src/state.h

   1 /* Type definitions for nondeterministic finite state machine for bison,
   2    Copyright 1984, 1989, 2000, 2001 Free Software Foundation, Inc.
   3
   4    This file is part of Bison, the GNU Compiler Compiler.
   5
   6    Bison is free software; you can redistribute it and/or modify
   7    it under the terms of the GNU General Public License as published by
   8    the Free Software Foundation; either version 2, or (at your option)
   9    any later version.
  10
  11    Bison is distributed in the hope that it will be useful,
  12    but WITHOUT ANY WARRANTY; without even the implied warranty of
  13    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  14    GNU General Public License for more details.
  15
  16    You should have received a copy of the GNU General Public License
  17    along with Bison; see the file COPYING.  If not, write to
  18    the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
  19    Boston, MA 02111-1307, USA.  */
  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 RITEMS 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_t 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.  The data is a list of
  65    nreds rules, represented by their rule numbers.  first_reduction
  66    points to the list of these structures.
  67
  68    Conflict resolution can decide that certain tokens in certain
  69    states should explicitly be errors (for implementing %nonassoc).
  70    For each state, the tokens that are errors for this reason are
  71    recorded in an errs structure, which holds the token numbers.
  72
  73    There is at least one goto transition present in state zero.  It
  74    leads to a next-to-final state whose accessing_symbol is the
  75    grammar's start symbol.  The next-to-final state has one shift to
  76    the final state, whose accessing_symbol is zero (end of input).
  77    The final state has one shift, which goes to the termination state.
  78    The reason for the extra state at the end is to placate the
  79    parser's strategy of making all decisions one token ahead of its
  80    actions.  */
  81
  82 #ifndef STATE_H_
  83 # define STATE_H_
  84
  85 # include "bitset.h"
  86
  87
  88 /*-------------------.
  89 | Numbering states.  |
  90 `-------------------*/
  91
  92 typedef short state_number_t;
  93 # define STATE_NUMBER_MAX ((state_number_t) SHRT_MAX)
  94
  95 /* Be ready to map a state_number_t to an int.  */
  96 # define state_number_as_int(Tok) ((int) (Tok))
  97
  98
  99 typedef struct state_s state_t;
 100
 101 /*--------------.
 102 | Transitions.  |
 103 `--------------*/
 104
 105 typedef struct transtion_s
 106 {
 107   short num;
 108   state_t *states[1];
 109 } transitions_t;
 110
 111
 112 /* What is the symbol labelling the transition to
 113    TRANSITIONS->states[Num]?  Can be a token (amongst which the error
 114    token), or non terminals in case of gotos.  */
 115
 116 #define TRANSITION_SYMBOL(Transitions, Num) \
 117   (Transitions->states[Num]->accessing_symbol)
 118
 119 /* Is the TRANSITIONS->states[Num] a shift? (as opposed to gotos).  */
 120
 121 #define TRANSITION_IS_SHIFT(Transitions, Num) \
 122   (ISTOKEN (TRANSITION_SYMBOL (Transitions, Num)))
 123
 124 /* Is the TRANSITIONS->states[Num] a goto?. */
 125
 126 #define TRANSITION_IS_GOTO(Transitions, Num) \
 127   (!TRANSITION_IS_SHIFT (Transitions, Num))
 128
 129 /* Is the TRANSITIONS->states[Num] labelled by the error token?  */
 130
 131 #define TRANSITION_IS_ERROR(Transitions, Num) \
 132   (TRANSITION_SYMBOL (Transitions, Num) == errtoken->number)
 133
 134 /* When resolving a SR conflicts, if the reduction wins, the shift is
 135    disabled.  */
 136
 137 #define TRANSITION_DISABLE(Transitions, Num) \
 138   (Transitions->states[Num] = NULL)
 139
 140 #define TRANSITION_IS_DISABLED(Transitions, Num) \
 141   (Transitions->states[Num] == NULL)
 142
 143
 144 /* Iterate over each transition over a token (shifts).  */
 145 #define FOR_EACH_SHIFT(Transitions, Iter)                       \
 146   for (Iter = 0;                                                \
 147        Iter < Transitions->num                                  \
 148          && (TRANSITION_IS_DISABLED (Transitions, Iter)         \
 149              || TRANSITION_IS_SHIFT (Transitions, Iter));       \
 150        ++Iter)                                                  \
 151     if (!TRANSITION_IS_DISABLED (Transitions, Iter))
 152
 153
 154 /* Return the state such these TRANSITIONS contain a shift/goto to it on
 155    SYMBOL.  Aborts if none found.  */
 156 struct state_s;
 157 struct state_s *transitions_to PARAMS ((transitions_t *state,
 158                                         symbol_number_t s));
 159
 160
 161 /*-------.
