X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/2c5f66eda72bb8b158bccf0a528d9dbb2c839f0c..914feea9d07b0a53ee958e8df755addb97196913:/src/reduce.c diff --git a/src/reduce.c b/src/reduce.c index b179f608..bef2ae90 100644 --- a/src/reduce.c +++ b/src/reduce.c @@ -28,62 +28,34 @@ #include "system.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 +64,7 @@ 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 +72,9 @@ 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[rule_table[i].rhs]; *r > 0; r++) - if (ISVAR (n = *r)) - if (!BITISSET (N0, n - ntokens)) - return FALSE; + for (r = &ritem[rules[i].rhs]; *r >= 0; r++) + if (ISVAR (n = *r) && !bitset_test (N0, n - ntokens)) + return FALSE; return TRUE; } @@ -115,13 +86,14 @@ 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 +113,21 @@ useless_nonterminals (void) while (1) { - for (i = WORDSIZE (nvars) - 1; i >= 0; i--) - Np[i] = N[i]; + bitset_copy (Np, N); for (i = 1; i <= nrules; i++) - { - if (!BITISSET (P, i)) - { - if (useful_production (i, N)) - { - SETBIT (Np, rule_table[i].lhs - ntokens); - SETBIT (P, i); - } - } - } - if (bits_equal (N, Np, WORDSIZE (nvars))) + if (!bitset_test (P, i) + && useful_production (i, N)) + { + bitset_set (Np, rules[i].lhs - 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 +135,7 @@ useless_nonterminals (void) static void inaccessable_symbols (void) { - BSet Vp, Vs, Pp; + bitset Vp, Vs, Pp; int i; short t; rule r; @@ -196,31 +163,30 @@ 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)) + if (bitset_test (N, start_symbol - ntokens)) { - SETBIT (V, start_symbol); + bitset_set (V, start_symbol); while (1) { - for (i = WORDSIZE (nsyms) - 1; i >= 0; i--) - Vp[i] = V[i]; + bitset_copy (Vp, V); for (i = 1; i <= nrules; i++) { - if (!BITISSET (Pp, i) - && BITISSET (P, i) - && BITISSET (V, rule_table[i].lhs)) + if (!bitset_test (Pp, i) + && bitset_test (P, i) + && bitset_test (V, rules[i].lhs)) { - for (r = &ritem[rule_table[i].rhs]; *r >= 0; r++) - if (ISTOKEN (t = *r) || BITISSET (N, t - ntokens)) - SETBIT (Vp, t); - SETBIT (Pp, i); + for (r = &ritem[rules[i].rhs]; *r >= 0; r++) + if (ISTOKEN (t = *r) || bitset_test (N, t - ntokens)) + bitset_set (Vp, t); + bitset_set (Pp, i); } } - if (bits_equal (V, Vp, WORDSIZE (nsyms))) + if (bitset_equal_p (V, Vp)) break; Vs = Vp; Vp = V; @@ -228,61 +194,68 @@ inaccessable_symbols (void) } } - 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, 0); /* end-of-input token */ + bitset_set (V, 1); /* error token */ + bitset_set (V, 2); /* 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 (rule_table[i].precsym != 0) - SETBIT (V1, rule_table[i].precsym); + if (rules[i].precsym != 0) + bitset_set (V1, rules[i].precsym); } 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)) + /* This is turned off because we would need to change the numbers in + the case statements in the actions file. + + We don't disable it via CPP so that it is still checked with the + rest of the code, to avoid its becoming completely obsolete. + + FIXME: I think the comment above demonstrates this code must be + turned off for *semantic* parser, not in the general case. Try + to understand this better --akim. */ + + 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 (bitset_test (P, pn)) { np++; if (pn != np) { - rule_table[np].lhs = rule_table[pn].lhs; - rline[np] = rline[pn]; - rule_table[np].prec = rule_table[pn].prec; - rule_table[np].assoc = rule_table[pn].assoc; - rule_table[np].rhs = rule_table[pn].rhs; - if (rule_table[np].rhs != ni) + rules[np].lhs = rules[pn].lhs; + rules[np].line = rules[pn].line; + rules[np].prec = rules[pn].prec; + rules[np].assoc = rules[pn].assoc; + rules[np].rhs = rules[pn].rhs; + if (rules[np].rhs != ni) { - pi = rule_table[np].rhs; - rule_table[np].rhs = ni; + pi = rules[np].rhs; + rules[np].rhs = ni; while (ritem[pi] >= 0) ritem[ni++] = ritem[pi++]; ritem[ni++] = -np; @@ -290,155 +263,135 @@ reduce_grammar_tables (void) } else { - while (ritem[ni++] >= 0); + while (ritem[ni++] >= 0) + /* Nothing. */; } } - } - ritem[ni] = 0; - nrules -= nuseless_productions; - nitems = ni; - /* Is it worth it to reduce the amount of memory for the - grammar? Probably not. */ + ritem[ni] = 0; + nrules -= nuseless_productions; + nitems = ni; + nritems = ni; - } -#endif /* 0 */ - /* Disable useless productions, - since they may contain useless nonterms - that would get mapped below to -1 and confuse everyone. */ + /* Is it worth it to reduce the amount of memory for the + grammar? Probably not. */ + } + + /* Disable useless productions. */ if (nuseless_productions > 0) { int pn; - for (pn = 1; pn <= nrules; pn++) - { - if (!BITISSET (P, pn)) - { - rule_table[pn].lhs = -1; - } - } + rules[pn].useful = bitset_test (P, pn); } +} - /* remove useless symbols */ - if (nuseless_nonterminals > 0) - { - int i, n; -/* short j; JF unused */ - short *nontermmap; - rule r; +/*------------------------------. +| Remove useless nonterminals. | +`------------------------------*/ - /* Create a map of nonterminal number to new nonterminal - number. -1 in the map means it was useless and is being - eliminated. */ +static void +nonterminals_reduce (void) +{ + int i, n; - nontermmap = XCALLOC (short, nvars) - ntokens; - for (i = ntokens; i < nsyms; i++) - nontermmap[i] = -1; + /* Map the nonterminals to their new index: useful first, useless + afterwards. Kept for later report. */ - n = ntokens; - for (i = ntokens; i < nsyms; i++) - if (BITISSET (V, i)) - nontermmap[i] = n++; + 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++; - /* Shuffle elements of tables indexed by symbol number. */ - 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]); - } - } + /* Shuffle elements of tables indexed by symbol number. */ + { + bucket **symbols_sorted = XMALLOC (bucket *, nvars) - ntokens; - /* Replace all symbol numbers in valid data structures. */ + 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); + } - for (i = 1; i <= nrules; i++) - { - /* Ignore the rules disabled above. */ - if (rule_table[i].lhs >= 0) - rule_table[i].lhs = nontermmap[rule_table[i].lhs]; - if (ISVAR (rule_table[i].precsym)) - /* Can this happen? */ - rule_table[i].precsym = nontermmap[rule_table[i].precsym]; - } + /* Replace all symbol numbers in valid data structures. */ + + for (i = 1; i <= nrules; i++) + { + rules[i].lhs = nontermmap[rules[i].lhs]; + if (ISVAR (rules[i].precsym)) + /* Can this happen? */ + rules[i].precsym = nontermmap[rules[i].precsym]; + } - for (r = ritem; *r; r++) - if (ISVAR (*r)) - *r = nontermmap[*r]; + for (i = 0; i < nritems; ++i) + if (ISVAR (ritem[i])) + ritem[i] = nontermmap[ritem[i]]; - start_symbol = nontermmap[start_symbol]; + start_symbol = nontermmap[start_symbol]; - nsyms -= nuseless_nonterminals; - nvars -= nuseless_nonterminals; + nsyms -= nuseless_nonterminals; + nvars -= nuseless_nonterminals; - free (&nontermmap[ntokens]); - } + free (nontermmap + ntokens); } -/*-----------------------------------------------------------------. -| Ouput the detailed results of the reductions. For FILE.output. | -`-----------------------------------------------------------------*/ +/*------------------------------------------------------------------. +| Output the detailed results of the reductions. For FILE.output. | +`------------------------------------------------------------------*/ void reduce_output (FILE *out) { - int i; - rule r; - bool b; - if (nuseless_nonterminals > 0) { - fprintf (out, _("Useless nonterminals:")); - fprintf (out, "\n\n"); - for (i = ntokens; i < nsyms; i++) - if (!BITISSET (V, i)) - fprintf (out, " %s\n", tags[i]); + int i; + fprintf (out, "%s\n\n", _("Useless nonterminals:")); + 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 (!bitset_test (V, i) && !bitset_test (V1, i)) { if (!b) - { - fprintf (out, "\n\n"); - fprintf (out, _("Terminals which are not used:")); - fprintf (out, "\n\n"); - b = TRUE; - } - fprintf (out, " %s\n", tags[i]); + fprintf (out, "%s\n\n", _("Terminals which are not used:")); + b = TRUE; + fprintf (out, " %s\n", symbols[i]->tag); } - } + if (b) + fputs ("\n\n", out); + } if (nuseless_productions > 0) { - fprintf (out, "\n\n"); - fprintf (out, _("Useless rules:")); - fprintf (out, "\n\n"); + int i; + fprintf (out, "%s\n\n", _("Useless rules:")); for (i = 1; i <= nrules; i++) - if (!BITISSET (P, i)) + if (!rules[i].useful) { - fprintf (out, "#%-4d ", i); - fprintf (out, "%s :\t", tags[rule_table[i].lhs]); - for (r = &ritem[rule_table[i].rhs]; *r >= 0; r++) - fprintf (out, " %s", tags[*r]); - fprintf (out, ";\n"); + rule r; + fprintf (out, "#%-4d ", i - 1); + fprintf (out, "%s:", symbols[rules[i].lhs]->tag); + for (r = &ritem[rules[i].rhs]; *r >= 0; r++) + fprintf (out, " %s", symbols[*r]->tag); + fputs (";\n", out); } + fputs ("\n\n", out); } - if (nuseless_nonterminals > 0 || nuseless_productions > 0 || b) - fprintf (out, "\n\n"); } -#if TRACE static void dump_grammar (FILE *out) { @@ -447,38 +400,44 @@ dump_grammar (FILE *out) fprintf (out, "REDUCED GRAMMAR\n\n"); fprintf (out, - "ntokens = %d, nvars = %d, nsyms = %d, nrules = %d, nitems = %d\n\n", - ntokens, nvars, nsyms, nrules, nitems); - fprintf (out, _("Variables\n---------\n\n")); - fprintf (out, _("Value Sprec Sassoc Tag\n")); + "ntokens = %d, nvars = %d, nsyms = %d, nrules = %d, nitems = %d\n\n", + ntokens, nvars, nsyms, nrules, nitems); + fprintf (out, "Variables\n---------\n\n"); + fprintf (out, "Value Sprec Sassoc Tag\n"); for (i = ntokens; i < nsyms; i++) - fprintf (out, "%5d %5d %5d %s\n", i, sprec[i], sassoc[i], tags[i]); + fprintf (out, "%5d %5d %5d %s\n", + i, + symbols[i]->prec, symbols[i]->assoc, symbols[i]->tag); fprintf (out, "\n\n"); - fprintf (out, _("Rules\n-----\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; i++) { - fprintf (out, "%-5d(%5d%5d)%5d : (@%-5d)", - i, - rule_table[i].prec, - rule_table[i].assoc, - rule_table[i].lhs, - rule_table[i].rhs); - for (r = &ritem[rule_table[i].rhs]; *r > 0; r++) - fprintf (out, "%5d", *r); - fprintf (out, " [%d]\n", -(*r)); + int rhs_count = 0; + /* Find the last RHS index in ritems. */ + for (r = &ritem[rules[i].rhs]; *r >= 0; ++r) + ++rhs_count; + fprintf (out, "%3d (%2d, %2d, %2d, %2d-%2d) %2d ->", + i - 1, + rules[i].prec, rules[i].assoc, rules[i].useful, + rules[i].rhs, rules[i].rhs + rhs_count - 1, + rules[i].lhs); + /* Dumped the RHS. */ + for (r = &ritem[rules[i].rhs]; *r >= 0; r++) + fprintf (out, "%3d", *r); + fprintf (out, " [%d]\n", -(*r) - 1); } fprintf (out, "\n\n"); - fprintf (out, _("Rules interpreted\n-----------------\n\n")); + fprintf (out, "Rules interpreted\n-----------------\n\n"); for (i = 1; i <= nrules; i++) { - fprintf (out, "%-5d %s :", i, tags[rule_table[i].lhs]); - for (r = &ritem[rule_table[i].rhs]; *r > 0; r++) - fprintf (out, " %s", tags[*r]); + fprintf (out, "%-5d %s :", i, symbols[rules[i].lhs]->tag); + for (r = &ritem[rules[i].rhs]; *r >= 0; r++) + fprintf (out, " %s", symbols[*r]->tag); fputc ('\n', out); } fprintf (out, "\n\n"); } -#endif @@ -490,24 +449,27 @@ static 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); } @@ -519,10 +481,10 @@ 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 (); @@ -534,25 +496,22 @@ reduce_grammar (void) reduce_print (); - 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]); + symbols[start_symbol]->tag); reduce_grammar_tables (); -#if TRACE - dump_grammar (stderr); -#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")); + if (nuseless_nonterminals > 0) + nonterminals_reduce (); + + if (trace_flag) + { + dump_grammar (stderr); + + fprintf (stderr, "reduced %s defines %d terminals, %d nonterminals\ +, and %d productions.\n", + infile, ntokens, nvars, nrules); + } } @@ -563,8 +522,8 @@ reduce_grammar (void) void reduce_free (void) { - XFREE (N); - XFREE (V); - XFREE (V1); - XFREE (P); + bitset_free (N); + bitset_free (V); + bitset_free (V1); + bitset_free (P); }