/* Output the generated parsing program for bison,
- Copyright 1984, 1986, 1989, 1992, 2000, 2001
+ Copyright (C) 1984, 1986, 1989, 1992, 2000, 2001, 2002
Free Software Foundation, Inc.
This file is part of Bison, the GNU Compiler Compiler.
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,
negative short int. Used to flag ?? */
#include "system.h"
+#include "bitsetv.h"
#include "quotearg.h"
+#include "error.h"
#include "getargs.h"
#include "files.h"
#include "gram.h"
#include "conflicts.h"
#include "muscle_tab.h"
+/* From src/scan-skel.l. */
+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 *state_count = NULL;
-static short *order = NULL;
-static short *base = NULL;
-static short *pos = NULL;
-static short *table = NULL;
-static short *check = 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 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;
-struct obstack muscle_obstack;
static struct obstack format_obstack;
int error_verbose = 0;
-/* Returns the number of lines of S. */
-static size_t
-get_lines_number (const char *s)
-{
- size_t lines = 0;
- size_t i;
- for (i = 0; s[i]; ++i)
- if (s[i] == '\n')
- ++lines;
+/*----------------------------------------------------------------.
+| If TABLE (and CHECK) appear to be small to be addressed at |
+| DESIRED, grow them. Note that TABLE[DESIRED] is to be used, so |
+| the desired size is at least DESIRED + 1. |
+`----------------------------------------------------------------*/
- return lines;
-}
+static void
+table_grow (size_t desired)
+{
+ size_t old_size = table_size;
+ while (table_size <= desired)
+ table_size *= 2;
-/* FIXME. */
+ if (trace_flag & trace_resource)
+ fprintf (stderr, "growing table and check from: %d to %d\n",
+ old_size, table_size);
-static inline void
-output_table_data (struct obstack *oout,
- short *table_data,
- short first,
- int begin,
- int end)
-{
- int i;
- int j = 1;
+ 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);
- obstack_fgrow1 (oout, "%6d", first);
- for (i = begin; i < end; ++i)
+ for (/* Nothing. */; old_size < table_size; ++old_size)
{
- obstack_1grow (oout, ',');
- if (j >= 10)
- {
- obstack_sgrow (oout, "\n ");
- j = 1;
- }
- else
- ++j;
- obstack_fgrow1 (oout, "%6d", table_data[i]);
+ table[old_size] = 0;
+ check[old_size] = -1;
}
- obstack_1grow (oout, 0);
}
-static void
-output_token_translations (void)
-{
- output_table_data (&format_obstack, token_translations,
- 0, 1, max_user_token_number + 1);
- muscle_insert ("translate", obstack_finish (&format_obstack));
- XFREE (token_translations);
+/*-------------------------------------------------------------------.
+| Create a function NAME which associates to the muscle NAME the |
+| result of formatting the FIRST and then TABLE_DATA[BEGIN..END[ (of |
+| TYPE), and to the muscle NAME_max, the max value of the |
+| TABLE_DATA. |
+`-------------------------------------------------------------------*/
+
+
+#define GENERATE_MUSCLE_INSERT_TABLE(Name, Type) \
+ \
+static void \
+Name (const char *name, \
+ Type *table_data, \
+ Type first, \
+ int begin, \
+ int end) \
+{ \
+ Type min = first; \
+ Type max = first; \
+ int i; \
+ int j = 1; \
+ \
+ obstack_fgrow1 (&format_obstack, "%6d", first); \
+ for (i = begin; i < end; ++i) \
+ { \
+ obstack_1grow (&format_obstack, ','); \
+ if (j >= 10) \
+ { \
+ obstack_sgrow (&format_obstack, "\n "); \
+ j = 1; \
+ } \
+ else \
+ ++j; \
+ obstack_fgrow1 (&format_obstack, "%6d", table_data[i]); \
+ 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_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), \
+ (long int) max); \
}
+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 the muscles related to the tokens: translate, tname, and |
+| toknum. |
+`-----------------------------------------------------------------*/
