X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/342b8b6e2156e923797745af2da2596e3086ccd5..9a636f47cee64ab6e1f7588c6261fc46ea134c39:/src/reduce.c diff --git a/src/reduce.c b/src/reduce.c index b2299b33..25681750 100644 --- a/src/reduce.c +++ b/src/reduce.c @@ -1,5 +1,5 @@ /* Grammar reduction for Bison. - Copyright 1988, 1989, 2000 Free Software Foundation, Inc. + Copyright (C) 1988, 1989, 2000, 2001, 2002 Free Software Foundation, Inc. This file is part of Bison, the GNU Compiler Compiler. @@ -26,64 +26,37 @@ user's parser. */ #include "system.h" +#include "quotearg.h" #include "getargs.h" #include "files.h" +#include "symtab.h" #include "gram.h" #include "complain.h" #include "reduce.h" #include "reader.h" #include "getargs.h" +#include "bitset.h" -typedef unsigned *BSet; typedef short *rule; -/* N is set of all nonterminals which are not useless. P is set of - all rules which have no useless nonterminals in their RHS. V is - the set of all accessible symbols. */ +/* Set of all nonterminals which are not useless. */ +static bitset N; -static BSet N, P, V, V1; +/* Set of all rules which have no useless nonterminals in their RHS. */ +static bitset P; + +/* Set of all accessible symbols. */ +static bitset V; + +/* Set of symbols used to define rule precedence (so they are + `useless', but no warning should be issued). */ +static bitset V1; static int nuseful_productions; static int nuseless_productions; static int nuseful_nonterminals; -static int nuseless_nonterminals; - -static bool -bits_equal (BSet L, BSet R, int n) -{ - int i; - - for (i = n - 1; i >= 0; i--) - if (L[i] != R[i]) - return FALSE; - return TRUE; -} - - -static int -nbits (unsigned i) -{ - int count = 0; - - while (i != 0) - { - i ^= (i & ((unsigned) (-(int) i))); - ++count; - } - return count; -} - - -static int -bits_size (BSet S, int n) -{ - int i, count = 0; - - for (i = n - 1; i >= 0; i--) - count += nbits (S[i]); - return count; -} +int nuseless_nonterminals; /*-------------------------------------------------------------------. | Another way to do this would be with a set for each production and | @@ -92,7 +65,7 @@ bits_size (BSet S, int n) `-------------------------------------------------------------------*/ static bool -useful_production (int i, BSet N0) +useful_production (int i, bitset N0) { rule r; short n; @@ -100,10 +73,9 @@ useful_production (int i, BSet N0) /* A production is useful if all of the nonterminals in its appear in the set of useful nonterminals. */ - for (r = &ritem[rrhs[i]]; *r > 0; r++) - if (ISVAR (n = *r)) - if (!BITISSET (N0, n - ntokens)) - return FALSE; + for (r = rules[i].rhs; *r >= 0; r++) + if (ISVAR (n = *r) && !bitset_test (N0, n - ntokens)) + return FALSE; return TRUE; } @@ -115,13 +87,14 @@ useful_production (int i, BSet N0) static void useless_nonterminals (void) { - BSet Np, Ns; + bitset Np, Ns; int i; /* N is set as built. Np is set being built this iteration. P is set of all productions which have a RHS all in N. */ - Np = XCALLOC (unsigned, WORDSIZE (nvars)); + Np = bitset_create (nvars, BITSET_FIXED); + /* The set being computed is a set of nonterminals which can derive the empty string or strings consisting of all terminals. At each @@ -141,26 +114,21 @@ useless_nonterminals (void) while (1) { - for (i = WORDSIZE (nvars) - 1; i >= 0; i--) - Np[i] = N[i]; - for (i = 1; i <= nrules; i++) - { - if (!BITISSET (P, i)) - { - if (useful_production (i, N)) - { - SETBIT (Np, rlhs[i] - ntokens); - SETBIT (P, i); - } - } - } - if (bits_equal (N, Np, WORDSIZE (nvars))) + bitset_copy (Np, N); + for (i = 1; i < nrules + 1; i++) + if (!bitset_test (P, i) + && useful_production (i, N)) + { + bitset_set (Np, rules[i].lhs->number - ntokens); + bitset_set (P, i); + } + if (bitset_equal_p (N, Np)) break; Ns = Np; Np = N; N = Ns; } - XFREE (N); + bitset_free (N); N = Np; } @@ -168,7 +136,7 @@ useless_nonterminals (void) static void inaccessable_symbols (void) { - BSet Vp, Vs, Pp; + bitset Vp, Vs, Pp; int i; short t; rule r; @@ -196,316 +164,300 @@ inaccessable_symbols (void) terminals are printed (if running in verbose mode) so that the user can know. */ - Vp = XCALLOC (unsigned, WORDSIZE (nsyms)); - Pp = XCALLOC (unsigned, WORDSIZE (nrules + 1)); + Vp = bitset_create (nsyms, BITSET_FIXED); + Pp = bitset_create (nrules + 1, BITSET_FIXED); /* If the start symbol isn't useful, then nothing will be useful. */ - if (!BITISSET (N, start_symbol - ntokens)) - goto end_iteration; - - SETBIT (V, start_symbol); - - while (1) + if (bitset_test (N, start_symbol - ntokens)) { - for (i = WORDSIZE (nsyms) - 1; i >= 0; i--) - Vp[i] = V[i]; - for (i = 1; i <= nrules; i++) + bitset_set (V, start_symbol); + + while (1) { - if (!BITISSET (Pp, i) && BITISSET (P, i) && BITISSET (V, rlhs[i])) + bitset_copy (Vp, V); + for (i = 1; i < nrules + 1; i++) { - for (r = &ritem[rrhs[i]]; *r >= 0; r++) + if (!bitset_test (Pp, i) + && bitset_test (P, i) + && bitset_test (V, rules[i].lhs->number)) { - if (ISTOKEN (t = *r) || BITISSET (N, t - ntokens)) - { - SETBIT (Vp, t); - } + for (r = rules[i].rhs; *r >= 0; r++) + if (ISTOKEN (t = *r) || bitset_test (N, t - ntokens)) + bitset_set (Vp, t); + bitset_set (Pp, i); } - SETBIT (Pp, i); } + if (bitset_equal_p (V, Vp)) + break; + Vs = Vp; + Vp = V; + V = Vs; } - if (bits_equal (V, Vp, WORDSIZE (nsyms))) - { - break; - } - Vs = Vp; - Vp = V; - V = Vs; } -end_iteration: - XFREE (V); + bitset_free (V); V = Vp; /* Tokens 0, 1, and 2 are internal to Bison. Consider them useful. */ - SETBIT (V, 0); /* end-of-input token */ - SETBIT (V, 1); /* error token */ - SETBIT (V, 2); /* some undefined token */ + bitset_set (V, eoftoken->number); /* end-of-input token */ + bitset_set (V, errtoken->number); /* error token */ + bitset_set (V, undeftoken->number); /* some undefined token */ - XFREE (P); + bitset_free (P); P = Pp; - nuseful_productions = bits_size (P, WORDSIZE (nrules + 1)); + nuseful_productions = bitset_count (P); nuseless_productions = nrules - nuseful_productions; nuseful_nonterminals = 0; for (i = ntokens; i < nsyms; i++) - if (BITISSET (V, i)) + if (bitset_test (V, i)) nuseful_nonterminals++; nuseless_nonterminals = nvars - nuseful_nonterminals; /* A token that was used in %prec should not be warned about. */ - for (i = 1; i < nrules; i++) - if (rprecsym[i] != 0) - SETBIT (V1, rprecsym[i]); + for (i = 1; i < nrules + 1; i++) + if (rules[i].precsym != 0) + bitset_set (V1, rules[i].precsym->number); } + +/*-------------------------------------------------------------------. +| Put the useless productions at the end of RULES, and adjust NRULES | +| accordingly. | +`-------------------------------------------------------------------*/ + static void reduce_grammar_tables (void) { -/* This is turned off because we would need to change the numbers - in the case statements in the actions file. */ -#if 0 - /* remove useless productions */ - if (nuseless_productions > 0) - { - short np, pn, ni, pi; - - np = 0; - ni = 0; - for (pn = 1; pn <= nrules; pn++) - { - if (BITISSET (P, pn)) - { - np++; - if (pn != np) - { - rlhs[np] = rlhs[pn]; - rline[np] = rline[pn]; - rprec[np] = rprec[pn]; - rassoc[np] = rassoc[pn]; - rrhs[np] = rrhs[pn]; - if (rrhs[np] != ni) - { - pi = rrhs[np]; - rrhs[np] = ni; - while (ritem[pi] >= 0) - ritem[ni++] = ritem[pi++]; - ritem[ni++] = -np; - } - } - else - { - while (ritem[ni++] >= 0); - } - } - } - ritem[ni] = 0; - nrules -= nuseless_productions; - nitems = ni; - - /* Is it worth it to reduce the amount of memory for the - grammar? Probably not. */ - - } -#endif /* 0 */ - /* Disable useless productions, - since they may contain useless nonterms - that would get mapped below to -1 and confuse everyone. */ - if (nuseless_productions > 0) - { - int pn; + /* Flag useless productions. */ + { + int pn; + for (pn = 1; pn < nrules + 1; pn++) + rules[pn].useful = bitset_test (P, pn); + } + + /* Map the nonterminals to their new index: useful first, useless + afterwards. Kept for later report. */ + { + int useful = 1; + int useless = nrules + 1 - nuseless_productions; + rule_t *rules_sorted = XMALLOC (rule_t, nrules + 1) - 1; + int i; + for (i = 1; i < nrules + 1; ++i) + rules_sorted[rules[i].useful ? useful++ : useless++] = rules[i]; + free (rules + 1); + rules = rules_sorted; + + /* Renumber the rules markers in RITEMS. */ + for (i = 1; i < nrules + 1; ++i) + { + short *rhsp = rules[i].rhs; + for (/* Nothing. */; *rhsp >= 0; ++rhsp) + /* Nothing. */; + *rhsp = -i; + rules[i].number = i; + } + nrules -= nuseless_productions; + } + + /* Adjust NRITEMS and NITEMS. */ + { + int r; + int length; + for (r = nrules + 1; r < nrules + 1 + nuseless_productions; ++r) + { + length = rule_rhs_length (&rules[r]); + nritems -= length + 1; + } + } +} - for (pn = 1; pn <= nrules; pn++) - { - if (!BITISSET (P, pn)) - { - rlhs[pn] = -1; - } - } - } - /* remove useless symbols */ - if (nuseless_nonterminals > 0) - { +/*------------------------------. +| Remove useless nonterminals. | +`------------------------------*/ - int i, n; -/* short j; JF unused */ - short *nontermmap; - rule r; +static void +nonterminals_reduce (void) +{ + int i, n; - /* Create a map of nonterminal number to new nonterminal - number. -1 in the map means it was useless and is being - eliminated. */ + /* Map the nonterminals to their new index: useful first, useless + afterwards. Kept for later report. */ - nontermmap = XCALLOC (short, nvars) - ntokens; - for (i = ntokens; i < nsyms; i++) - nontermmap[i] = -1; + short *nontermmap = XCALLOC (short, nvars) - ntokens; + n = ntokens; + for (i = ntokens; i < nsyms; i++) + if (bitset_test (V, i)) + nontermmap[i] = n++; + for (i = ntokens; i < nsyms; i++) + if (!bitset_test (V, i)) + nontermmap[i] = n++; - n = ntokens; - for (i = ntokens; i < nsyms; i++) - if (BITISSET (V, i)) - nontermmap[i] = n++; - /* Shuffle elements of tables indexed by symbol number. */ + /* Shuffle elements of