X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/23cbcc6c19c08edbede56f177fdde9d6f74d4e4f..da4160c30ae623f732447afa85ea08984abc2315:/src/LR0.c diff --git a/src/LR0.c b/src/LR0.c index f8fa5bd0..efe9589a 100644 --- a/src/LR0.c +++ b/src/LR0.c @@ -1,5 +1,5 @@ /* Generate the nondeterministic finite state machine for bison, - Copyright 1984, 1986, 1989, 2000, 2001 Free Software Foundation, Inc. + Copyright 1984, 1986, 1989, 2000, 2001, 2002 Free Software Foundation, Inc. This file is part of Bison, the GNU Compiler Compiler. @@ -23,6 +23,10 @@ The entry point is generate_states. */ #include "system.h" +#include "bitset.h" +#include "quotearg.h" +#include "symtab.h" +#include "gram.h" #include "getargs.h" #include "reader.h" #include "gram.h" @@ -30,39 +34,44 @@ #include "complain.h" #include "closure.h" #include "LR0.h" +#include "lalr.h" #include "reduce.h" -int nstates; -int final_state; -core *first_state = NULL; -shifts *first_shift = NULL; -reductions *first_reduction = NULL; +unsigned int nstates = 0; +/* Initialize the final state to -1, otherwise, it might be set to 0 + by default, and since we don't compute the reductions of the final + state, we end up not computing the reductions of the initial state, + which is of course needed. + + FINAL_STATE is properly set by new_state when it recognizes the + accessing symbol: EOF. */ +int final_state = -1; +static state_t *first_state = NULL; -static core *this_state = NULL; -static core *last_state = NULL; -static shifts *last_shift = NULL; -static reductions *last_reduction = NULL; +static state_t *this_state = NULL; +static state_t *last_state = NULL; static int nshifts; -static short *shift_symbol = NULL; +static symbol_number_t *shift_symbol = NULL; static short *redset = NULL; static short *shiftset = NULL; -static short **kernel_base = NULL; +static item_number_t **kernel_base = NULL; static int *kernel_size = NULL; -static short *kernel_items = NULL; +static item_number_t *kernel_items = NULL; /* hash table for states, to recognize equivalent ones. */ -#define STATE_TABLE_SIZE 1009 -static core **state_table = NULL; +#define STATE_HASH_SIZE 1009 +static state_t **state_hash = NULL; static void allocate_itemsets (void) { - int i; + int i, r; + item_number_t *rhsp; /* Count the number of occurrences of all the symbols in RITEMS. Note that useless productions (hence useless nonterminals) are @@ -71,11 +80,11 @@ allocate_itemsets (void) int count = 0; short *symbol_count = XCALLOC (short, nsyms + nuseless_nonterminals); - for (i = 0; ritem[i]; ++i) - if (ritem[i] > 0) + for (r = 1; r < nrules + 1; ++r) + for (rhsp = rules[r].rhs; *rhsp >= 0; ++rhsp) { count++; - symbol_count[ritem[i]]++; + symbol_count[*rhsp]++; } /* See comments before new_itemsets. All the vectors of items @@ -84,9 +93,9 @@ allocate_itemsets (void) appears as an item, which is symbol_count[symbol]. We allocate that much space for each symbol. */ - kernel_base = XCALLOC (short *, nsyms); + kernel_base = XCALLOC (item_number_t *, nsyms); if (count) - kernel_items = XCALLOC (short, count); + kernel_items = XCALLOC (item_number_t, count); count = 0; for (i = 0; i < nsyms; i++) @@ -107,7 +116,8 @@ allocate_storage (void) shiftset = XCALLOC (short, nsyms); redset = XCALLOC (short, nrules + 1); - state_table = XCALLOC (core *, STATE_TABLE_SIZE); + state_hash = XCALLOC (state_t *, STATE_HASH_SIZE); + shift_symbol = XCALLOC (symbol_number_t, nsyms); } @@ -120,7 +130,7 @@ free_storage (void) free (kernel_base); free (kernel_size); XFREE (kernel_items); - free (state_table); + free (state_hash); } @@ -141,7 +151,6 @@ static void new_itemsets (void) { int i; - int shiftcount; if (trace_flag) fprintf (stderr, "Entering new_itemsets, state = %d\n", @@ -150,26 +159,22 @@ new_itemsets (void) for (i = 0; i < nsyms; i++) kernel_size[i] = 0; - shift_symbol = XCALLOC (short, nsyms); - shiftcount = 0; - - for (i = 0; i < itemsetsize; ++i) - { - int symbol = ritem[itemset[i]]; - if (symbol > 0) - { - if (!kernel_size[symbol]) - { - shift_symbol[shiftcount] = symbol; - shiftcount++; - } - - kernel_base[symbol][kernel_size[symbol]] = itemset[i] + 1; - kernel_size[symbol]++; - } - } + nshifts = 0; - nshifts = shiftcount; + for (i = 0; i < nritemset; ++i) + if (ritem[itemset[i]] >= 0) + { + symbol_number_t symbol + = item_number_as_symbol_number (ritem[itemset[i]]); + if (!kernel_size[symbol]) + { + shift_symbol[nshifts] = symbol; + nshifts++; + } + + kernel_base[symbol][kernel_size[symbol]] = itemset[i] + 1; + kernel_size[symbol]++; + } } @@ -179,27 +184,38 @@ new_itemsets (void) | necessary. | `-----------------------------------------------------------------*/ -static core * -new_state (int symbol) +static state_t * +new_state (symbol_number_t symbol, size_t core_size, item_number_t *core) { - core *p; + state_t *p; if (trace_flag) fprintf (stderr, "Entering new_state, state = %d, symbol = %d (%s)\n", - this_state->number, symbol, tags[symbol]); + nstates, symbol, quotearg_style (escape_quoting_style, + symbols[symbol]->tag)); - if (nstates >= MAXSHORT) - fatal (_("too many states (max %d)"), MAXSHORT); + if (nstates >= SHRT_MAX) + fatal (_("too many states (max %d)"), SHRT_MAX); - p = CORE_ALLOC (kernel_size[symbol]); + p = STATE_ALLOC (core_size); p->accessing_symbol = symbol; p->number = nstates; - p->nitems = kernel_size[symbol]; + p->solved_conflicts = NULL; + + p->nitems = core_size; + memcpy (p->items, core, core_size * sizeof (core[0])); - shortcpy (p->items, kernel_base[symbol], kernel_size[symbol]); + /* If this is the eoftoken, and this is not the initial state, then + this is the final state. */ + if (symbol == 0 && first_state) + final_state = p->number; - last_state->next = p; + if (!first_state) + first_state = p; + if (last_state) + last_state->next = p; last_state = p; + nstates++; return p; @@ -213,34 +229,35 @@ new_state (int symbol) `--------------------------------------------------------------*/ static int -get_state (int symbol) +get_state (symbol_number_t symbol, size_t core_size, item_number_t *core) { int key; - int i; - core *sp; + size_t i; + state_t *sp; if (trace_flag) fprintf (stderr, "Entering get_state, state = %d, symbol = %d (%s)\n", - this_state->number, symbol, tags[symbol]); + this_state->number, symbol, quotearg_style (escape_quoting_style, + symbols[symbol]->tag)); /* Add up the target state's active item numbers to get a hash key. */ key = 0; - for (i = 0; i < kernel_size[symbol]; ++i) - key += kernel_base[symbol][i]; - key = key % STATE_TABLE_SIZE; - sp = state_table[key]; + for (i = 0; i < core_size; ++i) + key += core[i]; + key = key % STATE_HASH_SIZE; + sp = state_hash[key]; if (sp) { int found = 0; while (!found) { - if (sp->nitems == kernel_size[symbol]) + if (sp->nitems == core_size) { found = 1; - for (i = 0; i < kernel_size[symbol]; ++i) - if (kernel_base[symbol][i] != sp->items[i]) + for (i = 0; i < core_size; ++i) + if (core[i] != sp->items[i]) found = 0; } @@ -252,7 +269,7 @@ get_state (int symbol) } else /* bucket exhausted and no match */ { - sp = sp->link = new_state (symbol); + sp = sp->link = new_state (symbol, core_size, core); found = 1; } } @@ -260,7 +277,7 @@ get_state (int symbol) } else /* bucket is empty */ { - state_table[key] = sp = new_state (symbol); + state_hash[key] = sp = new_state (symbol, core_size, core); } if (trace_flag) @@ -281,7 +298,7 @@ append_states (void) { int i; int j; - int symbol; + symbol_number_t symbol; if (trace_flag) fprintf (stderr, "Entering append_states, state = %d\n", @@ -302,15 +319,21 @@ append_states (void) } for (i = 0; i < nshifts; i++) - shiftset[i] = get_state (shift_symbol[i]); + { + symbol = shift_symbol[i]; + shiftset[i] = get_state (symbol, + kernel_size[symbol], kernel_base[symbol]); + } } static void new_states (void) { - first_state = last_state = this_state = CORE_ALLOC (0); - nstates = 1; + /* The 0 at the lhs is the index of the item of this