static void
-output_gram (void)
+prepare_tokens (void)
{
+ muscle_insert_symbol_number_table ("translate",
+ token_translations,
+ token_translations[0],
+ 1, max_user_token_number + 1);
+
{
int i;
- short *values = XCALLOC (short, nrules + 1);
- for (i = 0; i < nrules + 1; ++i)
- values[i] = rules[i].rhs;
- output_table_data (&format_obstack, values,
- 0, 1, nrules + 1);
- XFREE (values);
- }
+ int j = 0;
+ for (i = 0; i < nsyms; i++)
+ {
+ /* Be sure not to use twice the same QUOTEARG slot:
+ SYMBOL_TAG_GET uses slot 0. */
+ const char *cp =
+ quotearg_n_style (1, c_quoting_style,
+ symbols[i]->tag);
+ /* Width of the next token, including the two quotes, the coma
+ and the space. */
+ int strsize = strlen (cp) + 2;
+
+ if (j + strsize > 75)
+ {
+ obstack_sgrow (&format_obstack, "\n ");
+ j = 2;
+ }
+
+ obstack_sgrow (&format_obstack, cp);
+ obstack_sgrow (&format_obstack, ", ");
+ j += strsize;
+ }
+ /* Add a NULL entry to list of tokens (well, 0, as NULL might not be
+ defined). */
+ obstack_sgrow (&format_obstack, "0");
- muscle_insert ("prhs", obstack_finish (&format_obstack));
+ /* Finish table and store. */
+ obstack_1grow (&format_obstack, 0);
+ muscle_insert ("tname", obstack_finish (&format_obstack));
+ }
+ /* Output YYTOKNUM. */
{
- short *yyrhs;
int i;
+ int *values = XCALLOC (int, ntokens);
+ for (i = 0; i < ntokens; ++i)
+ values[i] = symbols[i]->user_token_number;
+ muscle_insert_int_table ("toknum", values,
+ values[0], 1, ntokens);
+ free (values);
+ }
+}
- yyrhs = XMALLOC (short, nritems);
-
- for (i = 1; i < nritems; ++i)
- yyrhs[i] = ritem[i] >= 0 ? ritem[i] : -1;
-
- output_table_data (&format_obstack, yyrhs,
- ritem[0], 1, nritems);
- muscle_insert ("rhs", obstack_finish (&format_obstack));
- XFREE (yyrhs);
- }
+/*-------------------------------------------------------------.
+| Prepare the muscles related to the rules: rhs, prhs, r1, r2, |
+| rline, dprec, merger |
+`-------------------------------------------------------------*/
-#if 0
- if (!semantic_parser)
- obstack_sgrow (&table_obstack, "\n#endif\n");
-#endif
+static void
+prepare_rules (void)
+{
+ rule_number_t r;
+ unsigned int i = 0;
+ item_number_t *rhs = XMALLOC (item_number_t, nritems);
+ 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 = NULL;
+ /* Index of rule R in RHS. */
+ prhs[r] = i;
+ /* RHS of the rule R. */
+ for (rhsp = rules[r].rhs; *rhsp >= 0; ++rhsp)
+ rhs[i++] = *rhsp;
+ /* LHS of the rule R. */
+ r1[r] = rules[r].lhs->number;
+ /* Length of rule R's RHS. */
+ r2[r] = i - prhs[r];
+ /* Separator in RHS. */
+ rhs[i++] = -1;
+ /* Line where rule was defined. */
+ rline[r] = rules[r].location.first_line;
+ /* Dynamic precedence (GLR) */
+ dprec[r] = rules[r].dprec;
+ /* Merger-function index (GLR) */
+ merger[r] = rules[r].merger;
+ }
+ assert (i == nritems);
+
+ muscle_insert_item_number_table ("rhs", rhs, ritem[0], 1, nritems);
+ 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);
+ free (rline);
+ free (r1);
+ free (r2);
+ free (dprec);
+ free (merger);
}
+/*--------------------------------------------.
+| Prepare the muscles related to the states. |
+`--------------------------------------------*/
static void
-output_stos (void)
+prepare_states (void)
{
- int i;
- short *values = (short *) alloca (sizeof (short) * nstates);
+ state_number_t i;
+ symbol_number_t *values =
+ (symbol_number_t *) alloca (sizeof (symbol_number_t) * nstates);
for (i = 0; i < nstates; ++i)
values[i] = states[i]->accessing_symbol;
- output_table_data (&format_obstack, values,
- 0, 1, nstates);
- muscle_insert ("stos", obstack_finish (&format_obstack));
+ muscle_insert_symbol_number_table ("stos", values,
+ 0, 1, nstates);
}
+/*-------------------------------------------------------------------.
