* src/print-xml.c (escape_bufs): New static global variable

[bison.git] / src / reduce.c

diff --git a/src/reduce.c b/src/reduce.c
index 9f9c2ee7369337d19236a6b674307e56f28e02b9..2e309a44bd77081c43d1a9f4e749693603cac1bc 100644 (file)
--- a/src/reduce.c
+++ b/src/reduce.c
@@ -1,598 +1,523 @@
 /* Grammar reduction for Bison.
 /* Grammar reduction for Bison.

-   Copyright (C) 1988, 1989 Free Software Foundation, Inc.

 
 

-This file is part of Bison, the GNU Compiler Compiler.
+   Copyright (C) 1988, 1989, 2000, 2001, 2002, 2003, 2005, 2006,
+   2007 Free Software Foundation, Inc.

 
 

-Bison is free software; you can redistribute it and/or modify
-it under the terms of the GNU General Public License as published by
-the Free Software Foundation; either version 2, or (at your option)
-any later version.
+   This file is part of Bison, the GNU Compiler Compiler.

 
 

-Bison is distributed in the hope that it will be useful,
-but WITHOUT ANY WARRANTY; without even the implied warranty of
-MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-GNU General Public License for more details.
+   This program is free software: you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation, either version 3 of the License, or
+   (at your option) any later version.

 
 

-You should have received a copy of the GNU General Public License
-along with Bison; see the file COPYING.  If not, write to
-the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.  */
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.

 
 

+   You should have received a copy of the GNU General Public License
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.  */

 
 

-/*
- * Reduce the grammar:  Find and eliminate unreachable terminals,
- * nonterminals, and productions.  David S. Bakin.
- */

 
 

-/*
- * Don't eliminate unreachable terminals:  They may be used by the user's
- * parser.
- */
+/* Reduce the grammar: Find and eliminate unreachable terminals,
+   nonterminals, and productions.  David S. Bakin.  */

 
 

-#include <stdio.h>
-#include "system.h"
-#include "files.h"
-#include "gram.h"
-#include "machine.h"
-#include "new.h"
+/* Don't eliminate unreachable terminals: They may be used by the
+   user's parser.  */

 
 

+#include <config.h>
+#include "system.h"

 
 

-extern char   **tags;          /* reader.c */
-extern int      verboseflag;   /* getargs.c */
-static int      statisticsflag;        /* XXXXXXX */
-
-#ifndef TRUE
-#define TRUE   (1)
-#define FALSE  (0)
-#endif
-typedef int bool;
-typedef unsigned *BSet;
-typedef short  *rule;
-
+#include <bitset.h>
+#include <quotearg.h>

 
 

-/*
- * N is set of all nonterminals which are not useless.  P is set of all rules
- * which have no useless nonterminals in their RHS.  V is the set of all
- * accessible symbols.
- */
+#include "complain.h"
+#include "files.h"
+#include "getargs.h"
+#include "gram.h"
+#include "print-xml.h"
+#include "reader.h"
+#include "reduce.h"
+#include "symtab.h"

 
 

-static BSet     N, P, V, V1;
+/* Set of all nonterminals which are not useless.  */
+static bitset N;

 
 

-static int      nuseful_productions, nuseless_productions,
-                nuseful_nonterminals, nuseless_nonterminals;
+/* Set of all rules which have no useless nonterminals in their RHS.  */
+static bitset P;

 
 

+/* Set of all accessible symbols.  */
+static bitset V;

 
 

-static void useless_nonterminals();
-static void inaccessable_symbols();
-static void reduce_grammar_tables();
-static void print_results();
-static void print_notices();
-void dump_grammar();
+/* Set of symbols used to define rule precedence (so they are
+   `useless', but no warning should be issued).  */
+static bitset V1;

 
 

-extern void fatals ();
+static rule_number nuseful_productions;
+rule_number nuseless_productions;
+static int nuseful_nonterminals;
+symbol_number nuseless_nonterminals;

 \f
 \f

+/*-------------------------------------------------------------------.
+| Another way to do this would be with a set for each production and |
+| then do subset tests against N0, but even for the C grammar the    |
+| whole reducing process takes only 2 seconds on my 8Mhz AT.         |
+`-------------------------------------------------------------------*/

 
 

-bool
-bits_equal (L, R, n)
-BSet L;
-BSet R;
-int n;
+static bool
+useful_production (rule_number r, bitset N0)

