X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/f958596b446e6d032c614258e10297c4d060f857..574fb2d5682af3cf6917e414b4240879d3e7f008:/src/output.c diff --git a/src/output.c b/src/output.c index e630b150..858bb98f 100644 --- a/src/output.c +++ b/src/output.c @@ -115,14 +115,21 @@ static int nvectors; static int nentries; static short **froms = NULL; static short **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 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). */ @@ -157,6 +164,8 @@ table_grow (size_t desired) table = XREALLOC (table, short, table_size); check = XREALLOC (check, short, table_size); + if (glr_parser) + conflict_table = XREALLOC (conflict_table, unsigned int, table_size); for (/* Nothing. */; old_size < table_size; ++old_size) { @@ -216,6 +225,7 @@ GENERATE_MUSCLE_INSERT_TABLE(muscle_insert_unsigned_int_table, unsigned int) GENERATE_MUSCLE_INSERT_TABLE(muscle_insert_short_table, short) 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) /*-----------------------------------------------------------------. @@ -235,11 +245,11 @@ prepare_tokens (void) int j = 0; for (i = 0; i < nsyms; i++) { - /* Be sure not to use twice the same quotearg slot. */ + /* 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, - quotearg_style (escape_quoting_style, - symbols[i]->tag)); + symbol_tag_get (symbols[i])); /* Width of the next token, including the two quotes, the coma and the space. */ int strsize = strlen (cp) + 2; @@ -278,23 +288,25 @@ prepare_tokens (void) /*-------------------------------------------------------------. | Prepare the muscles related to the rules: rhs, prhs, r1, r2, | -| rline. | +| rline, dprec, merger | `-------------------------------------------------------------*/ 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) { - item_number_t *rhsp; + item_number_t *rhsp = NULL; /* Index of rule R in RHS. */ prhs[r] = i; /* RHS of the rule R. */ @@ -307,7 +319,11 @@ prepare_rules (void) /* Separator in RHS. */ rhs[i++] = -1; /* Line where rule was defined. */ - rline[r] = rules[r].line; + 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); @@ -316,12 +332,16 @@ prepare_rules (void) 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); free (rhs); free (prhs); free (rline); free (r1); free (r2); + free (dprec); + free (merger); } /*--------------------------------------------. @@ -331,13 +351,57 @@ prepare_rules (void) static void prepare_states (void) { - size_t i; + 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; muscle_insert_symbol_number_table ("stos", values, - 0, 1, nstates); + 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 +conflict_row (state_t *state) +{ + int i, j; + + if (! glr_parser) + return; + + 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] != -state->lookaheads_rule[i]->number) + { + assert (conflict_list_free > 0); + conflict_list[conflict_list_cnt] + = state->lookaheads_rule[i]->number; + conflict_list_cnt += 1; + conflict_list_free -= 1; + } + + /* Leave a 0 at the end */ + assert (conflict_list_free > 0); + conflict_list_cnt += 1; + conflict_list_free -= 1; + } } @@ -353,21 +417,27 @@ prepare_states (void) | 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 action_row (state_t *state) { int i; - int default_rule = 0; - reductions *redp = state->reductions; - shifts *shiftp = state->shifts; - errs *errp = state->errs; + rule_number_t default_rule = 0; + reductions_t *redp = state->reductions; + shifts_t *shiftp = state->shifts; + errs_t *errp = state->errs; /* set 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) { @@ -377,35 +447,34 @@ action_row (state_t *state) 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_EXECUTE (state->lookaheads[i], 0, j, + { /* and record this rule as the rule to use if that token follows. */ - if (bitset_test (LA[state->lookaheadsp + i], j)) - actrow[j] = -LArule[state->lookaheadsp + i]->number; + if (actrow[j] != 0) + conflicted = conflrow[j] = 1; + actrow[j] = -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++) - { - symbol_number_t