X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/b2ca4022ea6c61e10ac87a0a92bb0f998ecf8b48..be3517b007a833ffec1735691f256f6a37e4a67f:/src/reduce.c diff --git a/src/reduce.c b/src/reduce.c index 8175f29f..5338393c 100644 --- a/src/reduce.c +++ b/src/reduce.c @@ -1,22 +1,22 @@ /* Grammar reduction for Bison. - Copyright (C) 1988, 1989, 2000 Free Software Foundation, Inc. + + Copyright (C) 1988-1989, 2000-2003, 2005-2012 Free Software + Foundation, Inc. This file is part of Bison, the GNU Compiler Compiler. - Bison is free software; you can redistribute it and/or modify + This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by - the Free Software Foundation; either version 2, or (at your option) - any later version. + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. - Bison is distributed in the hope that it will be useful, + This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License - along with Bison; see the file COPYING. If not, write to - the Free Software Foundation, Inc., 59 Temple Place - Suite 330, - Boston, MA 02111-1307, USA. */ + along with this program. If not, see . */ /* Reduce the grammar: Find and eliminate unreachable terminals, @@ -25,66 +25,37 @@ /* Don't eliminate unreachable terminals: They may be used by the user's parser. */ +#include #include "system.h" -#include "getargs.h" + +#include + +#include "complain.h" #include "files.h" +#include "getargs.h" #include "gram.h" -#include "alloc.h" -#include "complain.h" -#include "reduce.h" +#include "print-xml.h" #include "reader.h" -#include "getargs.h" +#include "reduce.h" +#include "symtab.h" -typedef unsigned *BSet; -typedef short *rule; +/* Set of all nonterminals which are not useless. */ +static bitset N; +/* Set of all rules which have no useless nonterminals in their RHS. */ +static bitset P; -/* 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 accessible symbols. */ +static bitset V; -static BSet N, P, V, V1; +/* 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 rule_number nuseful_productions; +rule_number 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; -} +symbol_number nuseless_nonterminals; /*-------------------------------------------------------------------. | Another way to do this would be with a set for each production and | @@ -93,19 +64,17 @@ bits_size (BSet S, int n) `-------------------------------------------------------------------*/ static bool -useful_production (int i, BSet N0) +useful_production (rule_number r, bitset N0) { - rule r; - short n; + item_number *rhsp; /* 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; - return TRUE; + for (rhsp = rules[r].rhs; *rhsp >= 0; ++rhsp) + if (ISVAR (*rhsp) && !bitset_test (N0, *rhsp - ntokens)) + return false; + return true; } @@ -116,13 +85,14 @@ useful_production (int i, BSet N0) static void useless_nonterminals (void) { - BSet Np, Ns; - int i, n; + bitset Np, Ns; + rule_number r; /* 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 = NEW2 (WORDSIZE (nvars), unsigned); + 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 @@ -140,29 +110,23 @@ useless_nonterminals (void) saved to be used when finding useful productions: only productions in this set will appear in the final grammar. */ - n = 0; 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 (r = 0; r < nrules; r++) + if (!bitset_test (P, r) + && useful_production (r, N)) + { + bitset_set (Np, rules[r].lhs->number - ntokens); + bitset_set (P, r); + } + if (bitset_equal_p (N, Np)) break; Ns = Np; Np = N; N = Ns; } - FREE (N); + bitset_free (N); N = Np; } @@ -170,10 +134,7 @@ useless_nonterminals (void) static void inaccessable_symbols (void) { - BSet