 {
 {

-  int i;
-
-  for (i = n - 1; i >= 0; i--)
-    if (L[i] != R[i])
-      return FALSE;
-  return TRUE;
-}
+  item_number *rhsp;

 
 

+  /* A production is useful if all of the nonterminals in its appear
+     in the set of useful nonterminals.  */

 
 

-int
-nbits (i)
-unsigned i;
-{
-  int count = 0;
-
-  while (i != 0) {
-    i ^= (i & -i);
-    ++count;
-  }
-  return count;
+  for (rhsp = rules[r].rhs; *rhsp >= 0; ++rhsp)
+    if (ISVAR (*rhsp) && !bitset_test (N0, *rhsp - ntokens))
+      return false;
+  return true;

 }
 
 
 }
 
 

-int
-bits_size (S, n)
-BSet S;
-int n;
-{
-  int i, count = 0;
+/*---------------------------------------------------------.
+| Remember that rules are 1-origin, symbols are 0-origin.  |
+`---------------------------------------------------------*/

 
 

-  for (i = n - 1; i >= 0; i--)
-    count += nbits(S[i]);
-  return count;
-}
-\f
-void
-reduce_grammar ()
+static void
+useless_nonterminals (void)

 {
 {

-  bool reduced;
+  bitset Np, Ns;
+  rule_number r;

 
 

-  /* Allocate the global sets used to compute the reduced grammar */
+  /* N is set as built.  Np is set being built this iteration. P is
+     set of all productions which have a RHS all in N.  */

 
 

-  N = NEW2(WORDSIZE(nvars), unsigned);
-  P = NEW2(WORDSIZE(nrules + 1), unsigned);
-  V = NEW2(WORDSIZE(nsyms), unsigned);
-  V1 = NEW2(WORDSIZE(nsyms), unsigned);
+  Np = bitset_create (nvars, BITSET_FIXED);

 
 

-  useless_nonterminals();
-  inaccessable_symbols();

 
 

-  reduced = (bool) (nuseless_nonterminals + nuseless_productions > 0);
+  /* The set being computed is a set of nonterminals which can derive
+     the empty string or strings consisting of all terminals. At each
+     iteration a nonterminal is added to the set if there is a
+     production with that nonterminal as its LHS for which all the
+     nonterminals in its RHS are already in the set.  Iterate until
+     the set being computed remains unchanged.  Any nonterminals not
+     in the set at that point are useless in that they will never be
+     used in deriving a sentence of the language.

 
 

-  if (verboseflag)
-    print_results();
+     This iteration doesn't use any special traversal over the
+     productions.  A set is kept of all productions for which all the
+     nonterminals in the RHS are in useful.  Only productions not in
+     this set are scanned on each iteration.  At the end, this set is
+     saved to be used when finding useful productions: only
+     productions in this set will appear in the final grammar.  */

 
 

-  if (reduced == FALSE)
-    goto done_reducing;
-
-  print_notices();
-
-  if (!BITISSET(N, start_symbol - ntokens))
-    fatals(_("Start symbol %s does not derive any sentence"),
-          tags[start_symbol]);
-
-  reduce_grammar_tables();
-  /* if (verboseflag) {
-     fprintf(foutput, "REDUCED GRAMMAR\n\n");
-     dump_grammar();
-     }
-     */
-
-  /**/ statisticsflag = FALSE; /* someday getopts should handle this */
-  if (statisticsflag == TRUE)
-    fprintf(stderr,
-           _("reduced %s defines %d terminal%s, %d nonterminal%s\
-, and %d production%s.\n"), infile,
-           ntokens, (ntokens == 1 ? "" : "s"),
-           nvars,   (nvars   == 1 ? "" : "s"),
-           nrules,  (nrules  == 1 ? "" : "s"));
-
- done_reducing:
-
-  /* Free the global sets used to compute the reduced grammar */
-
-  FREE(N);
-  FREE(V);
-  FREE(P);
-
-}
-\f
-/*
- * Another way to do this would be with a set for each production and then do
- * subset tests against N, but even for the C grammar the whole reducing
- * process takes only 2 seconds on my 8Mhz AT.
- */
-
-static bool
-useful_production (i, N)
-int  i;
-BSet N;
-{
-  rule  r;
-  short n;
-
-  /*
-   * A production is useful if all of the nonterminals in its RHS
-   * appear in the set of useful nonterminals.
-   */
-
-  for (r = &ritem[rrhs[i]]; *r > 0; r++)
-    if (ISVAR(n = *r))
-      if (!BITISSET(N, n - ntokens))
-       return FALSE;
-  return TRUE;
-}
-
-
-/* Remember that rules are 1-origin, symbols are 0-origin. */
-
-static void
-useless_nonterminals ()
-{
-  BSet Np, Ns;
-  int  i, n;
-
-  /*
-   * N is set as built.  Np is set being built this iteration. P is set
-   * of all productions which have a RHS all in N.
-   */
-
-  Np = NEW2(WORDSIZE(nvars), unsigned);
-
-  /*
-   * The set being computed is a set of nonterminals which can derive
-   * the empty string or strings consisting of all terminals. At each
-   * iteration a nonterminal is added to the set if there is a
-   * production with that nonterminal as its LHS for which all the
-   * nonterminals in its RHS are already in the set.  Iterate until the
-   * set being computed remains unchanged.  Any nonterminals not in the
-   * set at that point are useless in that they will never be used in
-   * deriving a sentence of the language.
-   *
-   * This iteration doesn't use any special traversal over the
-   * productions.  A set is kept of all productions for which all the
-   * nonterminals in the RHS are in useful.  Only productions not in
-   * this set are scanned on each iteration.  At the end, this set is
-   * saved to be used when finding useful productions: only productions
-   * in this set will appear in the final grammar.
-   */
-
-  n = 0;

