#include "system.h"
#include "getargs.h"
#include "files.h"
+#include "symtab.h"
#include "gram.h"
#include "complain.h"
#include "reduce.h"
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[rule_table[i].rhs]; *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, rule_table[i].lhs - ntokens);
+ SETBIT (Np, rules[i].lhs - ntokens);
SETBIT (P, i);
}
}
{
if (!BITISSET (Pp, i)
&& BITISSET (P, i)
- && BITISSET (V, rule_table[i].lhs))
+ && BITISSET (V, rules[i].lhs))
{
- for (r = &ritem[rule_table[i].rhs]; *r >= 0; r++)
+ for (r = &ritem[rules[i].rhs]; *r >= 0; r++)
if (ISTOKEN (t = *r) || BITISSET (N, t - ntokens))
SETBIT (Vp, t);
SETBIT (Pp, i);
/* A token that was used in %prec should not be warned about. */
for (i = 1; i < nrules; i++)
- if (rule_table[i].precsym != 0)
- SETBIT (V1, rule_table[i].precsym);
+ 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)
{
- rule_table[np].lhs = rule_table[pn].lhs;
- rline[np] = rline[pn];
- rule_table[np].prec = rule_table[pn].prec;
- rule_table[np].assoc = rule_table[pn].assoc;
- rule_table[np].rhs = rule_table[pn].rhs;
- if (rule_table[np].rhs != 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 = rule_table[np].rhs;
- rule_table[np].rhs = 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))
- {
- rule_table[pn].lhs = -1;
- }
- }
+ rules[pn].useful = BITISSET (P, pn);
}
}
nonterminals_reduce (void)
{
int i, n;
- rule r;
- /* 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. */
short *nontermmap = XCALLOC (short, nvars) - ntokens;
n = ntokens;
/* Shuffle elements of tables indexed by symbol number. */
{
- short *sassoc_sorted = XMALLOC (short, nvars) - ntokens;
- short *sprec_sorted = XMALLOC (short, nvars) - ntokens;
- char **tags_sorted = XMALLOC (char *, nvars) - ntokens;
+ bucket **symbols_sorted = XMALLOC (bucket *, nvars) - ntokens;
for (i = ntokens; i < nsyms; i++)
- {
- n = nontermmap[i];
- sassoc_sorted[n] = sassoc[i];
- sprec_sorted[n] = sprec[i];
- tags_sorted[n] = tags[i];
- }
+ symbols_sorted[nontermmap[i]] = symbols[i];
for (i = ntokens; i < nsyms; i++)
- {
- sassoc[i] = sassoc_sorted[i];
- sprec[i] = sprec_sorted[i];
- tags[i] = tags_sorted[i];
- }
- free (sassoc_sorted + ntokens);
- free (sprec_sorted + ntokens);
- free (tags_sorted + ntokens);
+ symbols[i] = symbols_sorted[i];
+ free (symbols_sorted + ntokens);
}
/* Replace all symbol numbers in valid data structures. */
for (i = 1; i <= nrules; i++)
{
- /* Ignore the rules disabled above. */
- if (rule_table[i].lhs >= 0)
- rule_table[i].lhs = nontermmap[rule_table[i].lhs];
- if (ISVAR (rule_table[i].precsym))
+ rules[i].lhs = nontermmap[rules[i].lhs];
+ if (ISVAR (rules[i].precsym))
/* Can this happen? */
- rule_table[i].precsym = nontermmap[rule_table[i].precsym];
+ 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];
nsyms -= nuseless_nonterminals;
nvars -= nuseless_nonterminals;
- free (&nontermmap[ntokens]);
+ free (nontermmap + ntokens);
}
void
reduce_output (FILE *out)
{
- int i;
- rule r;
- bool b;
-
if (nuseless_nonterminals > 0)
{
- fprintf (out, _("Useless nonterminals:"));
- fprintf (out, "\n\n");
+ int i;
+ fprintf (out, "%s\n\n", _("Useless nonterminals:"));
for (i = 0; i < nuseless_nonterminals; ++i)
- fprintf (out, " %s\n", tags[nsyms + 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))
+ if (!rules[i].useful)
{
- fprintf (out, "#%-4d ", i);
- fprintf (out, "%s :\t", tags[rule_table[i].lhs]);
- for (r = &ritem[rule_table[i].rhs]; *r >= 0; r++)
- fprintf (out, " %s", tags[*r]);
- fprintf (out, ";\n");
+ 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
static void
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, "Num (Prec, Assoc) Lhs : (@Rhs) Ritems [Num?]\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,
- rule_table[i].prec,
- rule_table[i].assoc,
- rule_table[i].lhs,
- rule_table[i].rhs);
- for (r = &ritem[rule_table[i].rhs]; *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");
for (i = 1; i <= nrules; i++)
{
- fprintf (out, "%-5d %s :", i, tags[rule_table[i].lhs]);
- for (r = &ritem[rule_table[i].rhs]; *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");
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 (nuseless_nonterminals > 0)