src/gram.h

   1 /* Data definitions for internal representation of bison's input,
   2    Copyright (C) 1984, 1986, 1989, 1992, 2001, 2002
   3    Free Software Foundation, Inc.
   4
   5    This file is part of Bison, the GNU Compiler Compiler.
   6
   7    Bison is free software; you can redistribute it and/or modify
   8    it under the terms of the GNU General Public License as published by
   9    the Free Software Foundation; either version 2, or (at your option)
  10    any later version.
  11
  12    Bison is distributed in the hope that it will be useful,
  13    but WITHOUT ANY WARRANTY; without even the implied warranty of
  14    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  15    GNU General Public License for more details.
  16
  17    You should have received a copy of the GNU General Public License
  18    along with Bison; see the file COPYING.  If not, write to
  19    the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
  20    Boston, MA 02111-1307, USA.  */
  21
  22 #ifndef GRAM_H_
  23 # define GRAM_H_
  24
  25 /* Representation of the grammar rules:
  26
  27    NTOKENS is the number of tokens, and NVARS is the number of
  28    variables (nonterminals).  NSYMS is the total number, ntokens +
  29    nvars.
  30
  31    Each symbol (either token or variable) receives a symbol number.
  32    Numbers 0 to NTOKENS - 1 are for tokens, and NTOKENS to NSYMS - 1
  33    are for variables.  Symbol number zero is the end-of-input token.
  34    This token is counted in ntokens.  The true number of token values
  35    assigned is NTOKENS reduced by one for each alias declaration.
  36
  37    The rules receive rule numbers 1 to NRULES in the order they are
  38    written.  More precisely Bison augments the grammar with the
  39    initial rule, `$axiom: START-SYMBOL EOF', which is numbered 1, all
  40    the user rules are 2, 3 etc.  Each time a rule number is presented
  41    to the user, we subtract 1, so *displayed* rule numbers are 0, 1,
  42    2...
  43
  44    Internally, we cannot use the number 0 for a rule because for
  45    instance RITEM stores both symbol (the RHS) and rule numbers: the
  46    symbols are shorts >= 0, and rule number are stored negative.
  47    Therefore 0 cannot be used, since it would be both the rule number
  48    0, and the token EOF).
  49
  50    Actions are accessed via the rule number.
  51
  52    The rules themselves are described by several arrays: amongst which
  53    RITEM, and RULES.
  54
  55    RULES is an array of struct rule_s, which members are:
  56
  57    RULES[R].lhs -- the symbol of the left hand side of rule R.
  58
  59    RULES[R].rhs -- the index in RITEM of the beginning of the portion
  60    for rule R.
  61
  62    RULES[R].prec -- the symbol providing the precedence level of R.
  63
  64    RULES[R].precsym -- the symbol attached (via %prec) to give its
  65    precedence to R.  Of course, if set, it is equal to `prec', but we
  66    need to distinguish one from the other when reducing: a symbol used
  67    in a %prec is not useless.
  68
  69    RULES[R].assoc -- the associativity of R.
  70
  71    RULES[R].dprec -- the dynamic precedence level of R (for GLR parsing).
  72
  73    RULES[R].merger -- index of merging function for R (for GLR parsing).
  74
  75    RULES[R].line -- the line where R was defined.
  76
  77    RULES[R].useful -- TRUE iff the rule is used (i.e., FALSE if thrown
  78    away by reduce).
  79
  80    The right hand side is stored as symbol numbers in a portion of
  81    RITEM.
  82
  83    The length of the portion is one greater than the number of symbols
  84    in the rule's right hand side.  The last element in the portion
  85    contains minus R, which identifies it as the end of a portion and
  86    says which rule it is for.
  87
  88    The portions of RITEM come in order of increasing rule number.
  89    NRITEMS is the total length of RITEM.  Each element of RITEM is
  90    called an "item" and its index in RITEM is an item number.
  91
  92    Item numbers are used in the finite state machine to represent
  93    places that parsing can get to.
  94
  95    SYMBOLS[I]->prec records the precedence level of each symbol.
  96
  97    Precedence levels are assigned in increasing order starting with 1
  98    so that numerically higher precedence values mean tighter binding
  99    as they ought to.  Zero as a symbol or rule's precedence means none
 100    is assigned.