   while (1)
     {
   while (1)
     {

-      for (i = WORDSIZE(nvars) - 1; i >= 0; i--)
-       Np[i] = N[i];
-      for (i = 1; i <= nrules; i++)
-       {
-         if (!BITISSET(P, i))
-           {
-             if (useful_production(i, N))
-               {
-                 SETBIT(Np, rlhs[i] - ntokens);
-                 SETBIT(P, i);
-               }
-           }
-       }
-      if (bits_equal(N, Np, WORDSIZE(nvars)))
+      bitset_copy (Np, N);
+      for (r = 0; r < nrules; r++)
+       if (!bitset_test (P, r)
+           && useful_production (r, N))
+         {
+           bitset_set (Np, rules[r].lhs->number - ntokens);
+           bitset_set (P, r);
+         }
+      if (bitset_equal_p (N, Np))

        break;
       Ns = Np;
       Np = N;
       N = Ns;
     }
        break;
       Ns = Np;
       Np = N;
       N = Ns;
     }

-  FREE(N);
+  bitset_free (N);

   N = Np;
 }
   N = Np;
 }

-\f
+
+

 static void
 static void

-inaccessable_symbols ()
+inaccessable_symbols (void)

 {
 {

-  BSet  Vp, Vs, Pp;
-  int   i, n;
-  short t;
-  rule  r;
-
-  /*
-   * Find out which productions are reachable and which symbols are
-   * used.  Starting with an empty set of productions and a set of
-   * symbols which only has the start symbol in it, iterate over all
-   * productions until the set of productions remains unchanged for an
-   * iteration.  For each production which has a LHS in the set of
-   * reachable symbols, add the production to the set of reachable
-   * productions, and add all of the nonterminals in the RHS of the
-   * production to the set of reachable symbols.
-   *
-   * Consider only the (partially) reduced grammar which has only
-   * nonterminals in N and productions in P.
-   *
-   * The result is the set P of productions in the reduced grammar, and
-   * the set V of symbols in the reduced grammar.
-   *
-   * Although this algorithm also computes the set of terminals which are
-   * reachable, no terminal will be deleted from the grammar. Some
-   * terminals might not be in the grammar but might be generated by
-   * semantic routines, and so the user might want them available with
-   * specified numbers.  (Is this true?)  However, the nonreachable
-   * terminals are printed (if running in verbose mode) so that the user
-   * can know.
-   */
-
-  Vp = NEW2(WORDSIZE(nsyms), unsigned);
-  Pp = NEW2(WORDSIZE(nrules + 1), unsigned);
+  bitset Vp, Vs, Pp;
+
+  /* Find out which productions are reachable and which symbols are
+     used.  Starting with an empty set of productions and a set of
+     symbols which only has the start symbol in it, iterate over all
+     productions until the set of productions remains unchanged for an
+     iteration.  For each production which has a LHS in the set of
+     reachable symbols, add the production to the set of reachable
+     productions, and add all of the nonterminals in the RHS of the
+     production to the set of reachable symbols.
+
+     Consider only the (partially) reduced grammar which has only
+     nonterminals in N and productions in P.
+
+     The result is the set P of productions in the reduced grammar,
+     and the set V of symbols in the reduced grammar.
+
+     Although this algorithm also computes the set of terminals which
+     are reachable, no terminal will be deleted from the grammar. Some
+     terminals might not be in the grammar but might be generated by
+     semantic routines, and so the user might want them available with
+     specified numbers.  (Is this true?)  However, the nonreachable
+     terminals are printed (if running in verbose mode) so that the
+     user can know.  */
+
+  Vp = bitset_create (nsyms, BITSET_FIXED);
+  Pp = bitset_create (nrules, BITSET_FIXED);