tables indexed by symbol number. */ + { + symbol_t **symbols_sorted = XMALLOC (symbol_t *, nvars) - ntokens; - for (i = ntokens; i < nsyms; i++) - { - n = nontermmap[i]; - if (n >= 0) - { - sassoc[n] = sassoc[i]; - sprec[n] = sprec[i]; - tags[n] = tags[i]; - } - else - { - free (tags[i]); - } - } + for (i = ntokens; i < nsyms; i++) + symbols[i]->number = nontermmap[i]; + for (i = ntokens; i < nsyms; i++) + symbols_sorted[nontermmap[i]] = symbols[i]; + for (i = ntokens; i < nsyms; i++) + symbols[i] = symbols_sorted[i]; + free (symbols_sorted + ntokens); + } - /* Replace all symbol numbers in valid data structures. */ + for (i = 0; i < nritems; ++i) + if (ISVAR (ritem[i])) + ritem[i] = nontermmap[ritem[i]]; - for (i = 1; i <= nrules; i++) - { - /* Ignore the rules disabled above. */ - if (rlhs[i] >= 0) - rlhs[i] = nontermmap[rlhs[i]]; - if (ISVAR (rprecsym[i])) - /* Can this happen? */ - rprecsym[i] = nontermmap[rprecsym[i]]; - } + start_symbol = nontermmap[start_symbol]; - for (r = ritem; *r; r++) - if (ISVAR (*r)) - *r = nontermmap[*r]; + nsyms -= nuseless_nonterminals; + nvars -= nuseless_nonterminals; - start_symbol = nontermmap[start_symbol]; + free (nontermmap + ntokens); +} - nsyms -= nuseless_nonterminals; - nvars -= nuseless_nonterminals; - free (&nontermmap[ntokens]); - } -} +/*------------------------------------------------------------------. +| Output the detailed results of the reductions. For FILE.output. | +`------------------------------------------------------------------*/ -static void -print_results (void) +void +reduce_output (FILE *out) { - int i; -/* short j; JF unused */ - rule r; - bool b; - if (nuseless_nonterminals > 0) { - obstack_sgrow (&output_obstack, _("Useless nonterminals:")); - obstack_sgrow (&output_obstack, "\n\n"); - for (i = ntokens; i < nsyms; i++) - if (!BITISSET (V, i)) - obstack_fgrow1 (&output_obstack, " %s\n", tags[i]); + int i; + fprintf (out, "%s\n\n", _("Useless nonterminals:")); + for (i = 0; i < nuseless_nonterminals; ++i) + fprintf (out, " %s\n", quotearg_style (escape_quoting_style, + symbols[nsyms + i]->tag)); + fputs ("\n\n", out); } - b = FALSE; - for (i = 0; i < ntokens; i++) - { - if (!BITISSET (V, i) && !BITISSET (V1, i)) + + { + bool b = FALSE; + int i; + for (i = 0; i < ntokens; i++) + if (!bitset_test (V, i) && !bitset_test (V1, i)) { if (!b) - { - obstack_sgrow (&output_obstack, "\n\n"); - obstack_sgrow (&output_obstack, - _("Terminals which are not used:")); - obstack_sgrow (&output_obstack, "\n\n"); - b = TRUE; - } - obstack_fgrow1 (&output_obstack, " %s\n", tags[i]); + fprintf (out, "%s\n\n", _("Terminals which are not used:")); + b = TRUE; + fprintf (out, " %s\n", quotearg_style (escape_quoting_style, + symbols[i]->tag)); } - } + if (b) + fputs ("\n\n", out); + } if (nuseless_productions > 0) { - obstack_sgrow (&output_obstack, "\n\n"); - obstack_sgrow (&output_obstack, _("Useless rules:")); - obstack_sgrow (&output_obstack, "\n\n"); - for (i = 1; i <= nrules; i++) + int i; + fprintf (out, "%s\n\n", _("Useless rules:")); + for (i = nrules + 1; i < nuseless_productions + nrules + 1; i++) { - if (!BITISSET (P, i)) - { - obstack_fgrow1 (&output_obstack, "#%-4d ", i); - obstack_fgrow1 (&output_obstack, "%s :\t", tags[rlhs[i]]); - for (r = &ritem[rrhs[i]]; *r >= 0; r++) - obstack_fgrow1 (&output_obstack, " %s", tags[*r]); - obstack_sgrow (&output_obstack, ";\n"); - } + rule r; + fprintf (out, "#%-4d ", rules[i].user_number - 1); + fprintf (out, "%s:", quotearg_style (escape_quoting_style, + rules[i].lhs->tag)); + for (r = rules[i].rhs; *r >= 0; r++) + fprintf (out, " %s", quotearg_style (escape_quoting_style, + symbols[*r]->tag)); + fputs (";\n", out); } + fputs ("\n\n", out); } - if (nuseless_nonterminals > 0 || nuseless_productions > 0 || b) - obstack_sgrow (&output_obstack, "\n\n"); } -#if 0 /* XXX currently unused. */ static void -dump_grammar (void) +dump_grammar (FILE *out) { int i; rule r; - obstack_fgrow5 (&output_obstack, - "ntokens = %d, nvars = %d, nsyms = %d, nrules = %d, nitems = %d\n\n", - ntokens, nvars, nsyms, nrules, nitems); - obstack_sgrow (&output_obstack, - _("Variables\n---------\n\n")); - obstack_sgrow (&output_obstack, - _("Value Sprec Sassoc Tag\n")); + fprintf (out, "REDUCED GRAMMAR\n\n"); + fprintf (out, + "ntokens = %d, nvars = %d, nsyms = %d, nrules = %d, nritems = %d\n\n", + ntokens, nvars, nsyms, nrules, nritems); + fprintf (out, "Variables\n---------\n\n"); + fprintf (out, "Value Sprec Sassoc Tag\n"); for (i = ntokens; i < nsyms; i++) - obstack_fgrow4 (&output_obstack, - "%5d %5d %5d %s\n", i, sprec[i], sassoc[i], tags[i]); - obstack_sgrow (&output_obstack, "\n\n"); - obstack_sgrow (&output_obstack, _("Rules\n-----\n\n")); - for (i = 1; i <= nrules; i++) + fprintf (out, "%5d %5d %5d %s\n", + i, + symbols[i]->prec, symbols[i]->assoc, + quotearg_style (escape_quoting_style, symbols[i]->tag)); + fprintf (out, "\n\n"); + fprintf (out, "Rules\n-----\n\n"); + fprintf (out, "Num (Prec, Assoc, Useful, Ritem Range) Lhs -> Rhs (Ritem range) [Num]\n"); + for (i = 1; i < nrules + nuseless_productions + 1; i++) { - obstack_fgrow5 (&output_obstack, "%-5d(%5d%5d)%5d : (@%-5d)", - i, rprec[i], rassoc[i], rlhs[i], rrhs[i]); - for (r = &ritem[rrhs[i]]; *r > 0; r++) - obstack_fgrow1 (&output_obstack, "%5d", *r); - obstack_fgrow1 (&output_obstack, " [%d]\n", -(*r)); + int rhs_count = 0; + /* Find the last RHS index in ritems. */ + for (r = rules[i].rhs; *r >= 0; ++r) + ++rhs_count; + fprintf (out, "%3d (%2d, %2d, %2d, %2d-%2d) %2d ->", + i - 1, + rules[i].prec->prec, + rules[i].prec->assoc, + rules[i].useful, + rules[i].rhs - ritem, + rules[i].rhs - ritem + rhs_count - 1, + rules[i].lhs->number); + /* Dumped the RHS. */ + for (r = rules[i].rhs; *r >= 0; r++) + fprintf (out, "%3d", *r); + fprintf (out, " [%d]\n", -(*r) - 1); } - obstack_sgrow (&output_obstack, "\n\n"); - obstack_sgrow (&output_obstack, - _("Rules interpreted\n-----------------\n\n")); - for (i = 1; i <= nrules; i++) + fprintf (out, "\n\n"); + fprintf (out, "Rules interpreted\n-----------------\n\n"); + for (i = 1; i < nrules + nuseless_productions + 1; i++) { - obstack_fgrow2 (&output_obstack, "%-5d %s :", i, tags[rlhs[i]]); - for (r = &ritem[rrhs[i]]; *r > 0; r++) - obstack_fgrow1 (&output_obstack, " %s", tags[*r]); - obstack_grow1 (&output_obstack, '\n'); + fprintf (out, "%-5d %s :", + i, quotearg_style (escape_quoting_style, rules[i].lhs->tag)); + for (r = rules[i].rhs; *r >= 0; r++) + fprintf (out, " %s", + quotearg_style (escape_quoting_style, symbols[*r]->tag)); + fputc ('\n', out); } - obstack_sgrow (&output_obstack, "\n\n"); + fprintf (out, "\n\n"); } -#endif + +/*-------------------------------. +| Report the results to STDERR. | +`-------------------------------*/ + static void -print_notices (void) +reduce_print (void) { if (yacc_flag && nuseless_productions) - fprintf (stderr, _("%d rules never reduced\n"), nuseless_productions); + fprintf (stderr, ngettext ("%d rule never reduced\n", + "%d rules never reduced\n", + nuseless_productions), + nuseless_productions); fprintf (stderr, _("%s contains "), infile); if (nuseless_nonterminals > 0) - { - fprintf (stderr, _("%d useless nonterminal%s"), - nuseless_nonterminals, - (nuseless_nonterminals == 1 ? "" : "s")); - } + fprintf (stderr, ngettext ("%d useless nonterminal", + "%d useless nonterminals", + nuseless_nonterminals), + nuseless_nonterminals); + if (nuseless_nonterminals > 0 && nuseless_productions > 0) fprintf (stderr, _(" and ")); if (nuseless_productions > 0) - { - fprintf (stderr, _("%d useless rule%s"), - nuseless_productions, (nuseless_productions == 1 ? "" : "s")); - } + fprintf (stderr, ngettext ("%d useless rule", + "%d useless rules", + nuseless_productions), + nuseless_productions); fprintf (stderr, "\n"); fflush (stderr); } @@ -517,52 +469,49 @@ reduce_grammar (void) /* Allocate the global sets used to compute the reduced grammar */ - N = XCALLOC (unsigned, WORDSIZE (nvars)); - P = XCALLOC (unsigned, WORDSIZE (nrules + 1)); - V = XCALLOC (unsigned, WORDSIZE (nsyms)); - V1 = XCALLOC (unsigned, WORDSIZE (nsyms)); + N = bitset_create (nvars, BITSET_FIXED); + P = bitset_create (nrules + 1, BITSET_FIXED); + V = bitset_create (nsyms, BITSET_FIXED); + V1 = bitset_create (nsyms, BITSET_FIXED); useless_nonterminals (); inaccessable_symbols (); reduced = (bool) (nuseless_nonterminals + nuseless_productions > 0); + if (!reduced) + return; - if (verbose_flag) - print_results (); - - if (reduced == FALSE) - goto done_reducing; + reduce_print (); - print_notices (); - - if (!BITISSET (N, start_symbol - ntokens)) + if (!bitset_test (N, start_symbol - ntokens)) fatal (_("Start symbol %s does not derive any sentence"), - tags[start_symbol]); + quotearg_style (escape_quoting_style, symbols[start_symbol]->tag)); - reduce_grammar_tables (); -#if 0 - if (verbose_flag) + if (nuseless_productions > 0) + reduce_grammar_tables (); + if (nuseless_nonterminals > 0) + nonterminals_reduce (); + + if (trace_flag) { - obstack_fgrow1 (&output_obstack, "REDUCED GRAMMAR\n\n"); - dump_grammar (); + dump_grammar (stderr); + + fprintf (stderr, "reduced %s defines %d terminals, %d nonterminals\ +, and %d productions.\n", + infile, ntokens, nvars, nrules); } -#endif - - if (statistics_flag) - fprintf (stderr, _("reduced %s defines %d terminal%s, %d nonterminal%s\ -, and %d production%s.\n"), - infile, - ntokens, - (ntokens == 1 ? "" : "s"), - nvars, - (nvars == 1 ? "" : "s"), - nrules, - (nrules == 1 ? "" : "s")); - -done_reducing: - /* Free the global sets used to compute the reduced grammar */ - - XFREE (N); - XFREE (V); - XFREE (P); +} + + +/*-----------------------------------------------------------. +| Free the global sets used to compute the reduced grammar. | +`-----------------------------------------------------------*/ + +void +reduce_free (void) +{ + bitset_free (N); + bitset_free (V); + bitset_free (V1); + bitset_free (P); }