X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/11652ab3dc44bf12d5fa9095d1495b5cd3102ff7..9a636f47cee64ab6e1f7588c6261fc46ea134c39:/src/reduce.c

diff --git a/src/reduce.c b/src/reduce.c
index 58bf43ea..25681750 100644
--- a/src/reduce.c
+++ b/src/reduce.c
@@ -1,5 +1,5 @@
 /* Grammar reduction for Bison.
-   Copyright 1988, 1989, 2000, 2001  Free Software Foundation, Inc.
+   Copyright (C) 1988, 1989, 2000, 2001, 2002  Free Software Foundation, Inc.
 
    This file is part of Bison, the GNU Compiler Compiler.
 
@@ -26,6 +26,7 @@
    user's parser.  */
 
 #include "system.h"
+#include "quotearg.h"
 #include "getargs.h"
 #include "files.h"
 #include "symtab.h"
@@ -118,7 +119,7 @@ useless_nonterminals (void)
 	if (!bitset_test (P, i)
 	    && useful_production (i, N))
 	  {
-	    bitset_set (Np, rules[i].lhs - ntokens);
+	    bitset_set (Np, rules[i].lhs->number - ntokens);
 	    bitset_set (P, i);
 	  }
       if (bitset_equal_p (N, Np))
@@ -178,7 +179,7 @@ inaccessable_symbols (void)
 	    {
 	      if (!bitset_test (Pp, i)
 		  && bitset_test (P, i)
-		  && bitset_test (V, rules[i].lhs))
+		  && bitset_test (V, rules[i].lhs->number))
 		{
 		  for (r = rules[i].rhs; *r >= 0; r++)
 		    if (ISTOKEN (t = *r) || bitset_test (N, t - ntokens))
@@ -198,9 +199,9 @@ inaccessable_symbols (void)
   V = Vp;
 
   /* Tokens 0, 1, and 2 are internal to Bison.  Consider them useful. */
-  bitset_set (V, 0);		/* end-of-input token */
-  bitset_set (V, 1);		/* error token */
-  bitset_set (V, 2);		/* some undefined token */
+  bitset_set (V, eoftoken->number);		/* end-of-input token */
+  bitset_set (V, errtoken->number);		/* error token */
+  bitset_set (V, undeftoken->number);		/* some undefined token */
 
   bitset_free (P);
   P = Pp;
@@ -217,73 +218,59 @@ inaccessable_symbols (void)
   /* A token that was used in %prec should not be warned about.  */
   for (i = 1; i < nrules + 1; i++)
     if (rules[i].precsym != 0)
-      bitset_set (V1, rules[i].precsym);
+      bitset_set (V1, rules[i].precsym->number);
 }
 
+
+/*-------------------------------------------------------------------.
+| Put the useless productions at the end of RULES, and adjust NRULES |
+| accordingly.                                                       |
+`-------------------------------------------------------------------*/
+
 static void
 reduce_grammar_tables (void)
 {
-  /* This is turned off because we would need to change the numbers in
-     the case statements in the actions file.
-
-     We don't disable it via CPP so that it is still checked with the
-     rest of the code, to avoid its becoming completely obsolete.
+  /* Flag useless productions.  */
+  {
+    int pn;
+    for (pn = 1; pn < nrules + 1; pn++)
+      rules[pn].useful = bitset_test (P, pn);
+  }
 
-     FIXME: I think the comment above demonstrates this code must be
-     turned off for *semantic* parser, not in the general case.  Try
-     to understand this better --akim.  */
+  /* Map the nonterminals to their new index: useful first, useless
+     afterwards.  Kept for later report.  */
+  {
+    int useful = 1;
+    int useless = nrules + 1 - nuseless_productions;
+    rule_t *rules_sorted = XMALLOC (rule_t, nrules + 1) - 1;
+    int i;
+    for (i = 1; i < nrules + 1; ++i)
+      rules_sorted[rules[i].useful ? useful++ : useless++] = rules[i];
+    free (rules + 1);
+    rules = rules_sorted;
 