 
   /* If the start symbol isn't useful, then nothing will be useful. */
 
   /* If the start symbol isn't useful, then nothing will be useful. */

-  if (!BITISSET(N, start_symbol - ntokens))
-    goto end_iteration;
-
-  SETBIT(V, start_symbol);
-
-  n = 0;
-  while (1)
+  if (bitset_test (N, accept->number - ntokens))

     {
     {

-      for (i = WORDSIZE(nsyms) - 1; i >= 0; i--)
-       Vp[i] = V[i];
-      for (i = 1; i <= nrules; i++)
+      bitset_set (V, accept->number);
+
+      while (1)

        {
        {

-         if (!BITISSET(Pp, i) && BITISSET(P, i) &&
-             BITISSET(V, rlhs[i]))
+         rule_number r;
+         bitset_copy (Vp, V);
+         for (r = 0; r < nrules; r++)

            {
            {

-             for (r = &ritem[rrhs[i]]; *r >= 0; r++)
+             if (!bitset_test (Pp, r)
+                 && bitset_test (P, r)
+                 && bitset_test (V, rules[r].lhs->number))

                {
                {

-                 if (ISTOKEN(t = *r)
-                     || BITISSET(N, t - ntokens))
-                   {
-                     SETBIT(Vp, t);
-                   }
+                 item_number *rhsp;
+                 for (rhsp = rules[r].rhs; *rhsp >= 0; rhsp++)
+                   if (ISTOKEN (*rhsp) || bitset_test (N, *rhsp - ntokens))
+                     bitset_set (Vp, *rhsp);
+                 bitset_set (Pp, r);

                }
                }

-             SETBIT(Pp, i);

            }
            }

+         if (bitset_equal_p (V, Vp))
+           break;
+         Vs = Vp;
+         Vp = V;
+         V = Vs;

        }
        }

-      if (bits_equal(V, Vp, WORDSIZE(nsyms)))
-       {
-         break;
-       }
-      Vs = Vp;
-      Vp = V;
-      V = Vs;

     }
     }

- end_iteration:

 
 

-  FREE(V);
+  bitset_free (V);

   V = Vp;
 
   /* Tokens 0, 1, and 2 are internal to Bison.  Consider them useful. */
   V = Vp;
 
   /* Tokens 0, 1, and 2 are internal to Bison.  Consider them useful. */

-  SETBIT(V, 0);                        /* end-of-input token */
-  SETBIT(V, 1);                        /* error token */
-  SETBIT(V, 2);                        /* some undefined token */
+  bitset_set (V, endtoken->number);            /* end-of-input token */
+  bitset_set (V, errtoken->number);            /* error token */
+  bitset_set (V, undeftoken->number);          /* some undefined token */

 
 

-  FREE(P);
+  bitset_free (P);

   P = Pp;
 
   P = Pp;
 

-  nuseful_productions = bits_size(P, WORDSIZE(nrules + 1));
+  nuseful_productions = bitset_count (P);

   nuseless_productions = nrules - nuseful_productions;
 
   nuseful_nonterminals = 0;
   nuseless_productions = nrules - nuseful_productions;
 
   nuseful_nonterminals = 0;

-  for (i = ntokens; i < nsyms; i++)
-    if (BITISSET(V, i))
-      nuseful_nonterminals++;
+  {
+    symbol_number i;
+    for (i = ntokens; i < nsyms; i++)
+      if (bitset_test (V, i))
+       nuseful_nonterminals++;
+  }

   nuseless_nonterminals = nvars - nuseful_nonterminals;
 
   /* A token that was used in %prec should not be warned about.  */
   nuseless_nonterminals = nvars - nuseful_nonterminals;
 