+| For GLR parsers, for each conflicted token in STATE, as indicated |
+| 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. |
+`-------------------------------------------------------------------*/
+
static void
-output_rule_data (void)
+conflict_row (state_t *state)
{
- int i;
- int j;
- short *short_tab = NULL;
-
- {
- short *values = XCALLOC (short, nrules + 1);
- for (i = 0; i < nrules + 1; ++i)
- values[i] = rules[i].line;
- output_table_data (&format_obstack, values,
- 0, 1, nrules + 1);
- muscle_insert ("rline", obstack_finish (&format_obstack));
- XFREE (values);
- }
-
-
- j = 0;
- for (i = 0; i < nsyms; i++)
- {
- /* Be sure not to use twice the same quotearg slot. */
- const char *cp =
- quotearg_n_style (1, c_quoting_style,
- quotearg_style (escape_quoting_style, symbols[i]->tag));
- /* Width of the next token, including the two quotes, the coma
- and the space. */
- int strsize = strlen (cp) + 2;
-
- if (j + strsize > 75)
- {
- obstack_sgrow (&format_obstack, "\n ");
- j = 2;
- }
-
- obstack_sgrow (&format_obstack, cp);
- obstack_sgrow (&format_obstack, ", ");
- j += strsize;
- }
- /* add a NULL entry to list of tokens */
- obstack_sgrow (&format_obstack, "NULL");
-
- /* Finish table and store. */
- obstack_1grow (&format_obstack, 0);
- muscle_insert ("tname", obstack_finish (&format_obstack));
-
- /* Output YYTOKNUM. */
- {
- short *values = XCALLOC (short, ntokens + 1);
- for (i = 0; i < ntokens + 1; ++i)
- values[i] = symbols[i]->user_token_number;
- output_table_data (&format_obstack, values,
- 0, 1, ntokens + 1);
- muscle_insert ("toknum", obstack_finish (&format_obstack));
- XFREE (values);
- }
+ int i, j;
+ if (! glr_parser)
+ return;
- /* Output YYR1. */
- {
- short *values = XCALLOC (short, nrules + 1);
- for (i = 0; i < nrules + 1; ++i)
- values[i] = rules[i].lhs;
- output_table_data (&format_obstack, values,
- 0, 1, nrules + 1);
- muscle_insert ("r1", obstack_finish (&format_obstack));
- XFREE (values);
- }
+ for (j = 0; j < ntokens; j += 1)
+ if (conflrow[j])
+ {
+ conflrow[j] = conflict_list_cnt;
+
+ /* 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]
+ != 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 + 1;
+ conflict_list_cnt += 1;
+ conflict_list_free -= 1;
+ }
- /* Output YYR2. */
- short_tab = XMALLOC (short, nrules + 1);
- for (i = 1; i < nrules; i++)
- short_tab[i] = rules[i + 1].rhs - rules[i].rhs - 1;
- short_tab[nrules] = nritems - rules[nrules].rhs - 1;
- output_table_data (&format_obstack, short_tab,
- 0, 1, nrules + 1);
- muscle_insert ("r2", obstack_finish (&format_obstack));
- XFREE (short_tab);
+ /* 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 |
-| MINSHORT, 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) |
+| with symbol SYM. The default reduction is not used for a symbol |
+| 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 *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] = 0;
+ 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 (BITISSET (LA (state->lookaheadsp + i), j))
- actrow[j] = -LAruleno[state->lookaheadsp + i];
+ 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)
{
- int symbol;
- int 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];
- actrow[symbol] = shift_state;
+ if (actrow[symbol] != 0)
+ conflicted = conflrow[symbol] = 1;
+ actrow[symbol] = state_number_as_int (shift_state->number);
/* Do not use any default reduction if there is a shift for
error */
- if (symbol == error_token_number)
+ if (symbol == errtoken->number)
nodefault = 1;
}
/* See which tokens are an explicit error in this state (due to
- %nonassoc). For them, record MINSHORT 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] = MINSHORT;
+ 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 = -LAruleno[state->lookaheadsp + i];
- 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)
}
}
- /* actions which match the default are replaced with zero,
- which means "use the default" */
+ /* GLR parsers need space for conflict lists, so we can't
+ default conflicted entries. For non-conflicted entries
+ or as long as we are not building a GLR parser,
+ actions that match the default are replaced with zero,
+ which means "use the default". */
if (max > 0)
{
int j;
for (j = 0; j < ntokens; j++)
- if (actrow[j] == default_rule)
+ 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] == MINSHORT)
+ if (actrow[i] == ACTION_MIN)
actrow[i] = 0;
+ if (conflicted)
+ conflict_row (state);
+
return default_rule;
}
+/*--------------------------------------------.
