (0);
-
-      states[sp->number] = sp;
+      state_t *state = this->state;
+      if (!state->shifts)
+       state_shifts_set (state, 0, 0);
+      if (!state->errs)
+       state->errs = errs_new (0);
+      if (!state->reductions)
+       state->reductions = reductions_new (0);
+
+      states[state->number] = state;
+
+      first_state = this->next;
+      free (this);

     }
     }

+  first_state = NULL;
+  last_state = NULL;

 }
 
 }
 

+

 /*-------------------------------------------------------------------.
 | Compute the nondeterministic finite state machine (see state.h for |
 | details) from the grammar.                                         |
 /*-------------------------------------------------------------------.
 | Compute the nondeterministic finite state machine (see state.h for |
 | details) from the grammar.                                         |


@@ -398,34 +360,38 @@ set_states (void)
 void
 generate_states (void)
 {
 void
 generate_states (void)
 {

+  state_list_t *list = NULL;

   allocate_storage ();
   new_closure (nritems);
   new_states ();
   allocate_storage ();
   new_closure (nritems);
   new_states ();

+  list = first_state;

 
 

-  while (this_state)
+  while (list)

     {
     {

+      state_t *state = list->state;

       if (trace_flag)
        fprintf (stderr, "Processing state %d (reached by %s)\n",
       if (trace_flag)
        fprintf (stderr, "Processing state %d (reached by %s)\n",

-                this_state->number,
-                symbols[this_state->accessing_symbol]->tag);
+                state->number,
+                symbol_tag_get (symbols[state->accessing_symbol]));

       /* Set up ruleset and itemset for the transitions out of this
          state.  ruleset gets a 1 bit for each rule that could reduce
          now.  itemset gets a vector of all the items that could be
          accepted next.  */
       /* Set up ruleset and itemset for the transitions out of this
          state.  ruleset gets a 1 bit for each rule that could reduce
          now.  itemset gets a vector of all the items that could be
          accepted next.  */

-      closure (this_state->items, this_state->nitems);
-      /* record the reductions allowed out of this state */
-      save_reductions ();
-      /* find the itemsets of the states that shifts can reach */
-      new_itemsets ();
-      /* find or create the core structures for those states */
-      append_states ();
-
-      /* create the shifts structures for the shifts to those states,
-         now that the state numbers transitioning to are known */
-      save_shifts ();
-
-      /* states are queued when they are created; process them all */
-      this_state = this_state->next;
+      closure (state->items, state->nitems);
+      /* Record the reductions allowed out of this state.  */
+      save_reductions (state);
+      /* Find the itemsets of the states that shifts can reach.  */
+      new_itemsets (state);
+      /* Find or create the core structures for those states.  */
+      append_states (state);
+
+      /* Create the shifts structures for the shifts to those states,
+        now that the state numbers transitioning to are known.  */
+      state_shifts_set (state, nshifts, shiftset);
+
+      /* States are queued when they are created; process them all.
+        */
+      list = list->next;

     }
 
   /* discard various storage */
     }
 
   /* discard various storage */