   /* A token that was used in %prec should not be warned about.  */

-  for (i = 1; i < nrules; i++)
-    if (rprecsym[i] != 0)
-      SETBIT(V1, rprecsym[i]);
+  {
+    rule_number r;
+    for (r = 0; r < nrules; ++r)
+      if (rules[r].precsym != 0)
+       bitset_set (V1, rules[r].precsym->number);
+  }

 }
 }

-\f
-static void
-reduce_grammar_tables ()
-{
-/* This is turned off because we would need to change the numbers
-   in the case statements in the actions file.  */
-#if 0
-  /* remove useless productions */
-  if (nuseless_productions > 0)
-    {
-      short np, pn, ni, pi;
-
-      np = 0;
-      ni = 0;
-      for (pn = 1; pn <= nrules; pn++)
-       {
-         if (BITISSET(P, pn))
-           {
-             np++;
-             if (pn != np)
-               {
-                 rlhs[np] = rlhs[pn];
-                 rline[np] = rline[pn];
-                 rprec[np] = rprec[pn];
-                 rassoc[np] = rassoc[pn];
-                 rrhs[np] = rrhs[pn];
-                 if (rrhs[np] != ni)
-                   {
-                     pi = rrhs[np];
-                     rrhs[np] = ni;
-                     while (ritem[pi] >= 0)
-                       ritem[ni++] = ritem[pi++];
-                     ritem[ni++] = -np;
-                   }
-               } else {
-                 while (ritem[ni++] >= 0);
-               }
-           }
-       }
-      ritem[ni] = 0;
-      nrules -= nuseless_productions;
-      nitems = ni;

 
 

-      /*
-       * Is it worth it to reduce the amount of memory for the
-       * grammar? Probably not.
-       */

 
 

-    }
-#endif /* 0 */
-  /* Disable useless productions,
-     since they may contain useless nonterms
-     that would get mapped below to -1 and confuse everyone.  */
-  if (nuseless_productions > 0)
-    {
-      int pn;
+/*-------------------------------------------------------------------.
+| Put the useless productions at the end of RULES, and adjust NRULES |
+| accordingly.                                                       |
+`-------------------------------------------------------------------*/

 
 

-      for (pn = 1; pn <= nrules; pn++)
-       {
-         if (!BITISSET(P, pn))
-           {
-             rlhs[pn] = -1;
-           }
-       }
-    }
+static void
+reduce_grammar_tables (void)
+{
+  /* Report and flag useless productions.  */
+  {
+    rule_number r;
+    for (r = 0; r < nrules; r++)
+      rules[r].useful = bitset_test (P, r);
+    grammar_rules_never_reduced_report (_("useless rule"));
+  }

 
 

-  /* remove useless symbols */
-  if (nuseless_nonterminals > 0)
-    {
+  /* Map the nonterminals to their new index: useful first, useless
+     afterwards.  Kept for later report.  */
+  {
+    int useful = 0;
+    int useless = nrules - nuseless_productions;
+    rule *rules_sorted = xnmalloc (nrules, sizeof *rules_sorted);
+    rule_number r;
+    for (r = 0; r < nrules; ++r)
+      rules_sorted[rules[r].useful ? useful++ : useless++] = rules[r];
+    free (rules);
+    rules = rules_sorted;
+
+    /* Renumber the rules markers in RITEMS.  */
+    for (r = 0; r < nrules; ++r)
+      {
+       item_number *rhsp = rules[r].rhs;
+       for (/* Nothing. */; *rhsp >= 0; ++rhsp)
+         /* Nothing. */;
+       *rhsp = rule_number_as_item_number (r);
+       rules[r].number = r;
+      }
+    nrules -= nuseless_productions;
+  }

 
 

-      int    i, n;
-/*      short  j; JF unused */
-      short *nontermmap;
-      rule   r;
+  /* Adjust NRITEMS.  */
+  {
+    rule_number r;
+    int length;
+    for (r = nrules; r < nrules + nuseless_productions; ++r)
+      {
+       length = rule_rhs_length (&rules[r]);
+       nritems -= length + 1;
+      }
+  }
+}

 
 

-      /*
-       * create a map of nonterminal number to new nonterminal
-       * number. -1 in the map means it was useless and is being
-       * eliminated.
-       */

 
 

-      nontermmap = NEW2(nvars, short) - ntokens;
-      for (i = ntokens; i < nsyms; i++)
-       nontermmap[i] = -1;
+/*------------------------------.
+| Remove useless nonterminals.  |
+`------------------------------*/

 
 