+| Set FROMS, TOS, TALLY and WIDTH for STATE. |
+`--------------------------------------------*/
+
static void
-save_row (int state)
+save_row (state_number_t state)
{
- int i;
+ symbol_number_t i;
int count;
- short *sp;
- short *sp1;
- short *sp2;
+ 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)
{
*sp1++ = i;
*sp2++ = actrow[i];
+ if (glr_parser)
+ *sp3++ = conflrow[i];
}
tally[state] = count;
static void
token_actions (void)
{
- int i;
- short *yydefact = XCALLOC (short, nstates);
+ state_number_t i;
+ rule_number_t r;
+ int nconflict = conflicts_total_count ();
+
+ rule_number_t *yydefact = XCALLOC (rule_number_t, nstates);
+
+ 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;
+
+ if (glr_parser)
+ {
+ conflict_list = XCALLOC (unsigned int, 1 + 2 * nconflict);
+ conflict_list_free = 2 * nconflict;
+ conflict_list_cnt = 1;
+ }
+ else
+ conflict_list_free = conflict_list_cnt = 0;
- actrow = XCALLOC (short, ntokens);
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);
}
- output_table_data (&format_obstack, yydefact,
- yydefact[0], 1, nstates);
- muscle_insert ("defact", obstack_finish (&format_obstack));
+ 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);
}
| Output the actions to OOUT. |
`-----------------------------*/
-static void
-actions_output (FILE *out, size_t *line)
+void
+actions_output (FILE *out)
{
- int rule;
- for (rule = 1; rule < nrules + 1; ++rule)
- if (rules[rule].action)
+ rule_number_t r;
+
+ fputs ("m4_define([b4_actions], \n[[", out);
+ 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_line,
+ rules[r].action_location.first_line,
quotearg_style (c_quoting_style,
muscle_find ("filename")));
- /* As a Bison extension, add the ending semicolon. Since some
- Yacc don't do that, help people using bison as a Yacc
- finding their missing semicolons. */
- fprintf (out, "{ %s%s }\n break;\n\n",
- rules[rule].action,
- yacc_flag ? ";" : "");
-
- /* We always output 4 '\n' per action. */
- *line += 4;
- /* Plus one if !no_lines_flag. */
- if (!no_lines_flag)
- ++*line;
- /* Get the number of lines written by the user. */
- *line += get_lines_number (rules[rule].action);
+ fprintf (out, " %s\n break;\n\n",
+ rules[r].action);
}
+ fputs ("]])\n\n", out);
+}
+
+/*--------------------------------------.
+| Output the merge functions to OUT. |
+`--------------------------------------*/
+
+static void
+merger_output (FILE *out)
+{
+ int n;
+ merger_list* p;
+
+ fputs ("m4_define([b4_mergers], \n[[", out);
+ for (n = 1, p = merge_functions; p != NULL; n += 1, p = p->next)
+ {
+ if (p->type[0] == '\0')
+ fprintf (out, " case %d: yyval = %s (*yy0, *yy1); break;\n",
+ n, p->name);
+ else
+ fprintf (out, " case %d: yyval.%s = %s (*yy0, *yy1); break;\n",
+ n, p->type, p->name);
+ }
+ fputs ("]])\n\n", out);
+}
+
+/*---------------------------------------.
+| Output the tokens definition to OOUT. |
+`---------------------------------------*/
+
+void
+token_definitions_output (FILE *out)
+{
+ int i;
+ int first = 1;
+
+ fputs ("m4_define([b4_tokens], \n[", out);
+ for (i = 0; i < ntokens; ++i)
+ {
+ symbol_t *symbol = symbols[i];
+ int number = symbol->user_token_number;
+
+ /* At this stage, if there are literal aliases, they are part of
+ SYMBOLS, so we should not find symbols which are the aliases
+ here. */
+ assert (number != USER_NUMBER_ALIAS);
+
+ /* Skip error token. */
+ if (symbol == errtoken)
+ continue;
+
+ /* If this string has an alias, then it is necessarily the alias
+ which is to be output. */
+ if (symbol->alias)
+ symbol = symbol->alias;
+
+ /* Don't output literal chars or strings (when defined only as a
+ string). Note that must be done after the alias resolution:
+ think about `%token 'f' "f"'. */
+ if (symbol->tag[0] == '\'' || symbol->tag[0] == '\"')
+ continue;
+
+ /* Don't #define nonliteral tokens whose names contain periods
+ or '$' (as does the default value of the EOF token). */
+ if (strchr (symbol->tag, '.') || strchr (symbol->tag, '$'))
+ continue;
+
+ fprintf (out, "%s[[[%s]], [%d]]",
+ first ? "" : ",\n", symbol->tag, number);
+
+ first = 0;
+ }
+ fputs ("])\n\n", out);
}
-/*----------------------------.
-| Output the guards to OOUT. |
-`----------------------------*/
+/*----------------------------------------.