 101
 102    Associativities are recorded similarly in SYMBOLS[I]->assoc.  */
 103
 104 # include "location.h"
 105 # include "symtab.h"
 106
 107 # define ISTOKEN(s)     ((s) < ntokens)
 108 # define ISVAR(s)       ((s) >= ntokens)
 109
 110 extern int nrules;
 111 extern int nsyms;
 112 extern int ntokens;
 113 extern int nvars;
 114
 115 # define ITEM_NUMBER_MAX INT_MAX
 116 typedef int item_number_t;
 117 extern item_number_t *ritem;
 118 extern unsigned int nritems;
 119
 120 /* There is weird relationship between item_number_t and
 121    symbol_number_t: we store symbol_number_t in item_number_t, but in
 122    the latter we also store, as negative numbers, the rule numbers.
 123
 124    Therefore, an symbol_number_t must be a valid item_number_t, and we
 125    sometimes have to perform the converse transformation.  */
 126 # define symbol_number_as_item_number(Tok) ((item_number_t) (Tok))
 127 # define item_number_as_symbol_number(Ite) ((symbol_number_t) (Ite))
 128
 129 extern symbol_number_t start_symbol;
 130
 131
 132 typedef struct rule_s
 133 {
 134   /* The number of the rule in the source.  It is usually the index in
 135      RULES too, except if there are useless rules.  */
 136   short user_number;
 137
 138   /* The index in RULES.  Usually the rule number in the source,
 139      except if some rules are useless.  */
 140   short number;
 141
 142   symbol_t *lhs;
 143   item_number_t *rhs;
 144
 145   /* This symbol provides both the associativity, and the precedence. */
 146   symbol_t *prec;
 147
 148   short dprec;
 149   short merger;
 150
 151   /* This symbol was attached to the rule via %prec. */
 152   symbol_t *precsym;
 153
 154   location_t location;
 155   bool useful;
 156
 157   const char *action;
 158   location_t action_location;
 159 } rule_t;
 160
 161 extern struct rule_s *rules;
 162
 163 /* Table of the symbols, indexed by the symbol number. */
 164 extern symbol_t **symbols;
 165
 166 /* TOKEN_TRANSLATION -- a table indexed by a token number as returned
 167    by the user's yylex routine, it yields the internal token number
 168    used by the parser and throughout bison.  */
 169 extern symbol_number_t *token_translations;
 170 extern int max_user_token_number;
 171
 172
 173 /* GLR_PARSER is nonzero if the input file says to use the GLR
 174    (Generalized LR) parser, and to output some additional
 175    information used by the GLR algorithm. */
 176
 177 extern int glr_parser;
 178
 179 /* PURE_PARSER is nonzero if should generate a parser that is all pure
 180    and reentrant.  */
 181
 182 extern int pure_parser;
 183
 184 /* Return the length of the RHS.  */
 185 int rule_rhs_length PARAMS ((rule_t *rule));
 186
 187 /* Print this RULE's RHS on OUT.  */
 188 void rule_rhs_print PARAMS ((rule_t *rule, FILE *out));
 189
 190 /* Print this RULE on OUT.  */
 191 void rule_print PARAMS ((rule_t *rule, FILE *out));
 192
 193 /* Dump RITEM for traces. */
 194 void ritem_print PARAMS ((FILE *out));
 195
 196 /* Return the size of the longest rule RHS.  */
 197 size_t ritem_longest_rhs PARAMS ((void));
 198
 199 /* Print the grammar's rules numbers from BEGIN (inclusive) to END
 200    (exclusive) on OUT under TITLE.  */
 201 void grammar_rules_partial_print PARAMS ((FILE *out, const char *title,
 202                                           int begin, int end));
 203
 204 /* Print the grammar's rules on OUT.  */
 205 void grammar_rules_print PARAMS ((FILE *out));
 206
 207 /* Dump the grammar. */
 208 void grammar_dump PARAMS ((FILE *out, const char *title));
 209
 210 /* Free the packed grammar. */
 211 void grammar_free PARAMS ((void));
 212
 213 #endif /* !GRAM_H_ */