1/* Type definitions for the finite state machine for Bison. 2 3 Copyright (C) 1984, 1989, 2000-2004, 2007, 2009-2012 Free Software 4 Foundation, Inc. 5 6 This file is part of Bison, the GNU Compiler Compiler. 7 8 This program is free software: you can redistribute it and/or modify 9 it under the terms of the GNU General Public License as published by 10 the Free Software Foundation, either version 3 of the License, or 11 (at your option) any later version. 12 13 This program is distributed in the hope that it will be useful, 14 but WITHOUT ANY WARRANTY; without even the implied warranty of 15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 16 GNU General Public License for more details. 17 18 You should have received a copy of the GNU General Public License 19 along with this program. If not, see <http://www.gnu.org/licenses/>. */ 20 21 22/* These type definitions are used to represent a nondeterministic 23 finite state machine that parses the specified grammar. This 24 information is generated by the function generate_states in the 25 file LR0. 26 27 Each state of the machine is described by a set of items -- 28 particular positions in particular rules -- that are the possible 29 places where parsing could continue when the machine is in this 30 state. These symbols at these items are the allowable inputs that 31 can follow now. 32 33 A core represents one state. States are numbered in the NUMBER 34 field. When generate_states is finished, the starting state is 35 state 0 and NSTATES is the number of states. (FIXME: This sentence 36 is no longer true: A transition to a state whose state number is 37 NSTATES indicates termination.) All the cores are chained together 38 and FIRST_STATE points to the first one (state 0). 39 40 For each state there is a particular symbol which must have been 41 the last thing accepted to reach that state. It is the 42 ACCESSING_SYMBOL of the core. 43 44 Each core contains a vector of NITEMS items which are the indices 45 in the RITEM vector of the items that are selected in this state. 46 47 The two types of actions are shifts/gotos (push the lookahead token 48 and read another/goto to the state designated by a nterm) and 49 reductions (combine the last n things on the stack via a rule, 50 replace them with the symbol that the rule derives, and leave the 51 lookahead token alone). When the states are generated, these 52 actions are represented in two other lists. 53 54 Each transition structure describes the possible transitions out 55 of one state, the state whose number is in the number field. Each 56 contains a vector of numbers of the states that transitions can go 57 to. The accessing_symbol fields of those states' cores say what 58 kind of input leads to them. 59 60 A transition to state zero should be ignored: conflict resolution 61 deletes transitions by having them point to zero. 62 63 Each reductions structure describes the possible reductions at the 64 state whose number is in the number field. rules is an array of 65 num rules. lookahead_tokens is an array of bitsets, one per rule. 66 67 Conflict resolution can decide that certain tokens in certain 68 states should explicitly be errors (for implementing %nonassoc). 69 For each state, the tokens that are errors for this reason are 70 recorded in an errs structure, which holds the token numbers. 71 72 There is at least one goto transition present in state zero. It 73 leads to a next-to-final state whose accessing_symbol is the 74 grammar's start symbol. The next-to-final state has one shift to 75 the final state, whose accessing_symbol is zero (end of input). 76 The final state has one shift, which goes to the termination state. 77 The reason for the extra state at the end is to placate the 78 parser's strategy of making all decisions one token ahead of its 79 actions. */ 80 81#ifndefSTATE_H_ 82#defineSTATE_H_ 83 84#include<bitset.h> 85 86#include"gram.h" 87#include"symtab.h" 88 89 90/*-------------------. 91| Numbering states. | 92`-------------------*/ 93 94typedefintstate_number; 95#defineSTATE_NUMBER_MAXIMUM INT_MAX 96 97/* Be ready to map a state_number to an int. */ 98staticinlineint99state_number_as_int (state_number s) 100{ 101returns; 102} 103 104 105typedefstructstate state; 106 107/*--------------. 