+| Output the symbol destructors to OOUT. |
+`----------------------------------------*/
static void
-guards_output (FILE *out, size_t *line)
+symbol_destructors_output (FILE *out)
{
- int rule;
- for (rule = 1; rule < nrules + 1; ++rule)
- if (rules[rule].action)
+ int i;
+ int first = 1;
+
+ fputs ("m4_define([b4_symbol_destructors], \n[", out);
+ for (i = 0; i < nsyms; ++i)
+ if (symbols[i]->destructor)
{
- fprintf (out, " case %d:\n", rule);
+ symbol_t *symbol = symbols[i];
+
+ /* Filename, lineno,
+ Symbol-name, Symbol-number,
+ destructor, typename. */
+ fprintf (out, "%s[[[%s]], [[%d]], [[%s]], [[%d]], [[%s]], [[%s]]]",
+ first ? "" : ",\n",
+ infile, symbol->destructor_location.first_line,
+ symbol->tag,
+ symbol->number,
+ symbol->destructor,
+ symbol->type_name);
+
+ first = 0;
+ }
+ fputs ("])\n\n", out);
+}
- if (!no_lines_flag)
- fprintf (out, muscle_find ("linef"),
- rules[rule].guard_line,
- quotearg_style (c_quoting_style,
- muscle_find ("filename")));
- fprintf (out, "{ %s; }\n break;\n\n",
- rules[rule].guard);
- /* We always output 4 '\n' per action. */
- *line += 4;
- /* Plus one if !no_lines_flag. */
- if (!no_lines_flag)
- ++*line;
- /* Get the number of lines written by the user. */
- *line += get_lines_number (rules[rule].guard);
+/*-------------------------------------.
+| Output the symbol printers to OOUT. |
+`-------------------------------------*/
+
+static void
+symbol_printers_output (FILE *out)
+{
+ int i;
+ int first = 1;
+
+ fputs ("m4_define([b4_symbol_printers], \n[", out);
+ for (i = 0; i < nsyms; ++i)
+ if (symbols[i]->destructor)
+ {
+ symbol_t *symbol = symbols[i];
+
+ /* Filename, lineno,
+ Symbol-name, Symbol-number,
+ destructor, typename. */
+ fprintf (out, "%s[[[%s]], [[%d]], [[%s]], [[%d]], [[%s]], [[%s]]]",
+ first ? "" : ",\n",
+ infile, symbol->printer_location.first_line,
+ symbol->tag,
+ symbol->number,
+ symbol->printer,
+ symbol->type_name);
+
+ first = 0;
}
+ fputs ("])\n\n", out);
}
+/*------------------------------------------------------------------.
+| 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 (int symbol, int default_state)
+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 + nstates;
+ vector_number_t symno = symbol_number_to_vector_number (symbol);
- short begin = goto_map[symbol];
- short 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)
{
width[symno] = sp1[-1] - sp[0] + 1;
}
-static int
-default_goto (int symbol)
+
+/*----------------------------------------------------------------.
+| Return `the' most common destination GOTO on SYMBOL (a nterm). |
+`----------------------------------------------------------------*/
+
+static state_number_t
+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];
- int default_state = -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;
if (m == n)
- return -1;
+ return (state_number_t) -1;
- for (i = 0; i < nstates; i++)
- state_count[i] = 0;
+ for (s = 0; s < nstates; s++)
+ state_count[s] = 0;
for (i = m; i < n; i++)
state_count[to_state[i]]++;
- for (i = 0; i < nstates; i++)
- if (state_count[i] > max)
+ for (s = 0; s < nstates; s++)
+ if (state_count[s] > max)
{
- max = state_count[i];
- default_state = i;
+ max = state_count[s];
+ default_state = s;
}
return default_state;
static void
goto_actions (void)
{
- int i;
- short *yydefgoto = XMALLOC (short, nsyms - ntokens);
+ symbol_number_t i;
+ 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)
{
- int 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;
}
- output_table_data (&format_obstack, yydefgoto,
- yydefgoto[0], 1, nsyms - ntokens);
- muscle_insert ("defgoto", obstack_finish (&format_obstack));
-
+ muscle_insert_state_number_table ("defgoto", yydefgoto,
+ yydefgoto[0], 1, nsyms - ntokens);
XFREE (state_count);
XFREE (yydefgoto);
}
-/* 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 (i >= nstates)
+ /* If VECTOR is a nterm, return -1. */
+ if (i >= (int) nstates)
return -1;
t = tally[i];
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 (j = lowzero - from[0]; j < MAXTABLE; j++)
+ for (j = lowzero - from[0]; j < (int) table_size; j++)
{
int k;
int ok = 1;
for (k = 0; ok && k < t; k++)