symbol; - int shift_state = shiftp->shifts[i]; - if (!shift_state) - continue; - - symbol = states[shift_state]->accessing_symbol; - - if (ISVAR (symbol)) - break; + for (i = 0; i < shiftp->nshifts && SHIFT_IS_SHIFT (shiftp, i); i++) + if (!SHIFT_IS_DISABLED (shiftp, i)) + { + symbol_number_t symbol = SHIFT_SYMBOL (shiftp, i); + state_number_t shift_state = shiftp->shifts[i]; - actrow[symbol] = shift_state; + if (actrow[symbol] != 0) + conflicted = conflrow[symbol] = 1; + actrow[symbol] = state_number_as_int (shift_state); - /* Do not use any default reduction if there is a shift for - error */ - if (symbol == errtoken->number) - nodefault = 1; - } + /* Do not use any default reduction if there is a shift for + error */ + if (symbol == errtoken->number) + nodefault = 1; + } /* See which tokens are an explicit error in this state (due to %nonassoc). For them, record SHRT_MIN as the action. */ @@ -428,8 +497,8 @@ action_row (state_t *state) for (i = 0; i < state->nlookaheads; i++) { int count = 0; - int rule = -LArule[state->lookaheadsp + i]->number; - int j; + rule_number_t rule = -state->lookaheads_rule[i]->number; + symbol_number_t j; for (j = 0; j < ntokens; j++) if (actrow[j] == rule) @@ -442,18 +511,20 @@ action_row (state_t *state) } } - /* 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] == default_rule && ! (glr_parser && conflrow[j])) actrow[j] = 0; - - default_rule = -default_rule; } + default_rule = -default_rule; } } @@ -465,18 +536,22 @@ action_row (state_t *state) if (actrow[i] == SHRT_MIN) actrow[i] = 0; + if (conflicted) + conflict_row (state); + return default_rule; } 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; + short *sp = NULL; + short *sp1 = NULL; + short *sp2 = NULL; + unsigned int *sp3 = NULL; count = 0; for (i = 0; i < ntokens; i++) @@ -488,12 +563,18 @@ save_row (int state) froms[state] = sp1 = sp = XCALLOC (short, count); tos[state] = sp2 = XCALLOC (short, count); + if (glr_parser) + conflict_tos[state] = sp3 = XCALLOC (unsigned int, count); + else + conflict_tos[state] = NULL; for (i = 0; i < ntokens; i++) if (actrow[i] != 0) { *sp1++ = i; *sp2++ = actrow[i]; + if (glr_parser) + *sp3++ = conflrow[i]; } tally[state] = count; @@ -512,10 +593,23 @@ save_row (int state) static void token_actions (void) { - size_t i; + state_number_t i; + int nconflict = conflicts_total_count (); + short *yydefact = XCALLOC (short, nstates); actrow = XCALLOC (short, ntokens); + + conflrow = XCALLOC (short, ntokens); + 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; + for (i = 0; i < nstates; ++i) { yydefact[i] = action_row (states[i]); @@ -525,6 +619,7 @@ token_actions (void) muscle_insert_short_table ("defact", yydefact, yydefact[0], 1, nstates); XFREE (actrow); + XFREE (conflrow); XFREE (yydefact); } @@ -536,22 +631,47 @@ token_actions (void) 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 = 1; r < nrules + 1; ++r) + if (rules[r].action) { - fprintf (out, " case %d:\n", rule); + fprintf (out, " case %d:\n", r); 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"))); fprintf (out, " %s\n break;\n\n", - rules[rule].action); + 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. | @@ -562,6 +682,8 @@ 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]; @@ -597,21 +719,88 @@ token_definitions_output (FILE *out) first = 0; } + fputs ("])\n\n", out); +} + + +/*----------------------------------------. +| Output the symbol destructors to OOUT. | +`----------------------------------------*/ + +static void +symbol_destructors_output (FILE *out) +{ + int i; + int first = 1; + + fputs ("m4_define([b4_symbol_destructors], \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->destructor_location.first_line, + symbol_tag_get (symbol), + symbol->number, + symbol->destructor, + symbol->type_name); + + first = 0; + } + fputs ("])\n\n", out); +} + + +/*-------------------------------------. +| 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_get (symbol), + symbol->number, + symbol->printer, + symbol->type_name); + + first = 0; + } + fputs ("])\n\n", out); } 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; int count; - int symno = symbol - ntokens + nstates; + int symno = symbol - ntokens + state_number_as_int (nstates); - short begin = goto_map[symbol]; - short end = goto_map[symbol + 1]; + int begin = goto_map[symbol]; + int end = goto_map[symbol + 1]; count = 0; for (i = begin; i < end; i++) @@ -635,29 +824,31 @@ save_column (int symbol, int default_state) width[symno] = sp1[-1] - sp[0] + 1; } -static int -default_goto (int symbol) + +static state_number_t +default_goto (symbol_number_t symbol) { - size_t i; - size_t m = goto_map[symbol]; - size_t n = goto_map[symbol + 1]; - int default_state = -1; + state_number_t s; + int i; + int m = goto_map[symbol]; + int 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; @@ -676,19 +867,19 @@ default_goto (int symbol) 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, nsyms - ntokens); 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); save_column (i, default_state); yydefgoto[i - ntokens] = default_state; } - muscle_insert_short_table ("defgoto", yydefgoto, - yydefgoto[0], 1, nsyms - ntokens); + muscle_insert_state_number_table ("defgoto", yydefgoto, + yydefgoto[0], 1, nsyms - ntokens); XFREE (state_count); XFREE (yydefgoto); } @@ -772,6 +963,7 @@ pack_vector (int vector) int loc = 0; short *from = froms[i]; short *to = tos[i]; + unsigned int *conflict_to = conflict_tos[i]; assert (t); @@ -782,7 +974,7 @@ pack_vector (int vector) for (k = 0; ok && k < t; k++) { - loc = j + from[k]; + loc = j + state_number_as_int (from[k]); if (loc > (int) table_size) table_grow (loc); @@ -798,9 +990,11 @@ pack_vector (int vector) { for (k = 0; k < t; k++) { - loc = j + from[k]; - table[loc] = to[k]; - check[loc] = from[k]; + loc = j + state_number_as_int (from[k]); + table[loc] = state_number_as_int (to[k]); + if (glr_parser && conflict_to != NULL) + conflict_table[loc] = conflict_to[k]; + check[loc] = state_number_as_int (from[k]); } while (table[lowzero] != 0) @@ -828,6 +1022,8 @@ pack_table (void) base = XCALLOC (short, nvectors); pos = XCALLOC (short, nentries); table = XCALLOC (short, table_size); + if (glr_parser) + conflict_table = XCALLOC (unsigned int, table_size); check = XCALLOC (short, table_size); lowzero = 0; @@ -856,14 +1052,16 @@ pack_table (void) { XFREE (froms[i]); XFREE (tos[i]); + XFREE (conflict_tos[i]); } XFREE (froms); XFREE (tos); + XFREE (conflict_tos); XFREE (pos); } -/* the following functions output yytable, yycheck +/* the following functions output yytable, yycheck, yyconflp, yyconfl, and the vectors whose elements index the portion starts */ static void @@ -889,6 +1087,28 @@ output_table (void) } +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) { @@ -905,11 +1125,17 @@ output_check (void) static void output_actions (void) { - size_t 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); + conflict_tos = XCALLOC (unsigned int *, nvectors); tally = XCALLOC (short, nvectors); width = XCALLOC (short, nvectors); @@ -927,17 +1153,9 @@ output_actions (void) 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); } @@ -1011,13 +1229,11 @@ output_skeleton (void) fputs ("m4_changecom()\n", out); fputs ("m4_init()\n", out); - fputs ("m4_define([b4_actions], \n[[", out); actions_output (out); - fputs ("]])\n\n", out); - - fputs ("m4_define([b4_tokens], \n[", out); + merger_output (out); token_definitions_output (out); - fputs ("])\n\n", out); + symbol_destructors_output (out); + symbol_printers_output (out); muscles_m4_output (out); @@ -1042,7 +1258,7 @@ prepare (void) 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 ("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); @@ -1071,7 +1287,12 @@ prepare (void) /* Find the right skeleton file. */ if (!skeleton) - skeleton = "bison.simple"; + { + if (glr_parser) + skeleton = "glr.c"; + else + skeleton = "yacc.c"; + } /* Parse the skeleton file and output the needed parsers. */ muscle_insert ("skeleton", skeleton);