02111-1307, USA. */
-/* The parser tables consist of these tables. Marked ones needed only
- for the semantic parser. Double marked are output only if switches
- are set.
+/* The parser tables consist of these tables.
YYTRANSLATE = vector mapping yylex's token numbers into bison's
token numbers.
- ++ YYTNAME = vector of string-names indexed by bison token number.
+ YYTNAME = vector of string-names indexed by bison token number.
- ++ YYTOKNUM = vector of yylex token numbers corresponding to
- entries in YYTNAME.
+ YYTOKNUM = vector of yylex token numbers corresponding to entries
+ in YYTNAME.
YYRLINE = vector of line-numbers of all rules. For yydebug
printouts.
YYR2[R] = number of symbols composing right hand side of rule R.
- + YYSTOS[S] = the symbol number of the symbol that leads to state
- S.
+ YYSTOS[S] = the symbol number of the symbol that leads to state S.
YYDEFACT[S] = default rule to reduce with in state s, when YYTABLE
doesn't specify something else to do. Zero means the default is an
in a roundabout way, the bounds of the portion you are trying to
examine.
- Suppose that the portion of yytable starts at index P and the index
+ Suppose that the portion of YYTABLE starts at index P and the index
to be examined within the portion is I. Then if YYCHECK[P+I] != I,
I is outside the bounds of what is actually allocated, and the
default (from YYDEFACT or YYDEFGOTO) should be used. Otherwise,
#include "conflicts.h"
#include "muscle_tab.h"
-/* From lib/readpipe.h. */
-FILE *readpipe PARAMS ((const char *, ...));
-
/* From src/scan-skel.l. */
-int skel_lex PARAMS ((void));
-extern FILE *skel_in;
+void m4_invoke PARAMS ((const char *definitions));
+
+
+/* Several tables will be indexed both by state and nonterminal
+ numbers. We call `vector' such a thing (= either a state or a
+ symbol number.
+
+ Of course vector_number_t ought to be wide enough to contain
+ state_number_t and symbol_number_t. */
+typedef short vector_number_t;
+#define VECTOR_NUMBER_MAX ((vector_number_t) SHRT_MAX)
+#define VECTOR_NUMBER_MIN ((vector_number_t) SHRT_MIN)
+#define state_number_to_vector_number(State) \
+ ((vector_number_t) State)
+#define symbol_number_to_vector_number(Symbol) \
+ ((vector_number_t) (state_number_as_int (nstates) + Symbol - ntokens))
static int nvectors;
-static int nentries;
-static short **froms = NULL;
-static short **tos = NULL;
+
+
+/* FROMS and TOS are indexed by vector_number_t.
+
+ If VECTOR is a nonterminal, (FROMS[VECTOR], TOS[VECTOR]) form an
+ array of state numbers of the non defaulted GOTO on VECTOR.
+
+ If VECTOR is a state, TOS[VECTOR] is the array of actions to do on
+ the (array of) symbols FROMS[VECTOR].
+
+ In both cases, TALLY[VECTOR] is the size of the arrays
+ FROMS[VECTOR], TOS[VECTOR]; and WIDTH[VECTOR] =
+ (FROMS[VECTOR][SIZE] - FROMS[VECTOR][0] + 1) where SIZE =
+ TALLY[VECTOR].
+
+ FROMS therefore contains symbol_number_t and action_number_t,
+ TOS state_number_t and action_number_t,
+ TALLY sizes,
+ WIDTH differences of FROMS.
+
+ Let base_t be the type of FROMS, TOS, and WIDTH. */
+typedef int base_t;
+#define BASE_MAX ((base_t) INT_MAX)
+#define BASE_MIN ((base_t) INT_MIN)
+
+static base_t **froms = NULL;
+static base_t **tos = NULL;
static unsigned int **conflict_tos = NULL;
static short *tally = NULL;
-static short *width = NULL;
-static short *actrow = NULL;
-static short *conflrow = NULL;
-static short *state_count = NULL;
-static short *order = NULL;
-static short *base = NULL;
-static short *pos = NULL;
+static base_t *width = NULL;
+
+
+/* For a given state, N = ACTROW[SYMBOL]:
+ If N = 0, stands for `run the default action'.
+ If N = MIN, stands for `raise a parse error'.
+ If N > 0, stands for `shift SYMBOL and go to n'.
+ If N < 0, stands for `reduce -N'. */
+typedef short action_t;
+#define ACTION_MAX ((action_t) SHRT_MAX)
+#define ACTION_MIN ((action_t) SHRT_MIN)
+
+static action_t *actrow = NULL;
+
+/* FROMS and TOS are reordered to be compressed. ORDER[VECTOR] is the
+ new vector number of VECTOR. We skip `empty' vectors (i.e.,
+ TALLY[VECTOR] = 0), and call these `entries'. */
+static vector_number_t *order = NULL;
+static int nentries;
+
+static base_t *base = NULL;
+/* A distinguished value of BASE, negative infinite. During the
+ computation equals to BASE_MIN, later mapped to BASE_NINF to
+ keep parser tables small. */
+base_t base_ninf = 0;
+static base_t *pos = NULL;
+
+static unsigned int *conflrow = NULL;
static unsigned int *conflict_table = NULL;
static unsigned int *conflict_list = NULL;
static int conflict_list_cnt;
static int conflict_list_free;
-/* TABLE_SIZE is the allocated size of both TABLE and CHECK.
