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 link field is used for chaining symbols that hash states by
  48    their itemsets.  This is for recognizing equivalent states and
  49    combining them when the states are generated.
  50
  51    The two types of transitions are shifts (push the lookahead token
  52    and read another) and reductions (combine the last n things on the
  53    stack via a rule, replace them with the symbol that the rule
  54    derives, and leave the lookahead token alone).  When the states are
  55    generated, these transitions are represented in two other lists.
  56
  57    Each shifts structure describes the possible shift transitions out
  58    of one state, the state whose number is in the number field.  The
  59    shifts structures are linked through next and first_shift points to
  60    them.  Each contains a vector of numbers of the states that shift
  61    transitions can go to.  The accessing_symbol fields of those
  62    states' cores say what kind of input leads to them.
  63
  64    A shift to state zero should be ignored.  Conflict resolution
  65    deletes shifts by changing them to zero.
  66
  67    Each reductions structure describes the possible reductions at the
  68    state whose number is in the number field.  The data is a list of
  69    nreds rules, represented by their rule numbers.  first_reduction
  70    points to the list of these structures.
  71
  72    Conflict resolution can decide that certain tokens in certain
  73    states should explicitly be errors (for implementing %nonassoc).
  74    For each state, the tokens that are errors for this reason are
  75    recorded in an errs structure, which has the state number in its
  76    number field.  The rest of the errs structure is full of token
  77    numbers.
  78
  79    There is at least one shift transition present in state zero.  It
  80    leads to a next-to-final state whose accessing_symbol is the
  81    grammar's start symbol.  The next-to-final state has one shift to
  82    the final state, whose accessing_symbol is zero (end of input).
  83    The final state has one shift, which goes to the termination state
  84    (whose number is nstates-1).  The reason for the extra state at the
  85    end is to placate the parser's strategy of making all decisions one
  86    token ahead of its actions.  */
  87
  88 #ifndef STATE_H_
  89 # define STATE_H_
  90
  91 # include "bitsetv.h"
  92
  93 /*---------.
  94 | Shifts.  |
  95 `---------*/
  96
  97 typedef struct shifts
  98 {
  99   short nshifts;
 100   short shifts[1];
 101 } shifts;
 102
 103 shifts *shifts_new PARAMS ((int n));
 104
 105
 106 /* What is the symbol which is shifted by SHIFTS->shifts[Shift]?  Can
 107    be a token (amongst which the error token), or non terminals in
 108    case of gotos.  */
 109
 110 #define SHIFT_SYMBOL(Shifts, Shift) \
 111   (states[Shifts->shifts[Shift]]->accessing_symbol)
 112
 113 /* Is the SHIFTS->shifts[Shift] a real shift? (as opposed to gotos.) */
 114
 115 #define SHIFT_IS_SHIFT(Shifts, Shift) \
 116   (ISTOKEN (SHIFT_SYMBOL (Shifts, Shift)))
 117
 118 /* Is the SHIFTS->shifts[Shift] a goto?. */
 119
 120 #define SHIFT_IS_GOTO(Shifts, Shift) \
 121   (!SHIFT_IS_SHIFT (Shifts, Shift))
 122
 123 /* Is the SHIFTS->shifts[Shift] then handling of the error token?. */
 124
 125 #define SHIFT_IS_ERROR(Shifts, Shift) \
 126   (SHIFT_SYMBOL (Shifts, Shift) == errtoken->number)
 127
 128 /* When resolving a SR conflicts, if the reduction wins, the shift is
 129    disabled.  */
 130
 131 #define SHIFT_DISABLE(Shifts, Shift) \
 132   (Shifts->shifts[Shift] = 0)
 133
 134 #define SHIFT_IS_DISABLED(Shifts, Shift) \
 135   (Shifts->shifts[Shift] == 0)
 136
 137
 138 /*-------.
 139 | Errs.  |
 140 `-------*/
 141
 142 typedef struct errs
 143 {
 144   short nerrs;
 145   short errs[1];
 146 } errs;
 147
 148 errs *errs_new PARAMS ((int n));
 149 errs *errs_dup PARAMS ((errs *src));
 150
 151
 152 /*-------------.
 153 | Reductions.  |
 154 `-------------*/
 155
 156 typedef struct reductions
 157 {
 158   short nreds;
 159   short rules[1];
 160 } reductions;
 161
 162 reductions *reductions_new PARAMS ((int n));
 163
 164
 165 /*----------.
 166 | State_t.  |
 167 `----------*/
 168
 169 typedef struct state_s
 170 {
 171   struct state_s *next;
 172   struct state_s *link;
 173
 174   short number;
 175   symbol_number_t accessing_symbol;
 176   shifts     *shifts;
 177   reductions *reductions;
 178   errs       *errs;
 179
 180   /* Nonzero if no lookahead is needed to decide what to do in state S.  */
 181   char consistent;
 182
 183   /* Used in LALR, not LR(0). */
 184   int nlookaheads;
 185   bitsetv lookaheads;
 186   rule_t **lookaheads_rule;
 187
 188   /* If some conflicts were solved thanks to precedence/associativity,
 189      a human readable description of the resolution.  */
 190   const char *solved_conflicts;
 191
 192   /* Its items.  Must be last, since ITEMS can be arbitrarily large.
 193      */
 194   unsigned short nitems;
 195   item_number_t items[1];
 196 } state_t;
 197
 198 #define STATE_ALLOC(Nitems)                                             \
 199   (state_t *) xcalloc ((unsigned) (sizeof (state_t)                     \
 200                                   + (Nitems - 1) * sizeof (item_number_t)), 1)
 201
 202 /* Print on OUT all the lookaheads such that this STATE wants to
 203    reduce this RULE.  */
 204
 205 void state_rule_lookaheads_print PARAMS ((state_t *state, rule_t *rule,
 206                                           FILE *out));
 207
 208 #endif /* !STATE_H_ */