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 and guards 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].line -- the line where R was defined.
  72
  73    RULES[R].useful -- TRUE iff the rule is used (i.e., FALSE if thrown
  74    away by reduce).
  75
  76    The right hand side is stored as symbol numbers in a portion of
  77    RITEM.
  78
  79    The length of the portion is one greater than the number of symbols
  80    in the rule's right hand side.  The last element in the portion
  81    contains minus R, which identifies it as the end of a portion and
  82    says which rule it is for.
  83
  84    The portions of RITEM come in order of increasing rule number and
  85    are followed by an element which is zero to mark the end.  nritems
  86    is the total length of ritem, not counting the final zero.  Each
  87    element of RITEM is called an "item" and its index in RITEM is an
  88    item number.
  89
  90    Item numbers are used in the finite state machine to represent
  91    places that parsing can get to.
  92
  93    SYMBOLS[I]->prec records the precedence level of each symbol.
  94
  95    Precedence levels are assigned in increasing order starting with 1
  96    so that numerically higher precedence values mean tighter binding
  97    as they ought to.  Zero as a symbol or rule's precedence means none
  98    is assigned.
  99
 100    Associativities are recorded similarly in SYMBOLS[I]->assoc.  */
 101
 102 #include "symtab.h"
 103
 104 #define ISTOKEN(s)      ((s) < ntokens)
 105 #define ISVAR(s)        ((s) >= ntokens)
 106
 107 extern int nrules;
 108 extern int nsyms;
 109 extern int ntokens;
 110 extern int nvars;
 111
 112 #define ITEM_NUMBER_MAX INT_MAX
 113 typedef int item_number_t;
 114 extern item_number_t *ritem;
 115 extern int nritems;
 116
 117 /* There is weird relationship between item_number_t and
 118    token_number_t: we store token_number_t in item_number_t, but in
 119    the latter we also store, as negative numbers, the rule numbers.
 120
 121    Therefore, an token_number_t must be a valid item_number_t, and we
 122    sometimes have to perform the converse transformation.  */
 123 #define token_number_as_item_number(Tok) ((item_number_t) (Tok))
 124 #define item_number_as_token_number(Ite) ((token_number_t) (Ite))
 125
 126 extern token_number_t start_symbol;
 127
 128
 129 typedef struct rule_s
 130 {
 131   /* The number of the rule in the source.  It is usually the index in
 132      RULES too, except if there are useless rules.  */
 133   short user_number;
 134
 135   /* The index in RULES.  Usually the rule number in the source,
 136      except if some rules are useless.  */
 137   short number;
 138
 139   symbol_t *lhs;
 140   item_number_t *rhs;
 141
 142   /* This symbol provides both the associativity, and the precedence. */
 143   symbol_t *prec;
 144
 145   /* This symbol was attached to the rule via %prec. */
 146   symbol_t *precsym;
 147
 148   int line;
 149   bool useful;
 150
 151   const char *action;
 152   int action_line;
 153
 154   const char *guard;
 155   int guard_line;
 156 } rule_t;
 157
 158 extern struct rule_s *rules;
 159
 160 /* Table of the symbols, indexed by the symbol number. */
 161 extern symbol_t **symbols;
 162
 163 /* TOKEN_TRANSLATION -- a table indexed by a token number as returned
 164    by the user's yylex routine, it yields the internal token number
 165    used by the parser and throughout bison.  */
 166 extern token_number_t *token_translations;
 167 extern int max_user_token_number;
 168
 169 /* SEMANTIC_PARSER is nonzero if the input file says to use the hairy
 170    parser that provides for semantic error recovery.  If it is zero,
 171    the yacc-compatible simplified parser is used.  */
 172
 173 extern int semantic_parser;
 174
 175 /* PURE_PARSER is nonzero if should generate a parser that is all pure
 176    and reentrant.  */
 177
 178 extern int pure_parser;
 179
 180 /* Report the length of the RHS. */
 181 int rule_rhs_length PARAMS ((rule_t *rule));
 182
 183 /* Dump RITEM for traces. */
 184 void ritem_print PARAMS ((FILE *out));
 185
 186 /* Return the size of the longest rule RHS.  */
 187 size_t ritem_longest_rhs PARAMS ((void));
 188
 189 /* Dump the grammar. */
 190 void grammar_dump PARAMS ((FILE *out, const char *title));
 191
 192 #endif /* !GRAM_H_ */