- We start with the original hard-coded value: SHRT_MAX
- (yes, not USHRT_MAX). */
-static size_t table_size = SHRT_MAX;
-static short *table = NULL;
-static short *check = NULL;
+/* TABLE_SIZE is the allocated size of both TABLE and CHECK. We start
+ with more or less the original hard-coded value (which was
+ SHRT_MAX). */
+static size_t table_size = 32768;
+static base_t *table = NULL;
+static base_t *check = NULL;
+/* The value used in TABLE to denote explicit parse errors
+ (%nonassoc), a negative infinite. First defaults to ACTION_MIN,
+ but in order to keep small tables, renumbered as TABLE_ERROR, which
+ is the smallest (non error) value minus 1. */
+base_t table_ninf = 0;
static int lowzero;
static int high;
while (table_size <= desired)
table_size *= 2;
- if (trace_flag)
+ if (trace_flag & trace_resource)
fprintf (stderr, "growing table and check from: %d to %d\n",
old_size, table_size);
- table = XREALLOC (table, short, table_size);
- check = XREALLOC (check, short, table_size);
+ table = XREALLOC (table, base_t, table_size);
+ check = XREALLOC (check, base_t, table_size);
if (glr_parser)
conflict_table = XREALLOC (conflict_table, unsigned int, table_size);
int begin, \
int end) \
{ \
+ Type min = first; \
Type max = first; \
int i; \
int j = 1; \
else \
++j; \
obstack_fgrow1 (&format_obstack, "%6d", table_data[i]); \
- if (table_data[i] > max) \
+ if (table_data[i] < min) \
+ min = table_data[i]; \
+ if (max < table_data[i]) \
max = table_data[i]; \
} \
obstack_1grow (&format_obstack, 0); \
muscle_insert (name, obstack_finish (&format_obstack)); \
\
- /* Build `NAME_max' in the obstack. */ \
+ /* Build `NAME_min' and `NAME_max' in the obstack. */ \
+ obstack_fgrow1 (&format_obstack, "%s_min", name); \
+ obstack_1grow (&format_obstack, 0); \
+ MUSCLE_INSERT_LONG_INT (obstack_finish (&format_obstack), \
+ (long int) min); \
obstack_fgrow1 (&format_obstack, "%s_max", name); \
obstack_1grow (&format_obstack, 0); \
MUSCLE_INSERT_LONG_INT (obstack_finish (&format_obstack), \
}
GENERATE_MUSCLE_INSERT_TABLE(muscle_insert_unsigned_int_table, unsigned int)
+GENERATE_MUSCLE_INSERT_TABLE(muscle_insert_int_table, int)
GENERATE_MUSCLE_INSERT_TABLE(muscle_insert_short_table, short)
+GENERATE_MUSCLE_INSERT_TABLE(muscle_insert_base_table, base_t)
+GENERATE_MUSCLE_INSERT_TABLE(muscle_insert_rule_number_table, rule_number_t)
GENERATE_MUSCLE_INSERT_TABLE(muscle_insert_symbol_number_table, symbol_number_t)
GENERATE_MUSCLE_INSERT_TABLE(muscle_insert_item_number_table, item_number_t)
GENERATE_MUSCLE_INSERT_TABLE(muscle_insert_state_number_table, state_number_t)
prepare_tokens (void)
{
muscle_insert_symbol_number_table ("translate",
- token_translations,
- 0, 1, max_user_token_number + 1);
+ token_translations,
+ token_translations[0],
+ 1, max_user_token_number + 1);
{
int i;
SYMBOL_TAG_GET uses slot 0. */
const char *cp =
quotearg_n_style (1, c_quoting_style,
- symbol_tag_get (symbols[i]));
+ symbols[i]->tag);
/* Width of the next token, including the two quotes, the coma
and the space. */
int strsize = strlen (cp) + 2;
muscle_insert ("tname", obstack_finish (&format_obstack));
}
- /* Output YYTOKNUM. */
+ /* Output YYTOKNUM. */
{
int i;
- short *values = XCALLOC (short, ntokens + 1);
- for (i = 0; i < ntokens + 1; ++i)
+ int *values = XCALLOC (int, ntokens);
+ for (i = 0; i < ntokens; ++i)
values[i] = symbols[i]->user_token_number;
- muscle_insert_short_table ("toknum", values,
- 0, 1, ntokens + 1);
+ muscle_insert_int_table ("toknum", values,
+ values[0], 1, ntokens);