-      n = ntokens;
-      for (i = ntokens; i < nsyms; i++)
-       if (BITISSET(V, i))
-         nontermmap[i] = n++;
+static void
+nonterminals_reduce (void)
+{
+  symbol_number i, n;

 
 

-      /* Shuffle elements of tables indexed by symbol number.  */
+  /* Map the nonterminals to their new index: useful first, useless
+     afterwards.  Kept for later report.  */

 
 

-      for (i = ntokens; i < nsyms; i++)
-       {
-         n = nontermmap[i];
-         if (n >= 0)
-           {
-             sassoc[n] = sassoc[i];
-             sprec[n] = sprec[i];
-             tags[n] = tags[i];
-           } else {
-             free(tags[i]);
-           }
-       }
+  symbol_number *nontermmap = xnmalloc (nvars, sizeof *nontermmap);
+  n = ntokens;
+  for (i = ntokens; i < nsyms; i++)
+    if (bitset_test (V, i))
+      nontermmap[i - ntokens] = n++;
+  for (i = ntokens; i < nsyms; i++)
+    if (!bitset_test (V, i))
+      {
+       nontermmap[i - ntokens] = n++;
+       warn_at (symbols[i]->location, _("useless nonterminal: %s"),
+                symbols[i]->tag);
+      }
+
+
+  /* Shuffle elements of tables indexed by symbol number.  */
+  {
+    symbol **symbols_sorted = xnmalloc (nvars, sizeof *symbols_sorted);
+
+    for (i = ntokens; i < nsyms; i++)
+      symbols[i]->number = nontermmap[i - ntokens];
+    for (i = ntokens; i < nsyms; i++)
+      symbols_sorted[nontermmap[i - ntokens] - ntokens] = symbols[i];
+    for (i = ntokens; i < nsyms; i++)
+      symbols[i] = symbols_sorted[i - ntokens];
+    free (symbols_sorted);
+  }

 
 

-      /* Replace all symbol numbers in valid data structures.  */
+  {
+    rule_number r;
+    for (r = 0; r < nrules; ++r)
+      {
+       item_number *rhsp;
+       for (rhsp = rules[r].rhs; *rhsp >= 0; ++rhsp)
+         if (ISVAR (*rhsp))
+           *rhsp =  symbol_number_as_item_number (nontermmap[*rhsp
+                                                             - ntokens]);
+      }
+    accept->number = nontermmap[accept->number - ntokens];
+  }

 
 

-      for (i = 1; i <= nrules; i++)
-       {
-         /* Ignore the rules disabled above.  */
-         if (rlhs[i] >= 0)
-           rlhs[i] = nontermmap[rlhs[i]];
-         if (ISVAR (rprecsym[i]))
-           /* Can this happen?  */
-           rprecsym[i] = nontermmap[rprecsym[i]];
-       }
+  nsyms -= nuseless_nonterminals;
+  nvars -= nuseless_nonterminals;

 
 

-      for (r = ritem; *r; r++)
-       if (ISVAR(*r))
-         *r = nontermmap[*r];
+  free (nontermmap);
+}

 
 

-      start_symbol = nontermmap[start_symbol];

 
 

-      nsyms -= nuseless_nonterminals;
-      nvars -= nuseless_nonterminals;
+/*------------------------------------------------------------------.
+| Output the detailed results of the reductions.  For FILE.output.  |
+`------------------------------------------------------------------*/

 
 

-      free(&nontermmap[ntokens]);
-    }
-}
-\f
-static void
-print_results ()
+void
+reduce_output (FILE *out)

 {
 {

-  int   i;
-/*  short j; JF unused */
-  rule  r;
-  bool  b;
-

   if (nuseless_nonterminals > 0)
     {
   if (nuseless_nonterminals > 0)
     {

-      fprintf(foutput, _("Useless nonterminals:\n\n"));
-      for (i = ntokens; i < nsyms; i++)
-       if (!BITISSET(V, i))
-         fprintf(foutput, "   %s\n", tags[i]);
+      int i;
+      fprintf (out, "%s\n\n", _("Useless nonterminals"));
+      for (i = 0; i < nuseless_nonterminals; ++i)
+       fprintf (out, "   %s\n", symbols[nsyms + i]->tag);
+      fputs ("\n\n", out);

     }
     }

-  b = FALSE;
-  for (i = 0; i < ntokens; i++)
-    {
-      if (!BITISSET(V, i) && !BITISSET(V1, i))
+
+  {
+    bool b = false;
+    int i;
+    for (i = 0; i < ntokens; i++)
+      if (!bitset_test (V, i) && !bitset_test (V1, i))