Vp, Vs, Pp; - int i, n; - short t; - rule r; + bitset Vp, Vs, Pp; /* Find out which productions are reachable and which symbols are used. Starting with an empty set of productions and a set of @@ -198,312 +159,238 @@ inaccessable_symbols (void) terminals are printed (if running in verbose mode) so that the user can know. */ - Vp = NEW2 (WORDSIZE (nsyms), unsigned); - Pp = NEW2 (WORDSIZE (nrules + 1), unsigned); + Vp = bitset_create (nsyms, BITSET_FIXED); + Pp = bitset_create (nrules, 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); - - n = 0; - while (1) + if (bitset_test (N, accept->number - ntokens)) { - for (i = WORDSIZE (nsyms) - 1; i >= 0; i--) - Vp[i] = V[i]; - for (i = 1; i <= nrules; i++) + bitset_set (V, accept->number); + + while (1) { - if (!BITISSET (Pp, i) && BITISSET (P, i) && BITISSET (V, rlhs[i])) + rule_number r; + bitset_copy (Vp, V); + for (r = 0; r < nrules; r++) { - for (r = &ritem[rrhs[i]]; *r >= 0; r++) + if (!bitset_test (Pp, r) + && bitset_test (P, r) + && bitset_test (V, rules[r].lhs->number)) { - if (ISTOKEN (t = *r) || BITISSET (N, t - ntokens)) - { - SETBIT (Vp, t); - } + item_number *rhsp; + for (rhsp = rules[r].rhs; *rhsp >= 0; rhsp++) + if (ISTOKEN (*rhsp) || bitset_test (N, *rhsp - ntokens)) + bitset_set (Vp, *rhsp); + bitset_set (Pp, r); } - 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: - FREE (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, endtoken->number); /* end-of-input token */ + bitset_set (V, errtoken->number); /* error token */ + bitset_set (V, undeftoken->number); /* some undefined token */ - FREE (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)) - nuseful_nonterminals++; + { + symbol_number i; + for (i = ntokens; i < nsyms; 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]); + { + rule_number r; + for (r = 0; r < nrules; ++r) + if (rules[r].precsym != 0) + bitset_set (V1, rules[r].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; - - for (pn = 1; pn <= nrules; pn++) - { - if (!BITISSET (P, pn)) - { - rlhs[pn] = -1; - } - } - } - - /* remove useless symbols */ - if (nuseless_nonterminals > 0) - { - - int i, n; -/* short j; JF unused */ - short *nontermmap; - rule r; - - /* Create a map of nonterminal number to new nonterminal - number. -1 in the map means it was useless and is being - eliminated. */ - - nontermmap = NEW2 (nvars, short) - ntokens; - for (i = ntokens; i < nsyms; i++) - nontermmap[i] = -1; - - n = ntokens; - for (i = ntokens; i < nsyms; i++) - if (BITISSET (V, i)) - nontermmap[i] = n++; - - /* Shuffle elements of tables indexed by symbol number. */ + /* Report and flag useless productions. */ + { + rule_number r; + for (r = 0; r < nrules; r++) + rules[r].useful = bitset_test (P, r); + grammar_rules_useless_report (_("rule useless in grammar")); + } + + /* Map the nonterminals to their new index: useful first, useless + afterwards. Kept for later report. */ + { + int useful = 0; + int useless = nrules - nuseless_productions; + rule *rules_sorted = xnmalloc (nrules, sizeof *rules_sorted); + rule_number r; + for (r = 0; r < nrules; ++r) + rules_sorted[rules[r].useful ? useful++ : useless++] = rules[r]; + free (rules); + rules = rules_sorted; + + /* Renumber the rules markers in RITEMS. */ + for (r = 0; r < nrules; ++r) + { + item_number *rhsp = rules[r].rhs; + for (/* Nothing. */; *rhsp >= 0; ++rhsp) + /* Nothing. */; + *rhsp = rule_number_as_item_number (r); + rules[r].number = r; + } + nrules -= nuseless_productions; + } + + /* Adjust NRITEMS. */ + { + rule_number r; + int length; + for (r = nrules; r < nrules + nuseless_productions; ++r) + { + length = rule_rhs_length (&rules[r]); + nritems -= length + 