free (values);
}
}
static void
prepare_rules (void)
{
- int r;
+ rule_number_t r;
unsigned int i = 0;
item_number_t *rhs = XMALLOC (item_number_t, nritems);
- unsigned int *prhs = XMALLOC (unsigned int, nrules + 1);
- unsigned int *rline = XMALLOC (unsigned int, nrules + 1);
- symbol_number_t *r1 = XMALLOC (symbol_number_t, nrules + 1);
- unsigned int *r2 = XMALLOC (unsigned int, nrules + 1);
- short *dprec = XMALLOC (short, nrules + 1);
- short *merger = XMALLOC (short, nrules + 1);
-
- for (r = 1; r < nrules + 1; ++r)
+ unsigned int *prhs = XMALLOC (unsigned int, nrules);
+ unsigned int *rline = XMALLOC (unsigned int, nrules);
+ symbol_number_t *r1 = XMALLOC (symbol_number_t, nrules);
+ unsigned int *r2 = XMALLOC (unsigned int, nrules);
+ short *dprec = XMALLOC (short, nrules);
+ short *merger = XMALLOC (short, nrules);
+
+ for (r = 0; r < nrules; ++r)
{
- item_number_t *rhsp;
+ item_number_t *rhsp = NULL;
/* Index of rule R in RHS. */
prhs[r] = i;
/* RHS of the rule R. */
assert (i == nritems);
muscle_insert_item_number_table ("rhs", rhs, ritem[0], 1, nritems);
- muscle_insert_unsigned_int_table ("prhs", prhs, 0, 1, nrules + 1);
- muscle_insert_unsigned_int_table ("rline", rline, 0, 1, nrules + 1);
- muscle_insert_symbol_number_table ("r1", r1, 0, 1, nrules + 1);
- muscle_insert_unsigned_int_table ("r2", r2, 0, 1, nrules + 1);
- muscle_insert_short_table ("dprec", dprec, 0, 1, nrules + 1);
- muscle_insert_short_table ("merger", merger, 0, 1, nrules + 1);
+ muscle_insert_unsigned_int_table ("prhs", prhs, 0, 0, nrules);
+ muscle_insert_unsigned_int_table ("rline", rline, 0, 0, nrules);
+ muscle_insert_symbol_number_table ("r1", r1, 0, 0, nrules);
+ muscle_insert_unsigned_int_table ("r2", r2, 0, 0, nrules);
+ muscle_insert_short_table ("dprec", dprec, 0, 0, nrules);
+ muscle_insert_short_table ("merger", merger, 0, 0, nrules);
free (rhs);
free (prhs);
/*-------------------------------------------------------------------.
| For GLR parsers, for each conflicted token in STATE, as indicated |
-| by non-zero entries in conflrow, create a list of possible |
+| by non-zero entries in CONFLROW, create a list of possible |
| reductions that are alternatives to the shift or reduction |
| currently recorded for that token in STATE. Store the alternative |
-| reductions followed by a 0 in conflict_list, updating |
-| conflict_list_cnt, and storing an index to the start of the list |
-| back into conflrow. |
+| reductions followed by a 0 in CONFLICT_LIST, updating |
+| CONFLICT_LIST_CNT, and storing an index to the start of the list |
+| back into CONFLROW. |
`-------------------------------------------------------------------*/
static void
{
conflrow[j] = conflict_list_cnt;
- /* find all reductions for token j, and record all that do
- * not match actrow[j] */
+ /* Find all reductions for token J, and record all that do not
+ match ACTROW[J]. */
for (i = 0; i < state->nlookaheads; i += 1)
if (bitset_test (state->lookaheads[i], j)
- && actrow[j] != -state->lookaheads_rule[i]->number)
+ && (actrow[j]
+ != rule_number_as_item_number (state->lookaheads_rule[i]->number)))
{
assert (conflict_list_free > 0);
conflict_list[conflict_list_cnt]
- = state->lookaheads_rule[i]->number;
+ = state->lookaheads_rule[i]->number + 1;
conflict_list_cnt += 1;
conflict_list_free -= 1;
}
- /* Leave a 0 at the end */
+ /* Leave a 0 at the end. */
assert (conflict_list_free > 0);
conflict_list_cnt += 1;
conflict_list_free -= 1;
/*------------------------------------------------------------------.