initial rule. */ + kernel_base[0][0] = 0; + kernel_size[0] = 1; + this_state = new_state (0, kernel_size[0], kernel_base[0]); } @@ -322,208 +345,8 @@ static void save_shifts (void) { shifts *p = shifts_new (nshifts); - - p->number = this_state->number; - - shortcpy (p->shifts, shiftset, nshifts); - - if (last_shift) - last_shift->next = p; - else - first_shift = p; - last_shift = p; -} - - -/*------------------------------------------------------------------. -| Subroutine of augment_automaton. Create the next-to-final state, | -| to which a shift has already been made in the initial state. | -`------------------------------------------------------------------*/ - -static void -insert_start_shift (void) -{ - core *statep; - shifts *sp; - - statep = CORE_ALLOC (0); - statep->number = nstates; - statep->accessing_symbol = start_symbol; - - last_state->next = statep; - last_state = statep; - - /* Make a shift from this state to (what will be) the final state. */ - sp = shifts_new (1); - sp->number = nstates++; - sp->shifts[0] = nstates; - - last_shift->next = sp; - last_shift = sp; -} - - -/*------------------------------------------------------------------. -| Make sure that the initial state has a shift that accepts the | -| grammar's start symbol and goes to the next-to-final state, which | -| has a shift going to the final state, which has a shift to the | -| termination state. Create such states and shifts if they don't | -| happen to exist already. | -`------------------------------------------------------------------*/ - -static void -augment_automaton (void) -{ - core *statep; - shifts *sp; - shifts *sp1 = NULL; - - sp = first_shift; - - if (!sp->nshifts) - { - /* There are no shifts for any state. Make one shift, from the - initial state to the next-to-final state. */ - - sp = shifts_new (1); - sp->shifts[0] = nstates; - - /* Initialize the chain of shifts with sp. */ - first_shift = sp; - last_shift = sp; - - /* Create the next-to-final state, with shift to - what will be the final state. */ - insert_start_shift (); - } - else if (sp->number == 0) - { - statep = first_state->next; - - /* The states reached by shifts from FIRST_STATE are numbered - 1..(SP->NSHIFTS). Look for one reached by START_SYMBOL. */ - while (statep->accessing_symbol < start_symbol - && statep->number < sp->nshifts) - statep = statep->next; - - if (statep->accessing_symbol == start_symbol) - { - /* We already have a next-to-final state. - Make sure it has a shift to what will be the final state. */ - while (sp && sp->number < statep->number) - { - sp1 = sp; - sp = sp->next; - } - - if (sp && sp->number == statep->number) - { - int i; - shifts *sp2 = shifts_new (sp->nshifts + 1); - sp2->number = statep->number; - sp2->shifts[0] = nstates; - for (i = sp->nshifts; i > 0; i--) - sp2->shifts[i] = sp->shifts[i - 1]; - - /* Patch sp2 into the chain of shifts in place of sp, - following sp1. */ - sp2->next = sp->next; - sp1->next = sp2; - if (sp == last_shift) - last_shift = sp2; - XFREE (sp); - } - else - { - shifts *sp2 = shifts_new (1); - sp2->number = statep->number; - sp2->shifts[0] = nstates; - - /* Patch sp2 into the chain of shifts between sp1 and sp. */ - sp2->next = sp; - sp1->next = sp2; - if (sp == 0) - last_shift = sp2; - } - } - else - { - int i, k; - shifts *sp2; - - /* There is no next-to-final state as yet. */ - /* Add one more shift in first_shift, - going to the next-to-final state (yet to be made). */ - sp = first_shift; - - sp2 = shifts_new (sp->nshifts + 1); - - /* Stick this shift into the vector at the proper place. */ - statep = first_state->next; - for (k = 0, i = 0; i < sp->nshifts; k++, i++) - { - if (statep->accessing_symbol > start_symbol && i == k) - sp2->shifts[k++] = nstates; - sp2->shifts[k] = sp->shifts[i]; - statep = statep->next; - } - if (i == k) - sp2->shifts[k++] = nstates; - - /* Patch sp2 into the chain of shifts - in place of sp, at the beginning. */ - sp2->next = sp->next; - first_shift = sp2; - if (last_shift == sp) - last_shift = sp2; - - XFREE (sp); - - /* Create the next-to-final state, with shift to - what will be the final state. */ - insert_start_shift (); - } - } - else - { - /* The initial state didn't even have any shifts. - Give it one shift, to the next-to-final state. */ - sp = shifts_new (1); - sp->shifts[0] = nstates; - - /* Patch sp into the chain of shifts at the beginning. */ - sp->next = first_shift; - first_shift = sp; - - /* Create the next-to-final state, with shift to - what will be the final state. */ - insert_start_shift (); - } - - /* Make the final state--the one that follows a shift from the - next-to-final state. - The symbol for that shift is 0 (end-of-file). */ - statep = CORE_ALLOC (0); - statep->number = nstates; - last_state->next = statep; - last_state = statep; - - /* Make the shift from the final state to the termination state. */ - sp = shifts_new (1); - sp->number = nstates++; - sp->shifts[0] = nstates; - last_shift->next = sp; - last_shift = sp; - - /* Note that the variable `final_state' refers to what we sometimes call - the termination state. */ - final_state = nstates; - - /* Make the termination state. */ - statep = CORE_ALLOC (0); - statep->number = nstates++; - last_state->next = statep; - last_state = statep; + memcpy (p->shifts, shiftset, nshifts * sizeof (shiftset[0])); + this_state->shifts = p; } @@ -536,13 +359,16 @@ augment_automaton (void) static void save_reductions (void) { - int count; + int count = 0; int i; - /* Find and count the active items that represent ends of rules. */ + /* If this is the final state, we want it to have no reductions at + all, although it has one for `START_SYMBOL EOF .'. */ + if (this_state->number == final_state) + return; - count = 0; - for (i = 0; i < itemsetsize; ++i) + /* Find and count the active items that represent ends of rules. */ + for (i = 0; i < nritemset; ++i) { int item = ritem[itemset[i]]; if (item < 0) @@ -550,25 +376,37 @@ save_reductions (void) } /* Make a reductions structure and copy the data into it. */ + this_state->reductions = reductions_new (count); + memcpy (this_state->reductions->rules, redset, count * sizeof (redset[0])); +} - if (count) - { - reductions *p = REDUCTIONS_ALLOC (count); - - p->number = this_state->number; - p->nreds = count; + +/*---------------. +| Build STATES. | +`---------------*/ - shortcpy (p->rules, redset, count); +static void +set_states (void) +{ + state_t *sp; + states = XCALLOC (state_t *, nstates); - if (last_reduction) - last_reduction->next = p; - else - first_reduction = p; - last_reduction = p; + for (sp = first_state; sp; sp = sp->next) + { + /* Pessimization, but simplification of the code: make sure all + the states have a shifts, errs, and reductions, even if + reduced to 0. */ + if (!sp->shifts) + sp->shifts = shifts_new (0); + if (!sp->errs) + sp->errs = errs_new (0); + if (!sp->reductions) + sp->reductions = reductions_new (0); + + states[sp->number] = sp; } } - /*-------------------------------------------------------------------. | Compute the nondeterministic finite state machine (see state.h for | | details) from the grammar. | @@ -578,14 +416,16 @@ void generate_states (void) { allocate_storage (); - new_closure (nitems); + new_closure (nritems); new_states (); while (this_state) { if (trace_flag) fprintf (stderr, "Processing state %d (reached by %s)\n", - this_state->number, tags[this_state->accessing_symbol]); + this_state->number, + quotearg_style (escape_quoting_style, + symbols[this_state->accessing_symbol]->tag)); /* Set up ruleset and itemset for the transitions out of this state. ruleset gets a 1 bit for each rule that could reduce now. itemset gets a vector of all the items that could be @@ -610,6 +450,6 @@ generate_states (void) free_closure (); free_storage (); - /* set up initial and final states as parser wants them */ - augment_automaton (); + /* Set up STATES. */ + set_states (); }