        {
          if (!b)
        {
          if (!b)

-           {
-             fprintf(foutput, _("\n\nTerminals which are not used:\n\n"));
-             b = TRUE;
-           }
-         fprintf(foutput, "   %s\n", tags[i]);
+           fprintf (out, "%s\n\n", _("Terminals which are not used"));
+         b = true;
+         fprintf (out, "   %s\n", symbols[i]->tag);

        }
        }

-    }
+    if (b)
+      fputs ("\n\n", out);
+  }

 
   if (nuseless_productions > 0)
 
   if (nuseless_productions > 0)

-    {
-      fprintf(foutput, _("\n\nUseless rules:\n\n"));
-      for (i = 1; i <= nrules; i++)
-       {
-         if (!BITISSET(P, i))
-           {
-             fprintf(foutput, "#%-4d  ", i);
-             fprintf(foutput, "%s :\t", tags[rlhs[i]]);
-             for (r = &ritem[rrhs[i]]; *r >= 0; r++)
-               {
-                 fprintf(foutput, " %s", tags[*r]);
-               }
-             fprintf(foutput, ";\n");
-           }
-       }
-    }
-  if (nuseless_nonterminals > 0 || nuseless_productions > 0 || b)
-    fprintf(foutput, "\n\n");
+    grammar_rules_partial_print (out, _("Useless rules"),
+                                rule_useless_p);

 }
 \f
 }
 \f

+
+/*--------------------------------------------------------------.
+| Output the detailed results of the reductions.  For FILE.xml. |
+`---------------------------------------------------------------*/
+

 void
 void

-dump_grammar ()
+reduce_xml (FILE *out, int level)

 {
 {

-  int i;
-  rule r;
-
-  fprintf(foutput,
-         "ntokens = %d, nvars = %d, nsyms = %d, nrules = %d, nitems = %d\n\n",
-         ntokens, nvars, nsyms, nrules, nitems);
-  fprintf(foutput, _("Variables\n---------\n\n"));
-  fprintf(foutput, _("Value  Sprec    Sassoc    Tag\n"));
-  for (i = ntokens; i < nsyms; i++)
-    fprintf(foutput, "%5d  %5d  %5d  %s\n",
-           i, sprec[i], sassoc[i], tags[i]);
-  fprintf(foutput, "\n\n");
-  fprintf(foutput, _("Rules\n-----\n\n"));
-  for (i = 1; i <= nrules; i++)
-    {
-      fprintf(foutput, "%-5d(%5d%5d)%5d : (@%-5d)",
-             i, rprec[i], rassoc[i], rlhs[i], rrhs[i]);
-      for (r = &ritem[rrhs[i]]; *r > 0; r++)
-       fprintf(foutput, "%5d", *r);
-      fprintf(foutput, " [%d]\n", -(*r));
-    }
-  fprintf(foutput, "\n\n");
-  fprintf(foutput, _("Rules interpreted\n-----------------\n\n"));
-  for (i = 1; i <= nrules; i++)
+  fputc ('\n', out);
+  xml_puts (out, level, "<reductions>");
+  xml_puts (out, level + 1, "<useless>");
+
+  if (nuseless_nonterminals > 0)

     {
     {

-      fprintf(foutput, "%-5d  %s :", i, tags[rlhs[i]]);
-      for (r = &ritem[rrhs[i]]; *r > 0; r++)
-       fprintf(foutput, " %s", tags[*r]);
-      fprintf(foutput, "\n");
+      int i;
+      xml_puts (out, level + 2, "<nonterminals>");
+      for (i = 0; i < nuseless_nonterminals; ++i)
+       xml_printf (out, level + 3,
+                   "<nonterminal>%s</nonterminal>",
+                   symbols[nsyms + i]->tag);
+      xml_puts (out, level + 2, "</nonterminals>");