 162 | Errs.  |
 163 `-------*/
 164
 165 typedef struct errs_s
 166 {
 167   short num;
 168   symbol_t *symbols[1];
 169 } errs_t;
 170
 171 errs_t *errs_new PARAMS ((int num, symbol_t **tokens));
 172
 173
 174 /*-------------.
 175 | Reductions.  |
 176 `-------------*/
 177
 178 typedef struct reductions_s
 179 {
 180   short num;
 181   bitset *lookaheads;
 182   rule_t *rules[1];
 183 } reductions_t;
 184
 185
 186
 187 /*---------.
 188 | States.  |
 189 `---------*/
 190
 191 struct state_s
 192 {
 193   state_number_t number;
 194   symbol_number_t accessing_symbol;
 195   transitions_t     *transitions;
 196   reductions_t *reductions;
 197   errs_t       *errs;
 198
 199   /* Nonzero if no lookahead is needed to decide what to do in state S.  */
 200   char consistent;
 201
 202   /* If some conflicts were solved thanks to precedence/associativity,
 203      a human readable description of the resolution.  */
 204   const char *solved_conflicts;
 205
 206   /* Its items.  Must be last, since ITEMS can be arbitrarily large.
 207      */
 208   unsigned short nitems;
 209   item_number_t items[1];
 210 };
 211
 212 extern state_number_t nstates;
 213 extern state_t *final_state;
 214
 215 /* Create a new state with ACCESSING_SYMBOL for those items.  */
 216 state_t *state_new PARAMS ((symbol_number_t accessing_symbol,
 217                             size_t core_size, item_number_t *core));
 218
 219 /* Set the transitions of STATE.  */
 220 void state_transitions_set PARAMS ((state_t *state,
 221                                     int num, state_t **transitions));
 222
 223 /* Set the reductions of STATE.  */
 224 void state_reductions_set PARAMS ((state_t *state,
 225                                    int num, rule_t **reductions));
 226
 227 int state_reduction_find PARAMS ((state_t *state, rule_t *rule));
 228
 229 /* Set the errs of STATE.  */
 230 void state_errs_set PARAMS ((state_t *state,
 231                              int num, symbol_t **errs));
 232
 233 /* Print on OUT all the lookaheads such that this STATE wants to
 234    reduce this RULE.  */
 235 void state_rule_lookaheads_print PARAMS ((state_t *state, rule_t *rule,
 236                                           FILE *out));
 237
 238 /* Create/destroy the states hash table.  */
 239 void state_hash_new PARAMS ((void));
 240 void state_hash_free PARAMS ((void));
 241
 242 /* Find the state associated to the CORE, and return it.  If it does
 243    not exist yet, return NULL.  */
 244 state_t *state_hash_lookup PARAMS ((size_t core_size, item_number_t *core));
 245
 246 /* Insert STATE in the state hash table.  */
 247 void state_hash_insert PARAMS ((state_t *state));
 248
 249 /* All the states, indexed by the state number.  */
 250 extern state_t **states;
 251
 252 /* Free all the states.  */
 253 void states_free PARAMS ((void));
 254 #endif /* !STATE_H_ */