-  if (0)
-    /* remove useless productions */
-    if (nuseless_productions > 0)
+    /* Renumber the rules markers in RITEMS.  */
+    for (i = 1; i < nrules + 1; ++i)
       {
-	short np, pn, ni, pi;
-
-	np = 0;
-	ni = 0;
-	for (pn = 1; pn < nrules + 1; pn++)
-	  if (bitset_test (P, pn))
-	    {
-	      np++;
-	      if (pn != np)
-		{
-		  rules[np].lhs   = rules[pn].lhs;
-		  rules[np].line  = rules[pn].line;
-		  rules[np].prec  = rules[pn].prec;
-		  rules[np].assoc = rules[pn].assoc;
-		  rules[np].rhs   = rules[pn].rhs;
-		  if (rules[np].rhs - ritem != ni)
-		    {
-		      pi = rules[np].rhs - ritem;
-		      rules[np].rhs = ritem + ni;
-		      while (ritem[pi] >= 0)
-			ritem[ni++] = ritem[pi++];
-		      ritem[ni++] = -np;
-		    }
-		}
-	      else
-		{
-		  while (ritem[ni++] >= 0)
-		    /* Nothing. */;
-		}
-	    }
-
-	ritem[ni] = 0;
-	nrules -= nuseless_productions;
-	nitems = ni;
-	nritems = ni;
-
-	/* Is it worth it to reduce the amount of memory for the
-	   grammar? Probably not.  */
+	short *rhsp = rules[i].rhs;
+	for (/* Nothing. */; *rhsp >= 0; ++rhsp)
+	  /* Nothing. */;
+	*rhsp = -i;
+	rules[i].number = i;
       }
+    nrules -= nuseless_productions;
+  }
 
-  /* Disable useless productions. */
-  if (nuseless_productions > 0)
-    {
-      int pn;
-      for (pn = 1; pn < nrules + 1; pn++)
-	rules[pn].useful = bitset_test (P, pn);
-    }
+  /* Adjust NRITEMS and NITEMS.  */
+  {
+    int r;
+    int length;
+    for (r = nrules + 1; r < nrules + 1 + nuseless_productions; ++r)
+      {
+	length = rule_rhs_length (&rules[r]);
+	nritems -= length + 1;
+      }
+  }
 }
 
 
@@ -311,8 +298,10 @@ nonterminals_reduce (void)
 
   /* Shuffle elements of tables indexed by symbol number.  */
   {
-    bucket **symbols_sorted = XMALLOC (bucket *, nvars) - ntokens;
+    symbol_t **symbols_sorted = XMALLOC (symbol_t *, nvars) - ntokens;
 
+    for (i = ntokens; i < nsyms; i++)
+      symbols[i]->number = nontermmap[i];
     for (i = ntokens; i < nsyms; i++)
       symbols_sorted[nontermmap[i]] = symbols[i];
     for (i = ntokens; i < nsyms; i++)
@@ -320,16 +309,6 @@ nonterminals_reduce (void)
     free (symbols_sorted + ntokens);
   }
 
-  /* Replace all symbol numbers in valid data structures.  */
-
-  for (i = 1; i < nrules + 1; i++)
-    {
-      rules[i].lhs = nontermmap[rules[i].lhs];
-      if (ISVAR (rules[i].precsym))
-	/* Can this happen?  */
-	rules[i].precsym = nontermmap[rules[i].precsym];
-    }
-
   for (i = 0; i < nritems; ++i)
     if (ISVAR (ritem[i]))
       ritem[i] = nontermmap[ritem[i]];
@@ -355,7 +334,8 @@ reduce_output (FILE *out)
       int i;
       fprintf (out, "%s\n\n", _("Useless nonterminals:"));
       for (i = 0; i < nuseless_nonterminals; ++i)
-	fprintf (out, "   %s\n", symbols[nsyms + i]->tag);
+	fprintf (out, "   %s\n", quotearg_style (escape_quoting_style,
+						 symbols[nsyms + i]->tag));
       fputs ("\n\n", out);
     }
 
@@ -368,7 +348,8 @@ reduce_output (FILE *out)
 	  if (!b)
 	    fprintf (out, "%s\n\n", _("Terminals which are not used:"));
 	  b = TRUE;
-	  fprintf (out, "   %s\n", symbols[i]->tag);
+	  fprintf (out, "   %s\n", quotearg_style (escape_quoting_style,
+						   symbols[i]->tag));
 	}
     if (b)
       fputs ("\n\n", out);
@@ -378,16 +359,17 @@ reduce_output (FILE *out)
     {
       int i;
       fprintf (out, "%s\n\n", _("Useless rules:"));
-      for (i = 1; i < nrules + 1; i++)
-	if (!rules[i].useful)
-	  {
-	    rule r;
-	    fprintf (out, "#%-4d  ", i - 1);
-	    fprintf (out, "%s:", symbols[rules[i].lhs]->tag);
-	    for (r = rules[i].rhs; *r >= 0; r++)
-	      fprintf (out, " %s", symbols[*r]->tag);
-	    fputs (";\n", out);
-	  }
+      for (i = nrules + 1; i < nuseless_productions + nrules + 1; i++)
+	{
+	  rule r;
+	  fprintf (out, "#%-4d  ", rules[i].user_number - 1);
+	  fprintf (out, "%s:", quotearg_style (escape_quoting_style,
+					       rules[i].lhs->tag));
+	  for (r = rules[i].rhs; *r >= 0; r++)
+	    fprintf (out, " %s", quotearg_style (escape_quoting_style,
+						 symbols[*r]->tag));
+	  fputs (";\n", out);
+	}
       fputs ("\n\n", out);
     }
 }
@@ -400,18 +382,19 @@ dump_grammar (FILE *out)
 