| Decide what to do for each type of token if seen as the lookahead |
| token in specified state. The value returned is used as the |
-| default action (yydefact) for the state. In addition, actrow is |
+| default action (yydefact) for the state. In addition, ACTROW is |
| filled with what to do for each kind of token, index by symbol |
| number, with zero meaning do the default action. The value |
-| SHRT_MIN, a very negative number, means this situation is an |
+| ACTION_MIN, a very negative number, means this situation is an |
| error. The parser recognizes this value specially. |
| |
| This is where conflicts are resolved. The loop over lookahead |
| rules considered lower-numbered rules last, and the last rule |
| considered that likes a token gets to handle it. |
-| |
-| For GLR parsers, also sets conflrow[SYM] to an index into |
-| conflict_list iff there is an unresolved conflict (s/r or r/r) |
+| |
+| For GLR parsers, also sets CONFLROW[SYM] to an index into |
+| CONFLICT_LIST iff there is an unresolved conflict (s/r or r/r) |
| with symbol SYM. The default reduction is not used for a symbol |
-| that has any such conflicts. |
+| that has any such conflicts. |
`------------------------------------------------------------------*/
-static int
+static rule_t *
action_row (state_t *state)
{
int i;
- int default_rule = 0;
- reductions *redp = state->reductions;
- shifts_t *shiftp = state->shifts;
- errs *errp = state->errs;
- /* set nonzero to inhibit having any default reduction */
+ rule_t *default_rule = NULL;
+ reductions_t *redp = state->reductions;
+ transitions_t *transitions = state->transitions;
+ errs_t *errp = state->errs;
+ /* Set to nonzero to inhibit having any default reduction. */
int nodefault = 0;
int conflicted = 0;
for (i = 0; i < ntokens; i++)
actrow[i] = conflrow[i] = 0;
- if (redp->nreds >= 1)
+ if (redp->num >= 1)
{
int j;
+ bitset_iterator biter;
/* loop over all the rules available here which require
lookahead */
for (i = state->nlookaheads - 1; i >= 0; --i)
/* and find each token which the rule finds acceptable
to come next */
- for (j = 0; j < ntokens; j++)
+ BITSET_FOR_EACH (biter, state->lookaheads[i], j, 0)
+ {
/* and record this rule as the rule to use if that
token follows. */
- if (bitset_test (state->lookaheads[i], j))
- {
- if (actrow[j] != 0)
- conflicted = conflrow[j] = 1;
- actrow[j] = -state->lookaheads_rule[i]->number;
- }
+ if (actrow[j] != 0)
+ conflicted = conflrow[j] = 1;
+ actrow[j] = rule_number_as_item_number (state->lookaheads_rule[i]->number);
+ }
}
/* Now see which tokens are allowed for shifts in this state. For
them, record the shift as the thing to do. So shift is preferred
to reduce. */
- for (i = 0; i < shiftp->nshifts; i++)
+ FOR_EACH_SHIFT (transitions, i)
{
- symbol_number_t symbol;
- state_number_t shift_state = shiftp->shifts[i];
- if (!shift_state)
- continue;
-
- symbol = states[shift_state]->accessing_symbol;
-
- if (ISVAR (symbol))
- break;
+ symbol_number_t symbol = TRANSITION_SYMBOL (transitions, i);
+ state_t *shift_state = transitions->states[i];
if (actrow[symbol] != 0)
conflicted = conflrow[symbol] = 1;
- actrow[symbol] = state_number_as_int (shift_state);
+ actrow[symbol] = state_number_as_int (shift_state->number);
/* Do not use any default reduction if there is a shift for
error */
}
/* See which tokens are an explicit error in this state (due to
- %nonassoc). For them, record SHRT_MIN as the action. */
- for (i = 0; i < errp->nerrs; i++)
+ %nonassoc). For them, record ACTION_MIN as the action. */
+ for (i = 0; i < errp->num; i++)
{
- int symbol = errp->errs[i];
- actrow[symbol] = SHRT_MIN;
+ symbol_t *symbol = errp->symbols[i];
+ actrow[symbol->number] = ACTION_MIN;
}
/* Now find the most common reduction and make it the default action
for this state. */
- if (redp->nreds >= 1 && !nodefault)
+ if (redp->num >= 1 && !nodefault)
{
if (state->consistent)
default_rule = redp->rules[0];
for (i = 0; i < state->nlookaheads; i++)
{
int count = 0;
- int rule = -state->lookaheads_rule[i]->number;
- int j;
+ rule_t *rule = state->lookaheads_rule[i];
+ symbol_number_t j;
for (j = 0; j < ntokens; j++)
- if (actrow[j] == rule)
+ if (actrow[j] == rule_number_as_item_number (rule->number))
count++;
if (count > max)
{
int j;
for (j = 0; j < ntokens; j++)
- if (actrow[j] == default_rule && ! (glr_parser && conflrow[j]))
+ if (actrow[j] == rule_number_as_item_number (default_rule->number)
+ && ! (glr_parser && conflrow[j]))
actrow[j] = 0;
}
- default_rule = -default_rule;
}
}
+ /* Find the rules which are reduced. */
+ if (!glr_parser)
+ {
+ for (i = 0; i < ntokens; i++)
+ if (actrow[i] < 0 && actrow[i] != ACTION_MIN)
+ rules[item_number_as_rule_number (actrow[i])].useful = TRUE;
+ if (default_rule)
+ default_rule->useful = TRUE;
+ }
+
/* If have no default rule, the default is an error.