{
- loc = j + from[k];
- if (loc > MAXTABLE)
- fatal (_("maximum table size (%d) exceeded"), MAXTABLE);
+ loc = j + state_number_as_int (from[k]);
+ if (loc > (int) table_size)
+ table_grow (loc);
if (table[loc] != 0)
ok = 0;
{
loc = j + from[k];
table[loc] = to[k];
+ if (glr_parser && conflict_to != NULL)
+ conflict_table[loc] = conflict_to[k];
check[loc] = from[k];
}
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, MAXTABLE);
- check = XCALLOC (short, MAXTABLE);
+ 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 (base_t, table_size);
lowzero = 0;
high = 0;
for (i = 0; i < nvectors; i++)
- base[i] = MINSHORT;
+ base[i] = BASE_MIN;
- for (i = 0; i < MAXTABLE; i++)
+ 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 (tos[i]);
+ XFREE (conflict_tos[i]);
}
- XFREE (froms);
- XFREE (tos);
- XFREE (pos);
+ free (froms);
+ free (tos);
+ free (conflict_tos);
+ free (pos);
}
-/* the following functions output yytable, yycheck
- and the vectors whose elements index the portion starts */
+
+/* the following functions output yytable, yycheck, yyconflp, yyconfl,
+ and the vectors whose elements index the portion starts. */
static void
output_base (void)
{
- /* Output pact. */
- output_table_data (&format_obstack, base,
- base[0], 1, nstates);
- muscle_insert ("pact", obstack_finish (&format_obstack));
-
- /* Output pgoto. */
- output_table_data (&format_obstack, base,
- base[nstates], nstates + 1, nvectors);
- muscle_insert ("pgoto", obstack_finish (&format_obstack));
-
+ /* Output PACT. */
+ muscle_insert_base_table ("pact", base,
+ base[0], 1, nstates);
+ MUSCLE_INSERT_INT ("pact_ninf", base_ninf);
+
+ /* Output PGOTO. */
+ muscle_insert_base_table ("pgoto", base,
+ base[nstates], nstates + 1, nvectors);
XFREE (base);
}
static void
output_table (void)
{
- output_table_data (&format_obstack, table,
- table[0], 1, high + 1);
- muscle_insert ("table", obstack_finish (&format_obstack));
+ muscle_insert_base_table ("table", table,
+ table[0], 1, high + 1);
+ MUSCLE_INSERT_INT ("table_ninf", table_ninf);
XFREE (table);
}
+static void
+output_conflicts (void)
+{
+ /* GLR parsing slightly modifies yytable and yycheck
+ (and thus yypact) so that in states with unresolved conflicts,
+ the default reduction is not used in the conflicted entries, so
+ that there is a place to put a conflict pointer. This means that
+ yyconflp and yyconfl are nonsense for a non-GLR parser, so we
+ avoid accidents by not writing them out in that case. */
+ if (! glr_parser)
+ return;
+
+ muscle_insert_unsigned_int_table ("conflict_list_heads", conflict_table,
+ conflict_table[0], 1, high+1);
+ muscle_insert_unsigned_int_table ("conflicting_rules", conflict_list,
+ conflict_list[0], 1, conflict_list_cnt);
+
+ XFREE (conflict_table);
+ XFREE (conflict_list);
+}
+
+
static void
output_check (void)
{
- output_table_data (&format_obstack, check,
- check[0], 1, high + 1);
- muscle_insert ("check", obstack_finish (&format_obstack));
+ muscle_insert_base_table ("check", check,
+ check[0], 1, high + 1);
XFREE (check);
}
-/* compute and output yydefact, yydefgoto, yypact, yypgoto, yytable
- and yycheck. */
+/*-----------------------------------------------------------------.
+| Compute and output yydefact, yydefgoto, yypact, yypgoto, yytable |
+| and yycheck. |
+`-----------------------------------------------------------------*/
static void
-output_actions (void)
+prepare_actions (void)
{
- int i;
- nvectors = nstates + nvars;
+ /* 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
+ not useless. */
+ assert (sizeof (nvectors) >= sizeof (nstates));
+ assert (sizeof (nvectors) >= sizeof (nvars));
+
+ 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 ();
- XFREE (LA);
- XFREE (LAruleno);
+ bitsetv_free (LA);
+ free (LArule);
goto_actions ();
XFREE (goto_map + ntokens);
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 ();
+ output_conflicts ();
output_check ();
-
- for (i = 0; i < nstates; ++i)
- {
- free (states[i]->shifts);
- XFREE (states[i]->reductions);
- free (states[i]->errs);
- free (states[i]);
- }
- XFREE (states);
}
\f
-/*------------------------------------------------------------.
