#include "system.h"
#include "getargs.h"
#include "files.h"
+#include "symtab.h"
#include "gram.h"
#include "complain.h"
#include "reduce.h"
typedef short *rule;
-/* 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. */
+/* Set of all nonterminals which are not useless. */
+static BSet N;
-static BSet N, P, V, V1;
+/* Set of all rules which have no useless nonterminals in their RHS. */
+static BSet P;
+
+/* Set of all accessible symbols. */
+static BSet V;
+
+/* Set of symbols used to define rule precedence (so they are
+ `useless', but no warning should be issued). */
+static BSet V1;
static int nuseful_productions;
static int nuseless_productions;
static int nuseful_nonterminals;
-static int nuseless_nonterminals;
+int nuseless_nonterminals;
\f
static bool
bits_equal (BSet L, BSet R, int n)
/* A production is useful if all of the nonterminals in its appear
in the set of useful nonterminals. */
- for (r = &ritem[rrhs[i]]; *r > 0; r++)
+ for (r = &ritem[rules[i].rhs]; *r >= 0; r++)
if (ISVAR (n = *r))
if (!BITISSET (N0, n - ntokens))
return FALSE;
{
if (useful_production (i, N))
{
- SETBIT (Np, rlhs[i] - ntokens);
+ SETBIT (Np, rules[i].lhs - ntokens);
SETBIT (P, i);
}
}
Pp = XCALLOC (unsigned, WORDSIZE (nrules + 1));
/* 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);
-
- while (1)
+ if (BITISSET (N, start_symbol - ntokens))
{
- for (i = WORDSIZE (nsyms) - 1; i >= 0; i--)
- Vp[i] = V[i];
- for (i = 1; i <= nrules; i++)
+ SETBIT (V, start_symbol);
+
+ while (1)
{
- if (!BITISSET (Pp, i) && BITISSET (P, i) && BITISSET (V, rlhs[i]))
+ for (i = WORDSIZE (nsyms) - 1; i >= 0; i--)
+ Vp[i] = V[i];
+ for (i = 1; i <= nrules; i++)
{
- for (r = &ritem[rrhs[i]]; *r >= 0; r++)
+ if (!BITISSET (Pp, i)
+ && BITISSET (P, i)
+ && BITISSET (V, rules[i].lhs))
{
- if (ISTOKEN (t = *r) || BITISSET (N, t - ntokens))
- {
+ for (r = &ritem[rules[i].rhs]; *r >= 0; r++)
+ if (ISTOKEN (t = *r) || BITISSET (N, t - ntokens))
SETBIT (Vp, t);
- }
+ SETBIT (Pp, i);
}
- SETBIT (Pp, i);
}
+ if (bits_equal (V, Vp, WORDSIZE (nsyms)))
+ break;
+ Vs = Vp;
+ Vp = V;
+ V = Vs;
}
- if (bits_equal (V, Vp, WORDSIZE (nsyms)))
- {
- break;
- }
- Vs = Vp;
- Vp = V;
- V = Vs;
}
-end_iteration:
XFREE (V);
V = Vp;
/* 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]);
+ if (rules[i].precsym != 0)
+ SETBIT (V1, rules[i].precsym);
}
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. */
-#if 0
- /* remove useless productions */
- if (nuseless_productions > 0)
- {
- short np, pn, ni, pi;
+ /* This is turned off because we would need to change the numbers in
+ the case statements in the actions file.
- np = 0;
- ni = 0;
- for (pn = 1; pn <= nrules; pn++)
- {
+ 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.
+
+ 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. */
+
+ 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)
+ 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 != ni)
{
- pi = rrhs[np];
- rrhs[np] = ni;
+ pi = rules[np].rhs;
+ rules[np].rhs = ni;
while (ritem[pi] >= 0)
ritem[ni++] = ritem[pi++];
ritem[ni++] = -np;
}
else
{
- while (ritem[ni++] >= 0);
+ while (ritem[ni++] >= 0)
+ /* Nothing. */;
}
}
- }
- ritem[ni] = 0;
- nrules -= nuseless_productions;
- nitems = ni;
- /* Is it worth it to reduce the amount of memory for the
- grammar? Probably not. */
+ ritem[ni] = 0;
+ nrules -= nuseless_productions;
+ nitems = ni;
+ nritems = ni;
- }
-#endif /* 0 */
- /* Disable useless productions,
- since they may contain useless nonterms
- that would get mapped below to -1 and confuse everyone. */
+ /* Is it worth it to reduce the amount of memory for the
+ grammar? Probably not. */
+ }
+
+ /* Disable useless productions. */
if (nuseless_productions > 0)
{
int pn;
-
for (pn = 1; pn <= nrules; pn++)
- {
- if (!BITISSET (P, pn))
- {
- rlhs[pn] = -1;
- }
- }
+ rules[pn].useful = BITISSET (P, pn);
}
+}
- /* remove useless symbols */
- if (nuseless_nonterminals > 0)
- {
- int i, n;
-/* short j; JF unused */
- short *nontermmap;
- rule r;
+/*------------------------------.