So replace any action which says "error" with "use default". */
- if (default_rule == 0)
+ if (!default_rule)
for (i = 0; i < ntokens; i++)
- if (actrow[i] == SHRT_MIN)
+ if (actrow[i] == ACTION_MIN)
actrow[i] = 0;
if (conflicted)
}
+/*--------------------------------------------.
+| Set FROMS, TOS, TALLY and WIDTH for STATE. |
+`--------------------------------------------*/
+
static void
save_row (state_number_t state)
{
symbol_number_t i;
int count;
- short *sp = NULL;
- short *sp1 = NULL;
- short *sp2 = NULL;
+ base_t *sp = NULL;
+ base_t *sp1 = NULL;
+ base_t *sp2 = NULL;
unsigned int *sp3 = NULL;
+ /* Number of non default actions in STATE. */
count = 0;
for (i = 0; i < ntokens; i++)
if (actrow[i] != 0)
if (count == 0)
return;
- froms[state] = sp1 = sp = XCALLOC (short, count);
- tos[state] = sp2 = XCALLOC (short, count);
+ /* Allocate non defaulted actions. */
+ froms[state] = sp1 = sp = XCALLOC (base_t, count);
+ tos[state] = sp2 = XCALLOC (base_t, count);
if (glr_parser)
conflict_tos[state] = sp3 = XCALLOC (unsigned int, count);
else
conflict_tos[state] = NULL;
+ /* Store non defaulted actions. */
for (i = 0; i < ntokens; i++)
if (actrow[i] != 0)
{
token_actions (void)
{
state_number_t i;
+ rule_number_t r;
int nconflict = conflicts_total_count ();
- short *yydefact = XCALLOC (short, nstates);
+ rule_number_t *yydefact = XCALLOC (rule_number_t, nstates);
- actrow = XCALLOC (short, ntokens);
+ actrow = XCALLOC (action_t, ntokens);
+ conflrow = XCALLOC (unsigned int, ntokens);
+
+ /* Now that the parser was computed, we can find which rules are
+ really reduced, and which are not because of SR or RR conflicts.
+ */
+ if (!glr_parser)
+ for (r = 0; r < nrules; ++r)
+ rules[r].useful = FALSE;
- conflrow = XCALLOC (short, ntokens);
if (glr_parser)
{
conflict_list = XCALLOC (unsigned int, 1 + 2 * nconflict);
for (i = 0; i < nstates; ++i)
{
- yydefact[i] = action_row (states[i]);
+ rule_t *default_rule = action_row (states[i]);
+ yydefact[i] = default_rule ? default_rule->number + 1 : 0;
save_row (i);
}
- muscle_insert_short_table ("defact", yydefact,
- yydefact[0], 1, nstates);
+ muscle_insert_rule_number_table ("defact", yydefact,
+ yydefact[0], 1, nstates);
+
+ if (!glr_parser)
+ for (r = 0; r < nrules ; ++r)
+ if (!rules[r].useful)
+ {
+ LOCATION_PRINT (stderr, rules[r].location);
+ fprintf (stderr, ": %s: %s: ",
+ _("warning"), _("rule never reduced because of conflicts"));
+ rule_print (&rules[r], stderr);
+ }
+
XFREE (actrow);
XFREE (conflrow);
XFREE (yydefact);
void
actions_output (FILE *out)
{
- int rule;
+ rule_number_t r;
fputs ("m4_define([b4_actions], \n[[", out);
- for (rule = 1; rule < nrules + 1; ++rule)
- if (rules[rule].action)
+ for (r = 0; r < nrules; ++r)
+ if (rules[r].action)
{
- fprintf (out, " case %d:\n", rule);
+ fprintf (out, " case %d:\n", r + 1);
if (!no_lines_flag)
fprintf (out, muscle_find ("linef"),
- rules[rule].action_location.first_line,
+ rules[r].action_location.first_line,
quotearg_style (c_quoting_style,
muscle_find ("filename")));
fprintf (out, " %s\n break;\n\n",
- rules[rule].action);
+ rules[r].action);
}
fputs ("]])\n\n", out);
}
fprintf (out, "%s[[[%s]], [[%d]], [[%s]], [[%d]], [[%s]], [[%s]]]",
first ? "" : ",\n",
infile, symbol->destructor_location.first_line,
- symbol_tag_get (symbol),
+ symbol->tag,
symbol->number,
symbol->destructor,
symbol->type_name);
fprintf (out, "%s[[[%s]], [[%d]], [[%s]], [[%d]], [[%s]], [[%s]]]",
first ? "" : ",\n",
infile, symbol->printer_location.first_line,
- symbol_tag_get (symbol),
+ symbol->tag,
symbol->number,
symbol->printer,
symbol->type_name);
}
+/*------------------------------------------------------------------.
