X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/11652ab3dc44bf12d5fa9095d1495b5cd3102ff7..c02634922181a18cd6f2d430901a7747ea17b12b:/src/reduce.c diff --git a/src/reduce.c b/src/reduce.c index 58bf43ea..35aab2e7 100644 --- a/src/reduce.c +++ b/src/reduce.c @@ -1,5 +1,5 @@ /* Grammar reduction for Bison. - Copyright 1988, 1989, 2000, 2001 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. @@ -26,6 +26,7 @@ user's parser. */ #include "system.h" +#include "quotearg.h" #include "getargs.h" #include "files.h" #include "symtab.h" @@ -36,9 +37,6 @@ #include "getargs.h" #include "bitset.h" -typedef short *rule; - - /* Set of all nonterminals which are not useless. */ static bitset N; @@ -53,7 +51,7 @@ static bitset V; static bitset V1; static int nuseful_productions; -static int nuseless_productions; +int nuseless_productions; static int nuseful_nonterminals; int nuseless_nonterminals; @@ -66,7 +64,7 @@ int nuseless_nonterminals; static bool useful_production (int i, bitset N0) { - rule r; + item_number_t *r; short n; /* A production is useful if all of the nonterminals in its appear @@ -118,7 +116,7 @@ useless_nonterminals (void) if (!bitset_test (P, i) && useful_production (i, N)) { - bitset_set (Np, rules[i].lhs - ntokens); + bitset_set (Np, rules[i].lhs->number - ntokens); bitset_set (P, i); } if (bitset_equal_p (N, Np)) @@ -138,7 +136,7 @@ inaccessable_symbols (void) bitset Vp, Vs, Pp; int i; short t; - rule r; + item_number_t *r; /* Find out which productions are reachable and which symbols are used. Starting with an empty set of productions and a set of @@ -167,9 +165,9 @@ inaccessable_symbols (void) Pp = bitset_create (nrules + 1, BITSET_FIXED); /* If the start symbol isn't useful, then nothing will be useful. */ - if (bitset_test (N, start_symbol - ntokens)) + if (bitset_test (N, axiom->number - ntokens)) { - bitset_set (V, start_symbol); + bitset_set (V, axiom->number); while (1) { @@ -178,7 +176,7 @@ inaccessable_symbols (void) { if (!bitset_test (Pp, i) && bitset_test (P, i) - && bitset_test (V, rules[i].lhs)) + && bitset_test (V, rules[i].lhs->number)) { for (r = rules[i].rhs; *r >= 0; r++) if (ISTOKEN (t = *r) || bitset_test (N, t - ntokens)) @@ -198,9 +196,9 @@ inaccessable_symbols (void) V = Vp; /* Tokens 0, 1, and 2 are internal to Bison. Consider them useful. */ - bitset_set (V, 0); /* end-of-input token */ - bitset_set (V, 1); /* error token */ - bitset_set (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 */ bitset_free (P); P = Pp; @@ -217,73 +215,67 @@ inaccessable_symbols (void) /* A token that was used in %prec should not be warned about. */ for (i = 1; i < nrules + 1; i++) if (rules[i].precsym != 0) - bitset_set (V1, rules[i].precsym); + 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. - - 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) + /* Report and flag useless productions. */ + { + int r; + for (r = 1; r < nrules + 1; r++) { - short np, pn, ni, pi; - - np = 0; - ni = 0; - for (pn = 1; pn < nrules + 1; pn++) - if (bitset_test (P, pn)) - { - np++; - if (pn != np) - { - 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 - ritem != ni) - { - pi = rules[np].rhs - ritem; - rules[np].rhs = ritem + ni; - while (ritem[pi] >= 0) - ritem[ni++] = ritem[pi++]; - ritem[ni++] = -np; - } - } - else - { - while (ritem[ni++] >= 0) - /* Nothing. */; - } - } + rules[r].useful = bitset_test (P, r); + if (!rules[r].useful) + { + LOCATION_PRINT (stderr, rules[r].location); + fprintf (stderr, ": %s: %s: ", _("warning"), _("useless rule")); + rule_print (&rules[r], stderr); + } + } + } - ritem[ni] = 0; - nrules -= nuseless_productions; - nitems = ni; - nritems = ni; + /* 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; - /* Is it worth it to reduce the amount of memory for the - grammar? Probably not. */ + /* Renumber the rules markers in RITEMS. */ + for (i = 1; i < nrules + 1; ++i) + { + item_number_t *rhsp = rules[i].rhs; + for (/* Nothing. */; *rhsp >= 0; ++rhsp) + /* Nothing. */; + *rhsp = -i; + rules[i].number = i; } + nrules -= nuseless_productions; + } - /* Disable useless productions. */ - if (nuseless_productions > 0) - { - int pn; - for (pn = 1; pn < nrules + 1; pn++) - rules[pn].useful = bitset_test (P, pn); - } + /* Adjust NRITEMS. */ + { + int r; + int length; + for (r = nrules + 1; r < nrules + 1 + nuseless_productions; ++r) + { + length = rule_rhs_length (&rules[r]); + nritems -= length + 1; + } + } } @@ -294,25 +286,33 @@ reduce_grammar_tables (void) static void nonterminals_reduce (void) { - int i, n; + symbol_number_t i, n; /* Map the nonterminals to their new index: useful first, useless afterwards. Kept for later report. */ - short *nontermmap = XCALLOC (short, nvars) - ntokens; + symbol_number_t *nontermmap = XCALLOC (symbol_number_t, 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++; + { + nontermmap[i] = n++; + LOCATION_PRINT (stderr, symbols[i]->location); + fprintf (stderr, ": %s: %s: %s\n", + _("warning"), _("useless nonterminal"), + symbol_tag_get (symbols[i])); + } /* Shuffle elements of tables indexed by symbol number. */ { - bucket **symbols_sorted = XMALLOC (bucket *, nvars) - ntokens; + symbol_t **symbols_sorted = XMALLOC (symbol_t *, nvars) - ntokens; + 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++) @@ -320,21 +320,17 @@ nonterminals_reduce (void) free (symbols_sorted + ntokens); } - /* Replace all symbol numbers in valid data structures. */ - - for (i = 1; i < nrules + 1; i++) - { - rules[i].lhs = nontermmap[rules[i].lhs]; - if (ISVAR (rules[i].precsym)) - /* Can this happen? */ - rules[i].precsym = nontermmap[rules[i].precsym]; - } - - for (i = 0; i < nritems; ++i) - if (ISVAR (ritem[i])) - ritem[i] = nontermmap[ritem[i]]; - - start_symbol = nontermmap[start_symbol]; + { + int r; + for (r = 1; r < nrules + 1; ++r) + { + item_number_t *rhsp; + for (rhsp = rules[r].rhs; *rhsp >= 0; ++rhsp) + if (ISVAR (*rhsp)) + *rhsp = symbol_number_as_item_number (nontermmap[*rhsp]); + } + axiom->number = nontermmap[axiom->number]; + } nsyms -= nuseless_nonterminals; nvars -= nuseless_nonterminals; @@ -355,7 +351,7 @@ reduce_output (FILE *out) int i; fprintf (out, "%s\n\n", _("Useless nonterminals:")); for (i = 0; i < nuseless_nonterminals; ++i) - fprintf (out, " %s\n", symbols[nsyms + i]->tag); + fprintf (out, " %s\n", symbol_tag_get (symbols[nsyms + i])); fputs ("\n\n", out); } @@ -368,76 +364,19 @@ reduce_output (FILE *out) if (!b) fprintf (out, "%s\n\n", _("Terminals which are not used:")); b = TRUE; - fprintf (out, " %s\n", symbols[i]->tag); + fprintf (out, " %s\n", symbol_tag_get (symbols[i])); } if (b) fputs ("\n\n", out); } if (nuseless_productions > 0) - { - int i; - fprintf (out, "%s\n\n", _("Useless rules:")); - for (i = 1; i < nrules + 1; i++) - if (!rules[i].useful) - { - rule r; - fprintf (out, "#%-4d ", i - 1); - fprintf (out, "%s:", symbols[rules[i].lhs]->tag); - for (r = rules[i].rhs; *r >= 0; r++) - fprintf (out, " %s", symbols[*r]->tag); - fputs (";\n", out); - } - fputs ("\n\n", out); - } + grammar_rules_partial_print (out, _("Useless rules"), + nrules + 1, + nuseless_productions + nrules + 1); } -static void -dump_grammar (FILE *out) -{ - int i; - rule r; - - 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"); - for (i = ntokens; i < nsyms; 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, "Num (Prec, Assoc, Useful, Ritem Range) Lhs -> Rhs (Ritem range) [Num]\n"); - for (i = 1; i < nrules + 1; i++) - { - 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, rules[i].assoc, rules[i].useful, - rules[i].rhs - ritem, rules[i].rhs - ritem + rhs_count - 1, - rules[i].lhs); - /* Dumped the RHS. */ - for (r = 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"); - for (i = 1; i < nrules + 1; i++) - { - fprintf (out, "%-5d %s :", i, symbols[rules[i].lhs]->tag); - for (r = rules[i].rhs; *r >= 0; r++) - fprintf (out, " %s", symbols[*r]->tag); - fputc ('\n', out); - } - fprintf (out, "\n\n"); -} + @@ -454,7 +393,7 @@ reduce_print (void) nuseless_productions), nuseless_productions); - fprintf (stderr, _("%s contains "), infile); + fprintf (stderr, "%s: %s: ", infile, _("warning")); if (nuseless_nonterminals > 0) fprintf (stderr, ngettext ("%d useless nonterminal", @@ -495,17 +434,21 @@ reduce_grammar (void) reduce_print (); - if (!bitset_test (N, start_symbol - ntokens)) + if (!bitset_test (N, axiom->number - ntokens)) fatal (_("Start symbol %s does not derive any sentence"), - symbols[start_symbol]->tag); + symbol_tag_get (symbols[axiom->number])); - reduce_grammar_tables (); + /* 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 (); if (trace_flag) { - dump_grammar (stderr); + grammar_dump (stderr, "Reduced Grammar"); fprintf (stderr, "reduced %s defines %d terminals, %d nonterminals\ , and %d productions.\n",