108| Transitions. | 109`--------------*/ 110 111typedefstruct112{ 113intnum; 114 state *states[1]; 115} transitions; 116 117 118/* What is the symbol labelling the transition to 119 TRANSITIONS->states[Num]? Can be a token (amongst which the error 120 token), or non terminals in case of gotos. */ 121 122#defineTRANSITION_SYMBOL(Transitions, Num) \ 123 (Transitions->states[Num]->accessing_symbol) 124 125/* Is the TRANSITIONS->states[Num] a shift? (as opposed to gotos). */ 126 127#defineTRANSITION_IS_SHIFT(Transitions, Num) \ 128 (ISTOKEN (TRANSITION_SYMBOL (Transitions, Num))) 129 130/* Is the TRANSITIONS->states[Num] a goto?. */ 131 132#defineTRANSITION_IS_GOTO(Transitions, Num) \ 133 (!TRANSITION_IS_SHIFT (Transitions, Num)) 134 135/* Is the TRANSITIONS->states[Num] labelled by the error token? */ 136 137#defineTRANSITION_IS_ERROR(Transitions, Num) \ 138 (TRANSITION_SYMBOL (Transitions, Num) == errtoken->number) 139 140/* When resolving a SR conflicts, if the reduction wins, the shift is 141 disabled. */ 142 143#defineTRANSITION_DISABLE(Transitions, Num) \ 144 (Transitions->states[Num] = NULL) 145 146#defineTRANSITION_IS_DISABLED(Transitions, Num) \ 147 (Transitions->states[Num] == NULL) 148 149 150/* Iterate over each transition over a token (shifts). */ 151#defineFOR_EACH_SHIFT(Transitions, Iter) \ 152for(Iter = 0; \ 153 Iter < Transitions->num \ 154 && (TRANSITION_IS_DISABLED (Transitions, Iter) \ 155 || TRANSITION_IS_SHIFT (Transitions, Iter)); \ 156 ++Iter) \ 157if(!TRANSITION_IS_DISABLED (Transitions, Iter)) 158 159 160/* Return the state such SHIFTS contain a shift/goto to it on SYM. 161 Abort if none found. */ 162structstate *transitions_to (transitions *shifts, symbol_number sym); 163 164 165/*-------. 166| Errs. | 167`-------*/ 168 169typedefstruct170{ 171intnum; 172 symbol *symbols[1]; 173} errs; 174 175errs *errs_new (intnum, symbol **tokens); 176 177 178/*-------------. 179| Reductions. | 180`-------------*/ 181 182typedefstruct183{ 184intnum; 185 bitset *lookahead_tokens; 186 /* Sorted ascendingly on rule number. */ 187 rule *rules[1]; 188} reductions; 189 190 191 192/*---------. 193| states. | 194`---------*/ 195 196structstate_list; 197 198structstate 199{ 200 state_number number; 201 symbol_number accessing_symbol; 202 transitions *transitions; 203 reductions *reductions; 204 errs *errs; 205 206 /* When an includer (such as ielr.c) needs to store states in a list, the 207 includer can define struct state_list as the list node structure and can 208 store in this member a reference to the node containing each state. */ 209structstate_list *state_list; 210 211 /* If non-zero, then no lookahead sets on reduce actions are needed to 212 decide what to do in state S. */ 213charconsistent; 214 215 /* If some conflicts were solved thanks to precedence/associativity, 216 a human readable description of the resolution. */ 217constchar*solved_conflicts; 218constchar*solved_conflicts_xml; 219 220 /* Its items. Must be last, since ITEMS can be arbitrarily large. Sorted 221 ascendingly on item index in RITEM, which is sorted on rule number. */ 222 size_t nitems; 223 item_number items[1]; 224}; 225 226externstate_number nstates; 227externstate *final_state; 228 229/* Create a new state with ACCESSING_SYMBOL for those items. */ 230state *state_new (symbol_number accessing_symbol, 231 size_t core_size, item_number *core); 232state *state_new_isocore (stateconst*s); 233 234/* Set the transitions of STATE. */ 235voidstate_transitions_set (state *s,intnum, state **trans); 236 237/* Set the reductions of STATE. */ 238voidstate_reductions_set (state *s,intnum, rule **reds); 239 240intstate_reduction_find (state *s, rule *r); 241 242/* Set the errs of STATE. */ 243voidstate_errs_set (state *s,intnum, symbol **errors); 244 245/* Print on OUT all the lookahead tokens such that this STATE wants to 246 reduce R. */ 247voidstate_rule_lookahead_tokens_print (state *s, rule *r, FILE *out); 248voidstate_rule_lookahead_tokens_print_xml (state *s, rule *r, 249 FILE *out,intlevel); 250 251/* Create/destroy the states hash table. */ 252voidstate_hash_new (void); 253voidstate_hash_free (void); 254 255/* Find the state associated to the CORE, and return it. If it does 256 not exist yet, return NULL. */ 257state *state_hash_lookup (size_t core_size, item_number *core); 258 259/* Insert STATE in the state hash table. */ 260voidstate_hash_insert (state *s); 261 262/* Remove unreachable states, renumber remaining states, update NSTATES, and 263 write to OLD_TO_NEW a mapping of old state numbers to new state numbers such 264 that the old value of NSTATES is written as the new state number for removed 265 states. The size of OLD_TO_NEW must be the old value of NSTATES. */ 266voidstate_remove_unreachable_states (state_number old_to_new[]); 267 268/* All the states, indexed by the state number. */ 269externstate **states; 270 271/* Free all the states. */ 272voidstates_free (void); 273 274#endif/* !STATE_H_ */ 275