+| Compute FROMS[VECTOR], TOS[VECTOR], TALLY[VECTOR], WIDTH[VECTOR], |
+| i.e., the information related to non defaulted GOTO on the nterm |
+| SYMBOL. |
+| |
+| DEFAULT_STATE is the principal destination on SYMBOL, i.e., the |
+| default GOTO destination on SYMBOL. |
+`------------------------------------------------------------------*/
+
static void
save_column (symbol_number_t symbol, state_number_t default_state)
{
int i;
- short *sp;
- short *sp1;
- short *sp2;
+ base_t *sp;
+ base_t *sp1;
+ base_t *sp2;
int count;
- int symno = symbol - ntokens + state_number_as_int (nstates);
+ vector_number_t symno = symbol_number_to_vector_number (symbol);
- int begin = goto_map[symbol];
- int end = goto_map[symbol + 1];
+ goto_number_t begin = goto_map[symbol];
+ goto_number_t end = goto_map[symbol + 1];
+ /* Number of non default GOTO. */
count = 0;
for (i = begin; i < end; i++)
if (to_state[i] != default_state)
if (count == 0)
return;
- froms[symno] = sp1 = sp = XCALLOC (short, count);
- tos[symno] = sp2 = XCALLOC (short, count);
+ /* Allocate room for non defaulted gotos. */
+ froms[symno] = sp1 = sp = XCALLOC (base_t, count);
+ tos[symno] = sp2 = XCALLOC (base_t, count);
+ /* Store the state numbers of the non defaulted gotos. */
for (i = begin; i < end; i++)
if (to_state[i] != default_state)
{
}
+/*----------------------------------------------------------------.
+| Return `the' most common destination GOTO on SYMBOL (a nterm). |
+`----------------------------------------------------------------*/
+
static state_number_t
-default_goto (symbol_number_t symbol)
+default_goto (symbol_number_t symbol, short state_count[])
{
state_number_t s;
int i;
- int m = goto_map[symbol];
- int n = goto_map[symbol + 1];
+ goto_number_t m = goto_map[symbol];
+ goto_number_t n = goto_map[symbol + 1];
state_number_t default_state = (state_number_t) -1;
int max = 0;
goto_actions (void)
{
symbol_number_t i;
- state_number_t *yydefgoto = XMALLOC (state_number_t, nsyms - ntokens);
+ state_number_t *yydefgoto = XMALLOC (state_number_t, nvars);
- state_count = XCALLOC (short, nstates);
+ /* For a given nterm I, STATE_COUNT[S] is the number of times there
+ is a GOTO to S on I. */
+ short *state_count = XCALLOC (short, nstates);
for (i = ntokens; i < nsyms; ++i)
{
- state_number_t default_state = default_goto (i);
+ state_number_t default_state = default_goto (i, state_count);
save_column (i, default_state);
yydefgoto[i - ntokens] = default_state;
}
}
-/* The next few functions decide how to pack the actions and gotos
- information into yytable. */
+/*------------------------------------------------------------------.
+| Compute ORDER, a reordering of vectors, in order to decide how to |
+| pack the actions and gotos information into yytable. |
+`------------------------------------------------------------------*/
static void
sort_actions (void)
{
int i;
- order = XCALLOC (short, nvectors);
nentries = 0;
for (i = 0; i < nvectors; i++)
}
-static int
-matching_state (int vector)
+/* If VECTOR is a state which actions (reflected by FROMS, TOS, TALLY
+ and WIDTH of VECTOR) are common to a previous state, return this
+ state number.
+
+ In any other case, return -1. */
+
+static state_number_t
+matching_state (vector_number_t vector)
{
- int i = order[vector];
+ vector_number_t i = order[vector];
int t;
int w;
int prev;
+ /* If VECTOR is a nterm, return -1. */
if (i >= (int) nstates)
return -1;
for (prev = vector - 1; prev >= 0; prev--)
{
- int j = order[prev];
+ vector_number_t j = order[prev];
int k;
int match = 1;
+ /* Given how ORDER was computed, if the WIDTH or TALLY is
+ different, there cannot be a matching state. */
if (width[j] != w || tally[j] != t)
return -1;
}
-static int
-pack_vector (int vector)
+static base_t
+pack_vector (vector_number_t vector)
{
- int i = order[vector];
+ vector_number_t i = order[vector];
int j;
int t = tally[i];
int loc = 0;
- short *from = froms[i];
- short *to = tos[i];
+ base_t *from = froms[i];
+ base_t *to = tos[i];
unsigned int *conflict_to = conflict_tos[i];
assert (t);
{
for (k = 0; k < t; k++)
{
- loc = j + state_number_as_int (from[k]);
- table[loc] = state_number_as_int (to[k]);
+ loc = j + from[k];
+ table[loc] = to[k];
if (glr_parser && conflict_to != NULL)
conflict_table[loc] = conflict_to[k];
- check[loc] = state_number_as_int (from[k]);
+ check[loc] = from[k];
}
while (table[lowzero] != 0)
if (loc > high)
high = loc;
+ if (j < BASE_MIN || BASE_MAX < j)
+ fatal ("base_t too small to hold %d\n", j);
return j;
}
}
}
+/*-------------------------------------------------------------.