-| Copy the parser code from SKEL_FILENAME into OOUT obstack. |
-| and do the muscle substitution. |
-`------------------------------------------------------------*/
+/*---------------------------.
+| Call the skeleton parser. |
+`---------------------------*/
static void
-output_parser (const char *skel_filename, FILE *out)
+output_skeleton (void)
{
- int c;
- FILE *fskel;
- size_t output_line;
- size_t skeleton_line;
-
- fskel = xfopen (skel_filename, "r");
-
- /* New output code. */
- output_line = 1;
- skeleton_line = 1;
- c = getc (fskel);
- while (c != EOF)
- {
- if (c != '%')
- {
- if (c == '\n')
- {
- ++output_line;
- ++skeleton_line;
- }
- putc (c, out);
- c = getc (fskel);
- }
- else if ((c = getc (fskel)) == '%')
- {
- /* Read the muscle. */
- const char *muscle_key = 0;
- const char *muscle_value = 0;
-
- while (isalnum (c = getc (fskel)) || c == '-')
- obstack_1grow (&muscle_obstack, c);
- obstack_1grow (&muscle_obstack, 0);
-
- /* Output the right value, or see if it's something special. */
- muscle_key = obstack_finish (&muscle_obstack);
- muscle_value = muscle_find (muscle_key);
- if (!strcmp (muscle_key, "actions"))
- actions_output (out, &output_line);
- else if (!strcmp (muscle_key, "guards"))
- guards_output (out, &output_line);
- else if (!strcmp (muscle_key, "line"))
- fprintf (out, "%d", output_line);
- else if (!strcmp (muscle_key, "skeleton-line"))
- fprintf (out, "%d", skeleton_line);
- else if (muscle_value)
- {
- fputs (muscle_value, out);
- output_line += get_lines_number (muscle_value);
- }
- else
- {
- fputs ("%%", out);
- fputs (muscle_key, out);
- }
- }
- else
- putc ('%', out);
- }
-
- /* End. */
- xfclose (fskel);
-}
-
-/*----------------------------------------.
-| Prepare the master parser to be output |
-`----------------------------------------*/
-
-static void
-output_master_parser (void)
-{
- FILE *parser = xfopen (parser_file_name, "w");
- if (!skeleton)
- {
- if (semantic_parser)
- skeleton = skeleton_find ("BISON_HAIRY", BISON_HAIRY);
- else
- skeleton = skeleton_find ("BISON_SIMPLE", BISON_SIMPLE);
- }
- muscle_insert ("skeleton", skeleton);
- muscle_insert ("parser-file-name", parser_file_name);
-
- output_parser (skeleton, parser);
- xfclose (parser);
-}
-
-
-/* FIXME. */
+ /* Store the definition of all the muscles. */
+ const char *tempdir = getenv ("TMPDIR");
+ char *tempfile = NULL;
+ FILE *out = NULL;
+ int fd;
+
+ if (tempdir == NULL)
+ tempdir = DEFAULT_TMPDIR;
+ tempfile = xmalloc (strlen (tempdir) + 11);
+ sprintf (tempfile, "%s/bsnXXXXXX", tempdir);
+ fd = mkstemp (tempfile);
+ if (fd == -1)
+ error (EXIT_FAILURE, errno, "%s", tempfile);
+
+ out = fdopen (fd, "w");
+ if (out == NULL)
+ error (EXIT_FAILURE, errno, "%s", tempfile);
+
+ /* There are no comments, especially not `#': we do want M4 expansion
+ after `#': think of CPP macros! */
+ fputs ("m4_changecom()\n", out);
+ fputs ("m4_init()\n", out);
+
+ actions_output (out);
+ merger_output (out);
+ token_definitions_output (out);
+ symbol_destructors_output (out);
+ symbol_printers_output (out);
+
+ muscles_m4_output (out);
+
+ fputs ("m4_wrap([m4_divert_pop(0)])\n", out);
+ fputs ("m4_divert_push(0)dnl\n", out);
+ xfclose (out);
-#define MUSCLE_INSERT_INT(Key, Value) \
-{ \
- obstack_fgrow1 (&muscle_obstack, "%d", Value); \
- obstack_1grow (&muscle_obstack, 0); \
- muscle_insert (Key, obstack_finish (&muscle_obstack)); \
-}
+ m4_invoke (tempfile);
-#define MUSCLE_INSERT_STRING(Key, Value) \
-{ \
- obstack_sgrow (&muscle_obstack, Value); \
- obstack_1grow (&muscle_obstack, 0); \
