tokens they accept. */
#include "system.h"
+#include "bitset.h"
+#include "reader.h"
#include "types.h"
#include "LR0.h"
+#include "symtab.h"
#include "gram.h"
#include "complain.h"
#include "lalr.h"
#include "getargs.h"
/* All the decorated states, indexed by the state number. */
-state_t **state_table = NULL;
+state_t **states = NULL;
int tokensetsize;
-short *LAruleno;
-unsigned *LA;
+short *LAruleno = NULL;
+bitset *LA = NULL;
size_t nLA;
static int ngotos;
-short *goto_map;
-short *from_state;
-short *to_state;
+short *goto_map = NULL;
+short *from_state = NULL;
+short *to_state = NULL;
/* And for the famous F variable, which name is so descriptive that a
comment is hardly needed. <grin>. */
-static unsigned *F = NULL;
-#define F(Rule) (F + (Rule) * tokensetsize)
+static bitset *F = NULL;
static short **includes;
static shorts **lookback;
traverse (int i)
{
int j;
- size_t k;
int height;
- size_t size = F (i + 1) - F(i);
VERTICES[++top] = i;
INDEX[i] = height = top;
if (INDEX[i] > INDEX[R[i][j]])
INDEX[i] = INDEX[R[i][j]];
- for (k = 0; k < size; ++k)
- F (i)[k] |= F (R[i][j])[k];
+ bitset_or (F[i], F[i], F[R[i][j]]);
}
if (INDEX[i] == height)
if (i == j)
break;
- for (k = 0; k < size; ++k)
- F (j)[k] = F (i)[k];
+ bitset_copy (F[j], F[i]);
}
}
static void
initialize_LA (void)
{
- int i;
+ size_t i;
int j;
short *np;
- reductions *rp;
/* Avoid having to special case 0. */
if (!nLA)
nLA = 1;
- LA = XCALLOC (unsigned, nLA * tokensetsize);
+ LA = XCALLOC (bitset, nLA);
+ for (i = 0; i < nLA; ++i)
+ {
+ LA[i] = bitset_create (ntokens, BITSET_FIXED);
+ bitset_zero (LA[i]);
+ }
LAruleno = XCALLOC (short, nLA);
lookback = XCALLOC (shorts *, nLA);
np = LAruleno;
for (i = 0; i < nstates; i++)
- if (!state_table[i]->consistent)
- if ((rp = state_table[i]->reductions))
- for (j = 0; j < rp->nreds; j++)
- *np++ = rp->rules[j];
+ if (!states[i]->consistent)
+ for (j = 0; j < states[i]->reductions->nreds; j++)
+ *np++ = states[i]->reductions->rules[j];
}
static void
set_goto_map (void)
{
- int state;
+ size_t state;
int i;
- int symbol;
- int k;
short *temp_map;
- int state2;
goto_map = XCALLOC (short, nvars + 1) - ntokens;
temp_map = XCALLOC (short, nvars + 1) - ntokens;
ngotos = 0;
for (state = 0; state < nstates; ++state)
{
- shifts *sp = state_table[state]->shifts;
+ shifts *sp = states[state]->shifts;
for (i = sp->nshifts - 1; i >= 0 && SHIFT_IS_GOTO (sp, i); --i)
{
- symbol = state_table[sp->shifts[i]]->accessing_symbol;
-
if (ngotos == MAXSHORT)
fatal (_("too many gotos (max %d)"), MAXSHORT);
ngotos++;
- goto_map[symbol]++;
+ goto_map[SHIFT_SYMBOL (sp, i)]++;
}
}
- k = 0;
- for (i = ntokens; i < nsyms; i++)
- {
- temp_map[i] = k;
- k += goto_map[i];
- }
+ {
+ int k = 0;
+ for (i = ntokens; i < nsyms; i++)
+ {
+ temp_map[i] = k;
+ k += goto_map[i];
+ }
- for (i = ntokens; i < nsyms; i++)
- goto_map[i] = temp_map[i];
+ for (i = ntokens; i < nsyms; i++)
+ goto_map[i] = temp_map[i];
- goto_map[nsyms] = ngotos;
- temp_map[nsyms] = ngotos;
+ goto_map[nsyms] = ngotos;