+| Remap the negative infinite in TAB from NINF to the greatest |
+| possible smallest value. Return it. |
+| |
+| In most case this allows us to use shorts instead of ints in |
+| parsers. |
+`-------------------------------------------------------------*/
+
+static base_t
+table_ninf_remap (base_t tab[], size_t size, base_t ninf)
+{
+ base_t res = 0;
+ size_t i;
+
+ for (i = 0; i < size; i++)
+ if (tab[i] < res && tab[i] != ninf)
+ res = base[i];
+
+ --res;
+
+ for (i = 0; i < size; i++)
+ if (tab[i] == ninf)
+ tab[i] = res;
+
+ return res;
+}
+
static void
pack_table (void)
{
int i;
- int place;
- int state;
- base = XCALLOC (short, nvectors);
- pos = XCALLOC (short, nentries);
- table = XCALLOC (short, table_size);
+ base = XCALLOC (base_t, nvectors);
+ pos = XCALLOC (base_t, nentries);
+ table = XCALLOC (base_t, table_size);
if (glr_parser)
conflict_table = XCALLOC (unsigned int, table_size);
- check = XCALLOC (short, table_size);
+ check = XCALLOC (base_t, table_size);
lowzero = 0;
high = 0;
for (i = 0; i < nvectors; i++)
- base[i] = SHRT_MIN;
+ base[i] = BASE_MIN;
for (i = 0; i < (int) table_size; i++)
check[i] = -1;
for (i = 0; i < nentries; i++)
{
- state = matching_state (i);
+ state_number_t state = matching_state (i);
+ base_t place;
if (state < 0)
+ /* A new set of state actions, or a nonterminal. */
place = pack_vector (i);
else
+ /* Action of I were already coded for STATE. */
place = base[state];
pos[i] = place;
base[order[i]] = place;
}
+ /* Use the greatest possible negative infinites. */
+ base_ninf = table_ninf_remap (base, nvectors, BASE_MIN);
+ table_ninf = table_ninf_remap (table, high + 1, ACTION_MIN);
+
for (i = 0; i < nvectors; i++)
{
XFREE (froms[i]);
XFREE (conflict_tos[i]);
}
- XFREE (froms);
- XFREE (tos);
- XFREE (conflict_tos);
- XFREE (pos);
+ free (froms);
+ free (tos);
+ free (conflict_tos);
+ free (pos);
}
+
/* the following functions output yytable, yycheck, yyconflp, yyconfl,
- and the vectors whose elements index the portion starts */
+ and the vectors whose elements index the portion starts. */
static void
output_base (void)
{
- /* Output pact. */
- muscle_insert_short_table ("pact", base,
+ /* Output PACT. */
+ muscle_insert_base_table ("pact", base,
base[0], 1, nstates);
+ MUSCLE_INSERT_INT ("pact_ninf", base_ninf);
- /* Output pgoto. */
- muscle_insert_short_table ("pgoto", base,
+ /* Output PGOTO. */
+ muscle_insert_base_table ("pgoto", base,
base[nstates], nstates + 1, nvectors);
XFREE (base);
}
static void
output_table (void)
{
- muscle_insert_short_table ("table", table,
- table[0], 1, high + 1);
+ muscle_insert_base_table ("table", table,
+ table[0], 1, high + 1);
+ MUSCLE_INSERT_INT ("table_ninf", table_ninf);
XFREE (table);
}
static void
output_check (void)
{
- muscle_insert_short_table ("check", check,
- check[0], 1, high + 1);
+ muscle_insert_base_table ("check", check,
+ check[0], 1, high + 1);
XFREE (check);
}
`-----------------------------------------------------------------*/
static void
-output_actions (void)
+prepare_actions (void)
{
/* That's a poor way to make sure the sizes are properly corelated,
in particular the signedness is not taking into account, but it's
nvectors = state_number_as_int (nstates) + nvars;
- froms = XCALLOC (short *, nvectors);
- tos = XCALLOC (short *, nvectors);
+ froms = XCALLOC (base_t *, nvectors);
+ tos = XCALLOC (base_t *, nvectors);
conflict_tos = XCALLOC (unsigned int *, nvectors);
tally = XCALLOC (short, nvectors);
- width = XCALLOC (short, nvectors);
+ width = XCALLOC (base_t, nvectors);
token_actions ();
bitsetv_free (LA);
XFREE (from_state);
XFREE (to_state);
+ order = XCALLOC (vector_number_t, nvectors);
sort_actions ();
pack_table ();
+ free (order);
+
+ free (tally);
+ free (width);
output_base ();
output_table ();
}
\f
-/*----------------------.