     }
     }

-  fprintf(foutput, "\n\n");
+  else
+    xml_puts (out, level + 2, "<nonterminals/>");
+
+  if (nuseless_productions > 0)
+    grammar_rules_partial_print_xml (out, level + 1, true, rule_useless_p);
+  else
+    xml_puts (out, level + 2, "<rules/>");
+
+  xml_puts (out, level + 1, "</useless>");
+  xml_puts (out, level + 1, "<unused>");
+
+  {
+    bool b = false;
+    int i;
+    for (i = 0; i < ntokens; i++)
+      if (!bitset_test (V, i) && !bitset_test (V1, i))
+       {
+         if (!b)
+           xml_puts (out, level + 2, "<terminals>");
+         b = true;
+         xml_printf (out, level + 3,
+                     "<terminal>%s</terminal>",
+                     xml_escape (symbols[i]->tag));
+       }
+    if (b)
+      xml_puts (out, level + 2, "</terminals>");
+    else
+      xml_puts (out, level + 2, "<terminals/>");
+  }
+
+  xml_puts (out, level + 1, "</unused>");
+  xml_puts (out, level, "</reductions>");
+  fputc ('\n', out);

 }
 }

+\f

 
 

+/*-------------------------------.
+| Report the results to STDERR.  |
+`-------------------------------*/

 
 static void
 
 static void

-print_notices ()
+reduce_print (void)

 {
 {

-  extern int fixed_outfiles;
+  if (yacc_flag && nuseless_productions)
+    fprintf (stderr, ngettext ("%d rule never reduced\n",
+                              "%d rules never reduced\n",
+                              nuseless_productions),
+            nuseless_productions);

 
 

-  if (fixed_outfiles && nuseless_productions)
-    fprintf(stderr, _("%d rules never reduced\n"), nuseless_productions);
-
-  fprintf(stderr, _("%s contains "), infile);
+  fprintf (stderr, "%s: %s: ", grammar_file, _("warning"));

 
   if (nuseless_nonterminals > 0)
 
   if (nuseless_nonterminals > 0)

-    {
-      fprintf(stderr, _("%d useless nonterminal%s"),
-             nuseless_nonterminals,
-             (nuseless_nonterminals == 1 ? "" : "s"));
-    }
+    fprintf (stderr, ngettext ("%d useless nonterminal",
+                              "%d useless nonterminals",
+                              nuseless_nonterminals),
+            nuseless_nonterminals);
+

   if (nuseless_nonterminals > 0 && nuseless_productions > 0)
   if (nuseless_nonterminals > 0 && nuseless_productions > 0)

-    fprintf(stderr, _(" and "));
+    fprintf (stderr, _(" and "));
+
+  if (nuseless_productions > 0)
+    fprintf (stderr, ngettext ("%d useless rule",
+                              "%d useless rules",
+                              nuseless_productions),
+            nuseless_productions);
+  fprintf (stderr, "\n");
+}
+\f
+void
+reduce_grammar (void)
+{
+  bool reduced;
+
+  /* Allocate the global sets used to compute the reduced grammar */
+
+  N = bitset_create (nvars, BITSET_FIXED);
+  P =  bitset_create (nrules, BITSET_FIXED);
+  V = bitset_create (nsyms, BITSET_FIXED);
+  V1 = bitset_create (nsyms, BITSET_FIXED);
+
+  useless_nonterminals ();
+  inaccessable_symbols ();
+
+  reduced = (nuseless_nonterminals + nuseless_productions > 0);
+  if (!reduced)
+    return;
+
+  reduce_print ();

 
 

+  if (!bitset_test (N, accept->number - ntokens))
+    fatal_at (startsymbol_location,
+             _("start symbol %s does not derive any sentence"),
+             startsymbol->tag);
+
+  /* First reduce the nonterminals, as they renumber themselves in the
+     whole grammar.  If you change the order, nonterms would be
+     renumbered only in the reduced grammar.  */
+  if (nuseless_nonterminals > 0)
+    nonterminals_reduce ();

   if (nuseless_productions > 0)
   if (nuseless_productions > 0)

+    reduce_grammar_tables ();
+
+  if (trace_flag & trace_grammar)

     {
     {

-      fprintf(stderr, _("%d useless rule%s"),
-             nuseless_productions,
-             (nuseless_productions == 1 ? "" : "s"));
+      grammar_dump (stderr, "Reduced Grammar");
+
+      fprintf (stderr, "reduced %s defines %d terminals, %d nonterminals\
+, and %d productions.\n",
+              grammar_file, ntokens, nvars, nrules);

     }
     }

-  fprintf(stderr, "\n");
-  fflush(stderr);
+}
+
+
+/*-----------------------------------------------------------.
+| Free the global sets used to compute the reduced grammar.  |
+`-----------------------------------------------------------*/
+
+void
+reduce_free (void)
+{
+  bitset_free (N);
+  bitset_free (V);
+  bitset_free (V1);
+  bitset_free (P);

 }
 }