   fprintf (out, "REDUCED GRAMMAR\n\n");
   fprintf (out,
-	   "ntokens = %d, nvars = %d, nsyms = %d, nrules = %d, nitems = %d\n\n",
-	   ntokens, nvars, nsyms, nrules, nitems);
+	   "ntokens = %d, nvars = %d, nsyms = %d, nrules = %d, nritems = %d\n\n",
+	   ntokens, nvars, nsyms, nrules, nritems);
   fprintf (out, "Variables\n---------\n\n");
   fprintf (out, "Value  Sprec  Sassoc  Tag\n");
   for (i = ntokens; i < nsyms; i++)
     fprintf (out, "%5d  %5d   %5d  %s\n",
 	     i,
-	     symbols[i]->prec, symbols[i]->assoc, symbols[i]->tag);
+	     symbols[i]->prec, symbols[i]->assoc,
+	     quotearg_style (escape_quoting_style, symbols[i]->tag));
   fprintf (out, "\n\n");
   fprintf (out, "Rules\n-----\n\n");
   fprintf (out, "Num (Prec, Assoc, Useful, Ritem Range) Lhs -> Rhs (Ritem range) [Num]\n");
-  for (i = 1; i < nrules + 1; i++)
+  for (i = 1; i < nrules + nuseless_productions + 1; i++)
     {
       int rhs_count = 0;
       /* Find the last RHS index in ritems. */
@@ -419,9 +402,12 @@ dump_grammar (FILE *out)
 	++rhs_count;
       fprintf (out, "%3d (%2d, %2d, %2d, %2d-%2d)   %2d ->",
 	       i - 1,
-	       rules[i].prec, rules[i].assoc, rules[i].useful,
-	       rules[i].rhs - ritem, rules[i].rhs - ritem + rhs_count - 1,
-	       rules[i].lhs);
+	       rules[i].prec->prec,
+	       rules[i].prec->assoc,
+	       rules[i].useful,
+	       rules[i].rhs - ritem,
+	       rules[i].rhs - ritem + rhs_count - 1,
+	       rules[i].lhs->number);
       /* Dumped the RHS. */
       for (r = rules[i].rhs; *r >= 0; r++)
 	fprintf (out, "%3d", *r);
@@ -429,11 +415,13 @@ dump_grammar (FILE *out)
     }
   fprintf (out, "\n\n");
   fprintf (out, "Rules interpreted\n-----------------\n\n");
-  for (i = 1; i < nrules + 1; i++)
+  for (i = 1; i < nrules + nuseless_productions + 1; i++)
     {
-      fprintf (out, "%-5d  %s :", i, symbols[rules[i].lhs]->tag);
+      fprintf (out, "%-5d  %s :",
+	       i, quotearg_style (escape_quoting_style, rules[i].lhs->tag));
       for (r = rules[i].rhs; *r >= 0; r++)
-	fprintf (out, " %s", symbols[*r]->tag);
+	fprintf (out, " %s",
+		 quotearg_style (escape_quoting_style, symbols[*r]->tag));
       fputc ('\n', out);
     }
   fprintf (out, "\n\n");
@@ -497,9 +485,10 @@ reduce_grammar (void)
 
   if (!bitset_test (N, start_symbol - ntokens))
     fatal (_("Start symbol %s does not derive any sentence"),
-	   symbols[start_symbol]->tag);
+	   quotearg_style (escape_quoting_style, symbols[start_symbol]->tag));
 
-  reduce_grammar_tables ();
+  if (nuseless_productions > 0)
+    reduce_grammar_tables ();
   if (nuseless_nonterminals > 0)
     nonterminals_reduce ();