+ temp_map[nsyms] = ngotos;
+ }
from_state = XCALLOC (short, ngotos);
to_state = XCALLOC (short, ngotos);
for (state = 0; state < nstates; ++state)
{
- shifts *sp = state_table[state]->shifts;
+ shifts *sp = states[state]->shifts;
for (i = sp->nshifts - 1; i >= 0 && SHIFT_IS_GOTO (sp, i); --i)
{
- for (i = sp->nshifts - 1; i >= 0 && SHIFT_IS_GOTO (sp, i); --i)
- {
- state2 = sp->shifts[i];
- symbol = state_table[state2]->accessing_symbol;
-
- k = temp_map[symbol]++;
- from_state[k] = state;
- to_state[k] = state2;
- }
+ int k = temp_map[SHIFT_SYMBOL (sp, i)]++;
+ from_state[k] = state;
+ to_state[k] = sp->shifts[i];
}
}
int i;
- F = XCALLOC (unsigned, ngotos * tokensetsize);
+ F = XCALLOC (bitset, ngotos);
+ for (i = 0; i < ngotos; ++i)
+ {
+ F[i] = bitset_create (ntokens, BITSET_FIXED);
+ bitset_zero (F[i]);
+ }
for (i = 0; i < ngotos; i++)
{
int stateno = to_state[i];
- shifts *sp = state_table[stateno]->shifts;
+ shifts *sp = states[stateno]->shifts;
int j;
for (j = 0; j < sp->nshifts && SHIFT_IS_SHIFT (sp, j); j++)
- {
- int symbol = state_table[sp->shifts[j]]->accessing_symbol;
- SETBIT (F (i), symbol);
- }
+ bitset_set (F[i], SHIFT_SYMBOL (sp, j));
for (; j < sp->nshifts; j++)
{
- int symbol = state_table[sp->shifts[j]]->accessing_symbol;
+ int symbol = SHIFT_SYMBOL (sp, j);
if (nullable[symbol])
edge[nedges++] = map_goto (stateno, symbol);
}
static void
-add_lookback_edge (int stateno, int ruleno, int gotono)
+add_lookback_edge (state_t *state, int ruleno, int gotono)
{
int i;
shorts *sp;
- for (i = 0; i < state_table[stateno]->nlookaheads; ++i)
- if (LAruleno[state_table[stateno]->lookaheadsp + i] == ruleno)
+ for (i = 0; i < state->nlookaheads; ++i)
+ if (LAruleno[state->lookaheadsp + i] == ruleno)
break;
- assert (LAruleno[state_table[stateno]->lookaheadsp + i] == ruleno);
+ assert (LAruleno[state->lookaheadsp + i] == ruleno);
sp = XCALLOC (shorts, 1);
- sp->next = lookback[state_table[stateno]->lookaheadsp + i];
+ sp->next = lookback[state->lookaheadsp + i];
sp->value = gotono;
- lookback[state_table[stateno]->lookaheadsp + i] = sp;
+ lookback[state->lookaheadsp + i] = sp;
}
build_relations (void)
{
short *edge = XCALLOC (short, ngotos + 1);
- short *states = XCALLOC (short, ritem_longest_rhs () + 1);
+ short *states1 = XCALLOC (short, ritem_longest_rhs () + 1);
int i;
includes = XCALLOC (short *, ngotos);
for (i = 0; i < ngotos; i++)
{
int nedges = 0;
- int state1 = from_state[i];
- int symbol1 = state_table[to_state[i]]->accessing_symbol;
+ int symbol1 = states[to_state[i]]->accessing_symbol;
short *rulep;
for (rulep = derives[symbol1]; *rulep > 0; rulep++)
{
int done;
int length = 1;
- int stateno = state1;
short *rp;
- states[0] = state1;
+ state_t *state = states[from_state[i]];
+ states1[0] = state->number;
- for (rp = ritem + rule_table[*rulep].rhs; *rp > 0; rp++)
+ for (rp = &ritem[rules[*rulep].rhs]; *rp >= 0; rp++)
{
- shifts *sp = state_table[stateno]->shifts;
+ shifts *sp = state->shifts;
int j;
for (j = 0; j < sp->nshifts; j++)