+| Remove useless nonterminals. |
+`------------------------------*/
+
+static void
+nonterminals_reduce (void)
+{
+ int i, n;
- /* Create a map of nonterminal number to new nonterminal
- number. -1 in the map means it was useless and is being
- eliminated. */
+ /* Map the nonterminals to their new index: useful first, useless
+ afterwards. Kept for later report. */
- nontermmap = XCALLOC (short, nvars) - ntokens;
- for (i = ntokens; i < nsyms; i++)
- nontermmap[i] = -1;
+ short *nontermmap = XCALLOC (short, nvars) - ntokens;
+ n = ntokens;
+ for (i = ntokens; i < nsyms; i++)
+ if (BITISSET (V, i))
+ nontermmap[i] = n++;
+ for (i = ntokens; i < nsyms; i++)
+ if (!BITISSET (V, i))
+ nontermmap[i] = n++;
- n = ntokens;
- for (i = ntokens; i < nsyms; i++)
- if (BITISSET (V, i))
- nontermmap[i] = n++;
- /* Shuffle elements of tables indexed by symbol number. */
+ /* Shuffle elements of tables indexed by symbol number. */
+ {
+ bucket **symbols_sorted = XMALLOC (bucket *, nvars) - ntokens;
- 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]);
- }
- }
+ for (i = ntokens; i < nsyms; i++)
+ symbols_sorted[nontermmap[i]] = symbols[i];
+ for (i = ntokens; i < nsyms; i++)
+ symbols[i] = symbols_sorted[i];
+ free (symbols_sorted + ntokens);
+ }
- /* Replace all symbol numbers in valid data structures. */
+ /* Replace all symbol numbers in valid data structures. */
- 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]];
- }
+ for (i = 1; i <= nrules; i++)
+ {
+ rules[i].lhs = nontermmap[rules[i].lhs];
+ if (ISVAR (rules[i].precsym))
+ /* Can this happen? */
+ rules[i].precsym = nontermmap[rules[i].precsym];
+ }
- for (r = ritem; *r; r++)
- if (ISVAR (*r))
- *r = nontermmap[*r];
+ for (i = 0; i < nritems; ++i)
+ if (ISVAR (ritem[i]))
+ ritem[i] = nontermmap[ritem[i]];
- start_symbol = nontermmap[start_symbol];
+ start_symbol = nontermmap[start_symbol];
- nsyms -= nuseless_nonterminals;
- nvars -= nuseless_nonterminals;
+ nsyms -= nuseless_nonterminals;
+ nvars -= nuseless_nonterminals;
- free (&nontermmap[ntokens]);
- }
+ free (nontermmap + ntokens);
}
-/*-----------------------------------------------------------------.
-| Ouput the detailed results of the reductions. For FILE.output. |
-`-----------------------------------------------------------------*/
+/*------------------------------------------------------------------.