1; + } + } +} - 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]); - } - } - /* Replace all symbol numbers in valid data structures. */ +/*------------------------------. +| Remove useless nonterminals. | +`------------------------------*/ - 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]]; - } +static void +nonterminals_reduce (void) +{ + symbol_number i, n; - for (r = ritem; *r; r++) - if (ISVAR (*r)) - *r = nontermmap[*r]; + /* Map the nonterminals to their new index: useful first, useless + afterwards. Kept for later report. */ - start_symbol = nontermmap[start_symbol]; + symbol_number *nontermmap = xnmalloc (nvars, sizeof *nontermmap); + n = ntokens; + for (i = ntokens; i < nsyms; i++) + if (bitset_test (V, i)) + nontermmap[i - ntokens] = n++; + for (i = ntokens; i < nsyms; i++) + if (!bitset_test (V, i)) + { + nontermmap[i - ntokens] = n++; + warn_at (symbols[i]->location, _("nonterminal useless in grammar: %s"), + symbols[i]->tag); + } + + + /* Shuffle elements of tables indexed by symbol number. */ + { + symbol **symbols_sorted = xnmalloc (nvars, sizeof *symbols_sorted); + + for (i = ntokens; i < nsyms; i++) + symbols[i]->number = nontermmap[i - ntokens]; + for (i = ntokens; i < nsyms; i++) + symbols_sorted[nontermmap[i - ntokens] - ntokens] = symbols[i]; + for (i = ntokens; i < nsyms; i++) + symbols[i] = symbols_sorted[i - ntokens]; + free (symbols_sorted); + } + + { + rule_number r; + for (r = 0; r < nrules; ++r) + { + item_number *rhsp; + for (rhsp = rules[r].rhs; *rhsp >= 0; ++rhsp) + if (ISVAR (*rhsp)) + *rhsp = symbol_number_as_item_number (nontermmap[*rhsp + - ntokens]); + } + accept->number = nontermmap[accept->number - ntokens]; + } + + nsyms -= nuseless_nonterminals; + nvars -= nuseless_nonterminals; + + free (nontermmap); +} - 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) { - fprintf (foutput, _("Useless nonterminals:\n\n")); - for (i = ntokens; i < nsyms; i++) - if (!BITISSET (V, i)) - fprintf (foutput, " %s\n", tags[i]); + int i; + fprintf (out, "%s\n\n", _("Nonterminals useless in grammar")); + for (i = 0; i < nuseless_nonterminals; ++i) + fprintf (out, " %s\n", 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 (reduce_token_unused_in_grammar (i)) { if (!b) - { - fprintf (foutput, _("\n\nTerminals which are not used:\n\n")); - b = TRUE; - } - fprintf (foutput, " %s\n", tags[i]); + fprintf (out, "%s\n\n", _("Terminals unused in grammar")); + b = true; + fprintf (out, " %s\n", symbols[i]->tag); } - } + if (b) + fputs ("\n\n", out); + } if (nuseless_productions > 0) - { - fprintf (foutput, _("\n\nUseless rules:\n\n")); - for (i = 1; i <= nrules; i++) - { - if (!BITISSET (P, i)) - { - fprintf (foutput, "#%-4d ", i); - fprintf (foutput, "%s :\t", tags[rlhs[i]]); - for (r = &ritem[rrhs[i]]; *r >= 0; r++) - { - fprintf (foutput, " %s", tags[*r]); - } - fprintf (foutput, ";\n"); - } - } - } - if (nuseless_nonterminals > 0 || nuseless_productions > 0 || b) - fprintf (foutput, "\n\n"); + grammar_rules_partial_print (out, _("Rules useless in grammar"), + rule_useless_in_grammar_p); } -#if 0 /* XXX currently unused. */ -static void -dump_grammar (void) -{ - int i; - rule r; - - fprintf (foutput, - "ntokens = %d, nvars = %d, nsyms = %d, nrules = %d, nitems = %d\n\n", - ntokens, nvars, nsyms, nrules, nitems); - fprintf (foutput, _("Variables\n---------\n\n")); - fprintf (foutput, _("Value Sprec Sassoc Tag\n")); - for (i = ntokens; i < nsyms; i++) - fprintf (foutput, "%5d %5d %5d %s\n", i, sprec[i], sassoc[i], tags[i]); - fprintf (foutput, "\n\n"); - fprintf (foutput, _("Rules\n-----\n\n")); - for (i = 1; i <= nrules; i++) - { - fprintf (foutput, "%-5d(%5d%5d)%5d : (@%-5d)", - i, rprec[i], rassoc[i], rlhs[i], rrhs[i]); - for (r = &ritem[rrhs[i]]; *r > 0; r++) - fprintf (foutput, "%5d", *r); - fprintf (foutput, " [%d]\n", -(*r)); - } - fprintf (foutput, "\n\n"); - fprintf (foutput, _("Rules interpreted\n-----------------\n\n")); - for (i = 1; i <= nrules; i++) - { - fprintf (foutput, "%-5d %s :", i, tags[rlhs[i]]); - for (r = &ritem[rrhs[i]]; *r > 0; r++) - fprintf (foutput, " %s", tags[*r]); - fprintf (foutput, "\n"); - } - fprintf (foutput, "\n\n"); -} - -#endif +/*-------------------------------. +| Report the results to STDERR. | +`-------------------------------*/ static void -print_notices (void) +reduce_print (void) { - if (fixed_outfiles && nuseless_productions) - fprintf (stderr, _("%d rules never reduced\n"), nuseless_productions); - - fprintf (stderr, _("%s contains "), infile); - if (nuseless_nonterminals > 0) - { - fprintf (stderr, _("%d useless nonterminal%s"), - nuseless_nonterminals, - (nuseless_nonterminals == 1 ? "" : "s")); - } - if (nuseless_nonterminals > 0 && nuseless_productions > 0) - fprintf (stderr, _(" and ")); - + warn (ngettext ("%d nonterminal useless in grammar", + "%d nonterminals useless in grammar", + nuseless_nonterminals), + nuseless_nonterminals); if (nuseless_productions > 0) - { - fprintf (stderr, _("%d useless rule%s"), - nuseless_productions, (nuseless_productions == 1 ? "" : "s")); - } - fprintf (stderr, "\n"); - fflush (stderr); + warn (ngettext ("%d rule useless in grammar", + "%d rules useless in grammar", + nuseless_productions), + nuseless_productions); } void @@ -513,52 +400,66 @@ reduce_grammar (void) /* Allocate the global sets used to compute the reduced grammar */ - N = NEW2 (WORDSIZE (nvars), unsigned); - P = NEW2 (WORDSIZE (nrules + 1), unsigned); - V = NEW2 (WORDSIZE (nsyms), unsigned); - V1 = NEW2 (WORDSIZE (nsyms), unsigned); + N = bitset_create (nvars, BITSET_FIXED); + P = bitset_create (nrules, 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 (verboseflag) - print_results (); + reduced = (nuseless_nonterminals + nuseless_productions > 0); + if (!reduced) + return; - if (reduced == FALSE) - goto done_reducing; + reduce_print (); - print_notices (); + if (!bitset_test (N, accept->number - ntokens)) + fatal_at (startsymbol_location, + _("start symbol %s does not derive any sentence"), + startsymbol->tag); - if (!BITISSET (N, start_symbol - ntokens)) - fatal (_("Start symbol %s does not derive any sentence"), - tags[start_symbol]); + /* First reduce the nonterminals, as they renumber themselves in the + whole grammar. If you change the order, nonterms would be + renumbered only in the reduced grammar. */ + if (nuseless_nonterminals > 0) + nonterminals_reduce (); + if (nuseless_productions > 0) + reduce_grammar_tables (); - reduce_grammar_tables (); -#if 0 - if (verboseflag) + if (trace_flag & trace_grammar) { - fprintf (foutput, "REDUCED GRAMMAR\n\n"); - dump_grammar (); + grammar_dump (stderr, "Reduced Grammar"); + + fprintf (stderr, "reduced %s defines %d terminals, %d nonterminals\ +, and %d productions.\n", + grammar_file, ntokens, nvars, nrules); } -#endif - - if (statisticsflag) - 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 */ - - FREE (N); - FREE (V); - FREE (P); +} + +bool +reduce_token_unused_in_grammar (symbol_number i) +{ + aver (i < ntokens); + return !bitset_test (V, i) && !bitset_test (V1, i); +} + +bool +reduce_nonterminal_useless_in_grammar (symbol_number i) +{ + aver (ntokens <= i && i < nsyms + nuseless_nonterminals); + return nsyms <= i; +} + +/*-----------------------------------------------------------. +| 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); }