{
- stateno = sp->shifts[j];
- if (state_table[stateno]->accessing_symbol == *rp)
+ state = states[sp->shifts[j]];
+ if (state->accessing_symbol == *rp)
break;
}
- states[length++] = stateno;
+ states1[length++] = state->number;
}
- if (!state_table[stateno]->consistent)
- add_lookback_edge (stateno, *rulep, i);
+ if (!state->consistent)
+ add_lookback_edge (state, *rulep, i);
length--;
done = 0;
/* JF added rp>=ritem && I hope to god its right! */
if (rp >= ritem && ISVAR (*rp))
{
- stateno = states[--length];
- edge[nedges++] = map_goto (stateno, *rp);
+ edge[nedges++] = map_goto (states1[--length], *rp);
if (nullable[*rp])
done = 0;
}
}
XFREE (edge);
- XFREE (states);
+ XFREE (states1);
includes = transpose (includes, ngotos);
}
for (i = 0; i < nLA; i++)
for (sp = lookback[i]; sp; sp = sp->next)
- {
- int size = LA (i + 1) - LA (i);
- int j;
- for (j = 0; j < size; ++j)
- LA (i)[j] |= F (sp->value)[j];
- }
+ bitset_or (LA[i], LA[i], F[sp->value]);
/* Free LOOKBACK. */
for (i = 0; i < nLA; i++)
LIST_FREE (shorts, lookback[i]);
XFREE (lookback);
+ for (i = 0; i < (unsigned) ngotos; ++i)
+ bitset_free (F[i]);
XFREE (F);
}
static void
initialize_lookaheads (void)
{
- int i;
+ size_t i;
nLA = 0;
for (i = 0; i < nstates; i++)
{
int k;
int nlookaheads = 0;
- reductions *rp = state_table[i]->reductions;
- shifts *sp = state_table[i]->shifts;
-
- if (rp
- && (rp->nreds > 1 || (sp->nshifts && SHIFT_IS_SHIFT (sp, 0))))
+ reductions *rp = states[i]->reductions;
+ shifts *sp = states[i]->shifts;
+
+ /* We need a lookahead either to distinguish different
+ reductions (i.e., there are two or more), or to distinguish a
+ reduction from a shift. Otherwise, it is straightforward,
+ and the state is `consistent'. */
+ if (rp->nreds > 1
+ || (rp->nreds == 1 && sp->nshifts && SHIFT_IS_SHIFT (sp, 0)))
nlookaheads += rp->nreds;
else
- state_table[i]->consistent = 1;
+ states[i]->consistent = 1;
for (k = 0; k < sp->nshifts; k++)
if (SHIFT_IS_ERROR (sp, k))
{
- state_table[i]->consistent = 0;
+ states[i]->consistent = 0;
break;
}
- state_table[i]->nlookaheads = nlookaheads;
- state_table[i]->lookaheadsp = nLA;
+ states[i]->nlookaheads = nlookaheads;
+ states[i]->lookaheadsp = nLA;
nLA += nlookaheads;
}
}
+
+/*---------------------------------------.
+| Output the lookaheads for each state. |
+`---------------------------------------*/
+
+static void
+lookaheads_print (FILE *out)
+{
+ size_t i;
+ int j, k;
+ fprintf (out, "Lookaheads: BEGIN\n");
+ for (i = 0; i < nstates; ++i)
+ {
+ fprintf (out, "State %d: %d lookaheads\n",
+ i, states[i]->nlookaheads);
+
+ for (j = 0; j < states[i]->nlookaheads; ++j)
+ for (k = 0; k < ntokens; ++k)
+ if (bitset_test (LA[states[i]->lookaheadsp + j], j))
+ fprintf (out, " on %d (%s) -> rule %d\n",
+ k, symbols[k]->tag,
+ -LAruleno[states[i]->lookaheadsp + j] - 1);
+ }
+ fprintf (out, "Lookaheads: END\n");
+}
+
void
lalr (void)
{
build_relations ();
compute_FOLLOWS ();
compute_lookaheads ();
+
+ if (trace_flag)
+ lookaheads_print (stderr);
}