- muscle_insert (Key, obstack_finish (&muscle_obstack)); \
-}
+ /* If `debugging', keep this file alive. */
+ if (!(trace_flag & trace_tools))
+ unlink (tempfile);
-#define MUSCLE_INSERT_PREFIX(Key, Value) \
-{ \
- obstack_fgrow2 (&muscle_obstack, "%s%s", spec_name_prefix, Value); \
- obstack_1grow (&muscle_obstack, 0); \
- muscle_insert (Key, obstack_finish (&muscle_obstack)); \
+ free (tempfile);
}
static void
prepare (void)
{
- MUSCLE_INSERT_INT ("last", high);
- MUSCLE_INSERT_INT ("flag", MINSHORT);
+ /* 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);
- MUSCLE_INSERT_INT ("maxtok", max_user_token_number);
- MUSCLE_INSERT_INT ("error-verbose", error_verbose);
- MUSCLE_INSERT_STRING ("prefix", spec_name_prefix);
- MUSCLE_INSERT_INT ("nnts", nvars);
- MUSCLE_INSERT_INT ("nrules", nrules);
- MUSCLE_INSERT_INT ("nstates", nstates);
- MUSCLE_INSERT_INT ("ntokens", ntokens);
+ /* 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 ("locations-flag", locations_flag);
-}
+ /* 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);
+ /* Rules. */
+ MUSCLE_INSERT_INT ("rules_number", nrules);
-/*-------------------------.
-| Output the header file. |
-`-------------------------*/
+ /* States. */
+ MUSCLE_INSERT_INT ("last", high);
+ MUSCLE_INSERT_INT ("final_state_number", final_state->number);
+ MUSCLE_INSERT_INT ("states_number", nstates);
-static void
-header_output (void)
-{
- FILE *out = xfopen (spec_defines_file, "w");
- char *macro_name = compute_header_macro ();
-
- fprintf (out, "#ifndef %s\n", macro_name);
- fprintf (out, "# define %s\n\n", macro_name);
-
- fputs (muscle_find ("tokendef"), out);
- fprintf (out, "\
-#ifndef YYSTYPE\n\
-typedef %s
-yystype;\n\
-# define YYSTYPE yystype\n\
-#endif\n",
- muscle_find ("stype"));
-
- if (!pure_parser)
- fprintf (out, "\nextern YYSTYPE %slval;\n",
- spec_name_prefix);
- if (semantic_parser)
- {
- int i;
+ /* User Code. */
+ obstack_1grow (&pre_prologue_obstack, 0);
+ obstack_1grow (&post_prologue_obstack, 0);
+ muscle_insert ("pre_prologue", obstack_finish (&pre_prologue_obstack));
+ muscle_insert ("post_prologue", obstack_finish (&post_prologue_obstack));
- for (i = ntokens; i < nsyms; i++)
- /* don't make these for dummy nonterminals made by gensym. */
- if (*symbols[i]->tag != '@')
- fprintf (out, "# define NT%s\t%d\n", symbols[i]->tag, i);
+ /* Find the right skeleton file. */
+ if (!skeleton)
+ {
+ if (glr_parser)
+ skeleton = "glr.c";
+ else
+ skeleton = "yacc.c";
}
- fprintf (out, "\n#endif /* not %s */\n", macro_name);
- free (macro_name);
- xfclose (out);
+ /* Parse the skeleton file and output the needed parsers. */
+ muscle_insert ("skeleton", skeleton);
}
{
obstack_init (&format_obstack);
- output_token_translations ();
- output_gram ();
-
- XFREE (ritem);
- if (semantic_parser)
- output_stos ();
- output_rule_data ();
- output_actions ();
+ prepare_tokens ();
+ prepare_rules ();
+ prepare_states ();
+ prepare_actions ();
prepare ();
- /* Copy definitions in directive. */
- obstack_1grow (&attrs_obstack, 0);
- muscle_insert ("prologue", obstack_finish (&attrs_obstack));
-
- /* Output the parser. */
- output_master_parser ();
- /* Output the header if needed. */
- if (defines_flag)
- header_output ();
-
- free (rules + 1);
- obstack_free (&muscle_obstack, NULL);
+
+ /* Process the selected skeleton file. */
+ output_skeleton ();
+
obstack_free (&format_obstack, NULL);
- obstack_free (&action_obstack, NULL);
- obstack_free (&attrs_obstack, NULL);
+ obstack_free (&pre_prologue_obstack, NULL);
+ obstack_free (&post_prologue_obstack, NULL);
}
projects / bison.git / blobdiff