-| Run our backend, M4. |
-`----------------------*/
-
-static void
-m4_invoke (const char *definitions)
-{
- /* Invoke m4 on the definition of the muscles, and the skeleton. */
- const char *bison_pkgdatadir = getenv ("BISON_PKGDATADIR");
- const char *m4 = getenv ("M4");
- int pkg_data_len;
- char *full_skeleton;
-
- if (!m4)
- m4 = M4;
- if (!bison_pkgdatadir)
- bison_pkgdatadir = PKGDATADIR;
- pkg_data_len = strlen (bison_pkgdatadir);
- full_skeleton = XMALLOC (char, pkg_data_len + strlen (skeleton) + 2);
- if (bison_pkgdatadir[pkg_data_len-1] == '/')
- sprintf (full_skeleton, "%s%s", bison_pkgdatadir, skeleton);
- else
- sprintf (full_skeleton, "%s/%s", bison_pkgdatadir, skeleton);
- if (trace_flag)
- fprintf (stderr,
- "running: %s -I %s m4sugar/m4sugar.m4 %s %s\n",
- m4, bison_pkgdatadir, definitions, full_skeleton);
- skel_in = readpipe (m4,
- "-I", bison_pkgdatadir,
- "m4sugar/m4sugar.m4",
- definitions,
- full_skeleton,
- NULL);
- XFREE (full_skeleton);
- if (!skel_in)
- error (EXIT_FAILURE, errno, "cannot run m4");
- skel_lex ();
-
-}
-
/*---------------------------.
| Call the skeleton parser. |
`---------------------------*/
m4_invoke (tempfile);
/* If `debugging', keep this file alive. */
- if (!trace_flag)
+ if (!(trace_flag & trace_tools))
unlink (tempfile);
free (tempfile);
static void
prepare (void)
{
- MUSCLE_INSERT_INT ("last", high);
- MUSCLE_INSERT_INT ("flag", SHRT_MIN);
+ /* Flags. */
+ MUSCLE_INSERT_INT ("locations_flag", locations_flag);
+ MUSCLE_INSERT_INT ("defines_flag", defines_flag);
+ MUSCLE_INSERT_INT ("error_verbose", error_verbose);
MUSCLE_INSERT_INT ("pure", pure_parser);
- MUSCLE_INSERT_INT ("nsym", nsyms);
MUSCLE_INSERT_INT ("debug", debug_flag);
- MUSCLE_INSERT_INT ("final", final_state->number);
- MUSCLE_INSERT_INT ("undef_token_number", undeftoken->number);
- MUSCLE_INSERT_INT ("user_token_number_max", max_user_token_number);
- MUSCLE_INSERT_INT ("error_verbose", error_verbose);
- MUSCLE_INSERT_STRING ("prefix", spec_name_prefix ? spec_name_prefix : "yy");
/* FIXME: This is wrong: the muscles should decide whether they hold
a copy or not, but the situation is too obscure currently. */
+ MUSCLE_INSERT_STRING ("prefix", spec_name_prefix ? spec_name_prefix : "yy");
MUSCLE_INSERT_STRING ("output_infix", output_infix ? output_infix : "");
MUSCLE_INSERT_STRING ("output_prefix", short_base_name);
MUSCLE_INSERT_STRING ("output_parser_name", parser_file_name);
MUSCLE_INSERT_STRING ("output_header_name", spec_defines_file);
- MUSCLE_INSERT_INT ("nnts", nvars);
- MUSCLE_INSERT_INT ("nrules", nrules);
- MUSCLE_INSERT_INT ("nstates", nstates);
- MUSCLE_INSERT_INT ("ntokens", ntokens);
+ /* Symbols. */
+ MUSCLE_INSERT_INT ("tokens_number", ntokens);
+ MUSCLE_INSERT_INT ("nterms_number", nvars);
+ MUSCLE_INSERT_INT ("undef_token_number", undeftoken->number);
+ MUSCLE_INSERT_INT ("user_token_number_max", max_user_token_number);
- MUSCLE_INSERT_INT ("locations_flag", locations_flag);
- MUSCLE_INSERT_INT ("defines_flag", defines_flag);
+ /* Rules. */
+ MUSCLE_INSERT_INT ("rules_number", nrules);
+
+ /* States. */
+ MUSCLE_INSERT_INT ("last", high);
+ MUSCLE_INSERT_INT ("final_state_number", final_state->number);
+ MUSCLE_INSERT_INT ("states_number", nstates);
- /* Copy definitions in directive. */
+ /* User Code. */
obstack_1grow (&pre_prologue_obstack, 0);
obstack_1grow (&post_prologue_obstack, 0);
muscle_insert ("pre_prologue", obstack_finish (&pre_prologue_obstack));
prepare_tokens ();
prepare_rules ();
prepare_states ();
- output_actions ();
+ prepare_actions ();
prepare ();
projects / bison.git / blobdiff