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, `$accept: START-SYMBOL $end', which is numbered 1,
  40    all the user rules are 2, 3 etc.  Each time a rule number is
  41    presented to the user, we subtract 1, so *displayed* rule numbers
  42    are 0, 1, 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 $end).
  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
  72    parsing).
  73
  74    RULES[R].merger -- index of merging function for R (for GLR
  75    parsing).
  76
  77    RULES[R].line -- the line where R was defined.
  78
  79    RULES[R].useful -- true iff the rule is used (i.e., false if thrown
  80    away by reduce).
  81
  82    The right hand side is stored as symbol numbers in a portion of
  83    RITEM.
  84
  85    The length of the portion is one greater than the number of symbols
  86    in the rule's right hand side.  The last element in the portion
  87    contains minus R, which identifies it as the end of a portion and
  88    says which rule it is for.
  89
  90    The portions of RITEM come in order of increasing rule number.
  91    NRITEMS is the total length of RITEM.  Each element of RITEM is
  92    called an "item" and its index in RITEM is an item number.
  93
  94    Item numbers are used in the finite state machine to represent
  95    places that parsing can get to.
  96
  97    SYMBOLS[I]->prec records the precedence level of each symbol.
  98
  99    Precedence levels are assigned in increasing order starting with 1
 100    so that numerically higher precedence values mean tighter binding
 101    as they ought to.  Zero as a symbol or rule's precedence means none
 102    is assigned.
 103
 104    Associativities are recorded similarly in SYMBOLS[I]->assoc.  */
 105
 106 # include "location.h"
 107 # include "symtab.h"
 108
 109 # define ISTOKEN(s)     ((s) < ntokens)
 110 # define ISVAR(s)       ((s) >= ntokens)
 111
 112 extern int nsyms;
 113 extern int ntokens;
 114 extern int nvars;
 115
 116 typedef int item_number_t;
 117 # define ITEM_NUMBER_MAX ((item_number_t) INT_MAX)
 118 # define ITEM_NUMBER_MIN ((item_number_t) INT_MIN)
 119 extern item_number_t *ritem;
 120 extern unsigned int nritems;
 121
 122 /* There is weird relationship between OT1H item_number_t and OTOH
 123    symbol_number_t and rule_number_t: we store the latter in
 124    item_number_t.  symbol_number_t are stored as are, while
 125    the negation of (rule_number_t + 1) are stored.
 126
 127    Therefore, an symbol_number_t must be a valid item_number_t, and we
 128    sometimes have to perform the converse transformation.  */
 129 # define symbol_number_as_item_number(Tok) ((item_number_t) (Tok))
 130 # define item_number_as_symbol_number(Ite) ((symbol_number_t) (Ite))
 131
 132 extern symbol_number_t start_symbol;
 133
 134 /* Rules numbers. */
 135 typedef short rule_number_t;
 136 # define RULE_NUMBER_MAX ((rule_number_t) SHRT_MAX)
 137 extern rule_number_t nrules;
 138 # define int_of_rule_number(RNum) ((int) (RNum))
 139 # define rule_number_as_item_number(RNum) ((item_number_t) (- RNum - 1))
 140 # define item_number_as_rule_number(INum) ((rule_number_t) (- INum - 1))
 141
 142
 143 /*--------.
 144 | Rules.  |
 145 `--------*/
 146
 147 typedef struct rule_s
 148 {
 149   /* The number of the rule in the source.  It is usually the index in
 150      RULES too, except if there are useless rules.  */
 151   rule_number_t user_number;
 152
 153   /* The index in RULES.  Usually the rule number in the source,
 154      except if some rules are useless.  */
 155   rule_number_t number;
 156
 157   symbol_t *lhs;
 158   item_number_t *rhs;
 159
 160   /* This symbol provides both the associativity, and the precedence. */
 161   symbol_t *prec;
 162
 163   short dprec;
 164   short merger;
 165
 166   /* This symbol was attached to the rule via %prec. */
 167   symbol_t *precsym;
 168
 169   location_t location;
 170   bool useful;
 171
 172   const char *action;
 173   location_t action_location;
 174 } rule_t;
 175
 176 extern struct rule_s *rules;
 177
 178 /* A function that selects a rule.  */
 179 typedef bool (*rule_filter_t) (rule_t *r);
 180
 181 /* Return true IFF the rule has a `number' smaller than NRULES.  */
 182 bool rule_useful_p (rule_t *r);
 183
 184 /* Return true IFF the rule has a `number' higher than NRULES.  */
 185 bool rule_useless_p (rule_t *r);
 186
 187 /* Return true IFF the rule is not flagged as useful *and* is useful.
 188    In other words, it was discarded because of conflicts.  */
 189 bool rule_never_reduced_p (rule_t *r);
 190
 191 /* Print this RULE's number and lhs on OUT.  If a PREVIOUS_LHS was
 192    already displayed (by a previous call for another rule), avoid
 193    useless repetitions.  */
 194 void rule_lhs_print (rule_t *rule, symbol_t *previous_lhs, FILE *out);
 195
 196 /* Return the length of the RHS.  */
 197 int rule_rhs_length (rule_t *rule);
 198
 199 /* Print this RULE's RHS on OUT.  */
 200 void rule_rhs_print (rule_t *rule, FILE *out);
 201
 202 /* Print this RULE on OUT.  */
 203 void rule_print (rule_t *rule, FILE *out);
 204
 205
 206
 207
 208 /* Table of the symbols, indexed by the symbol number. */
 209 extern symbol_t **symbols;
 210
 211 /* TOKEN_TRANSLATION -- a table indexed by a token number as returned
 212    by the user's yylex routine, it yields the internal token number
 213    used by the parser and throughout bison.  */
 214 extern symbol_number_t *token_translations;
 215 extern int max_user_token_number;
 216
 217
 218
 219 /* GLR_PARSER is nonzero if the input file says to use the GLR
 220    (Generalized LR) parser, and to output some additional
 221    information used by the GLR algorithm. */
 222
 223 extern int glr_parser;
 224
 225 /* PURE_PARSER is nonzero if should generate a parser that is all pure
 226    and reentrant.  */
 227
 228 extern int pure_parser;
 229
 230 /* Dump RITEM for traces. */
 231 void ritem_print (FILE *out);
 232
 233 /* Return the size of the longest rule RHS.  */
 234 size_t ritem_longest_rhs (void);
 235
 236 /* Print the grammar's rules numbers from BEGIN (inclusive) to END
 237    (exclusive) on OUT under TITLE.  */
 238 void grammar_rules_partial_print (FILE *out, const char *title,
 239                                   rule_filter_t filter);
 240
 241 /* Print the grammar's rules on OUT.  */
 242 void grammar_rules_print (FILE *out);
 243
 244 /* Dump the grammar. */
 245 void grammar_dump (FILE *out, const char *title);
 246
 247 /* Report on STDERR the rules that are not flagged USEFUL, using the
 248    MESSAGE (which can be `useless rule' when invoked after grammar
 249    reduction, or `never reduced' after conflicts were taken into
 250    account).  */
 251 void grammar_rules_never_reduced_report (const char *message);
 252
 253 /* Free the packed grammar. */
 254 void grammar_free (void);
 255
 256 #endif /* !GRAM_H_ */