+| Output the detailed results of the reductions. For FILE.output. |
+`------------------------------------------------------------------*/
void
reduce_output (FILE *out)
{
- int i;
- rule r;
- bool b;
-
if (nuseless_nonterminals > 0)
{
- fprintf (out, _("Useless nonterminals:"));
- fprintf (out, "\n\n");
- for (i = ntokens; i < nsyms; i++)
- if (!BITISSET (V, i))
- fprintf (out, " %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++)
- {
+
+ {
+ bool b = FALSE;
+ int i;
+ for (i = 0; i < ntokens; i++)
if (!BITISSET (V, i) && !BITISSET (V1, i))
{
if (!b)
- {
- fprintf (out, "\n\n");
- fprintf (out, _("Terminals which are not used:"));
- fprintf (out, "\n\n");
- b = TRUE;
- }
- fprintf (out, " %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)
{
- fprintf (out, "\n\n");
- fprintf (out, _("Useless rules:"));
- fprintf (out, "\n\n");
+ int i;
+ fprintf (out, "%s\n\n", _("Useless rules:"));
for (i = 1; i <= nrules; i++)
- {
- if (!BITISSET (P, i))
- {
- fprintf (out, "#%-4d ", i);
- fprintf (out, "%s :\t", tags[rlhs[i]]);
- for (r = &ritem[rrhs[i]]; *r >= 0; r++)
- fprintf (out, " %s", tags[*r]);
- fprintf (out, ";\n");
- }
- }
+ if (!rules[i].useful)
+ {
+ rule r;
+ fprintf (out, "#%-4d ", i - 1);
+ fprintf (out, "%s:", symbols[rules[i].lhs]->tag);
+ for (r = &ritem[rules[i].rhs]; *r >= 0; r++)
+ fprintf (out, " %s", symbols[*r]->tag);
+ fputs (";\n", out);
+ }
+ fputs ("\n\n", out);
}
- if (nuseless_nonterminals > 0 || nuseless_productions > 0 || b)
- fprintf (out, "\n\n");
}
\f
-#if 0 /* XXX currently unused. */
static void
dump_grammar (FILE *out)
{
int i;
rule r;
+ 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);
- fprintf (out, _("Variables\n---------\n\n"));
- fprintf (out, _("Value Sprec Sassoc Tag\n"));
+ "ntokens = %d, nvars = %d, nsyms = %d, nrules = %d, nitems = %d\n\n",
+ ntokens, nvars, nsyms, nrules, nitems);
+ 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, sprec[i], sassoc[i], tags[i]);
+ fprintf (out, "%5d %5d %5d %s\n",
+ i,
+ symbols[i]->prec, symbols[i]->assoc, symbols[i]->tag);
fprintf (out, "\n\n");
- fprintf (out, _("Rules\n-----\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; i++)
{
- fprintf (out, "%-5d(%5d%5d)%5d : (@%-5d)",
- i, rprec[i], rassoc[i], rlhs[i], rrhs[i]);
- for (r = &ritem[rrhs[i]]; *r > 0; r++)
- fprintf (out, "%5d", *r);
- fprintf (out, " [%d]\n", -(*r));
+ int rhs_count = 0;
+ /* Find the last RHS index in ritems. */
+ for (r = &ritem[rules[i].rhs]; *r >= 0; ++r)
+ ++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, rules[i].rhs + rhs_count - 1,
+ rules[i].lhs);
+ /* Dumped the RHS. */
+ for (r = &ritem[rules[i].rhs]; *r >= 0; r++)
+ fprintf (out, "%3d", *r);
+ fprintf (out, " [%d]\n", -(*r) - 1);
}
fprintf (out, "\n\n");
- fprintf (out, _("Rules interpreted\n-----------------\n\n"));
+ fprintf (out, "Rules interpreted\n-----------------\n\n");
for (i = 1; i <= nrules; i++)
{
- fprintf (out, "%-5d %s :", i, tags[rlhs[i]]);
- for (r = &ritem[rrhs[i]]; *r > 0; r++)
- fprintf (out, " %s", tags[*r]);
+ fprintf (out, "%-5d %s :", i, symbols[rules[i].lhs]->tag);
+ for (r = &ritem[rules[i].rhs]; *r >= 0; r++)
+ fprintf (out, " %s", symbols[*r]->tag);
fputc ('\n', out);
}
fprintf (out, "\n\n");
}
-#endif
reduce_print (void)
{
if (yacc_flag && nuseless_productions)
- fprintf (stderr, _("%d rules never reduced\n"), nuseless_productions);
+ fprintf (stderr, ngettext ("%d rule never reduced\n",
+ "%d rules never reduced\n",
+ nuseless_productions),
+ nuseless_productions);
fprintf (stderr, _("%s contains "), infile);
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)
fprintf (stderr, _(" and "));
if (nuseless_productions > 0)
- {
- fprintf (stderr, _("%d useless rule%s"),
- nuseless_productions, (nuseless_productions == 1 ? "" : "s"));
- }
+ fprintf (stderr, ngettext ("%d useless rule",
+ "%d useless rules",
+ nuseless_productions),
+ nuseless_productions);
fprintf (stderr, "\n");
fflush (stderr);
}
if (!BITISSET (N, start_symbol - ntokens))
fatal (_("Start symbol %s does not derive any sentence"),
- tags[start_symbol]);
+ symbols[start_symbol]->tag);
reduce_grammar_tables ();
-#if 0
- if (verbose_flag)
+ if (nuseless_nonterminals > 0)
+ nonterminals_reduce ();
+
+ if (trace_flag)
{
- fprintf (out, "REDUCED GRAMMAR\n\n");
- dump_grammar ();
+ dump_grammar (stderr);
+
+ fprintf (stderr, "reduced %s defines %d terminals, %d nonterminals\
+, and %d productions.\n",
+ infile, ntokens, nvars, nrules);
}
-#endif
-
- if (statistics_flag)
- 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"));
}