src/gram.h

   1 /* Data definitions for internal representation of Bison's input.
   2
   3    Copyright (C) 1984, 1986, 1989, 1992, 2001, 2002
   4    Free Software Foundation, Inc.
   5
   6    This file is part of Bison, the GNU Compiler Compiler.
   7
   8    Bison is free software; you can redistribute it and/or modify
   9    it under the terms of the GNU General Public License as published by
  10    the Free Software Foundation; either version 2, or (at your option)
  11    any later version.
  12
  13    Bison is distributed in the hope that it will be useful,
  14    but WITHOUT ANY WARRANTY; without even the implied warranty of
  15    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  16    GNU General Public License for more details.
  17
  18    You should have received a copy of the GNU General Public License
  19    along with Bison; see the file COPYING.  If not, write to
  20    the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
  21    Boston, MA 02111-1307, USA.  */
  22
  23 #ifndef GRAM_H_
  24 # define GRAM_H_
  25
  26 /* Representation of the grammar rules:
  27
  28    NTOKENS is the number of tokens, and NVARS is the number of
  29    variables (nonterminals).  NSYMS is the total number, ntokens +
  30    nvars.
  31
  32    Each symbol (either token or variable) receives a symbol number.
  33    Numbers 0 to NTOKENS - 1 are for tokens, and NTOKENS to NSYMS - 1
  34    are for variables.  Symbol number zero is the end-of-input token.
  35    This token is counted in ntokens.  The true number of token values
  36    assigned is NTOKENS reduced by one for each alias declaration.
  37
  38    The rules receive rule numbers 1 to NRULES in the order they are
  39    written.  More precisely Bison augments the grammar with the
  40    initial rule, `$accept: START-SYMBOL $end', which is numbered 1,
  41    all the user rules are 2, 3 etc.  Each time a rule number is
  42    presented to the user, we subtract 1, so *displayed* rule numbers
  43    are 0, 1, 2...
  44
  45    Internally, we cannot use the number 0 for a rule because for
  46    instance RITEM stores both symbol (the RHS) and rule numbers: the
  47    symbols are shorts >= 0, and rule number are stored negative.
  48    Therefore 0 cannot be used, since it would be both the rule number
  49    0, and the token $end).
  50
  51    Actions are accessed via the rule number.
  52
  53    The rules themselves are described by several arrays: amongst which
  54    RITEM, and RULES.
  55
  56    RULES is an array of rules, whose members are:
  57
  58    RULES[R].lhs -- the symbol of the left hand side of rule R.
  59
  60    RULES[R].rhs -- the index in RITEM of the beginning of the portion
  61    for rule R.
  62
  63    RULES[R].prec -- the symbol providing the precedence level of R.
  64
  65    RULES[R].precsym -- the symbol attached (via %prec) to give its
  66    precedence to R.  Of course, if set, it is equal to `prec', but we
  67    need to distinguish one from the other when reducing: a symbol used
  68    in a %prec is not useless.
  69
  70    RULES[R].assoc -- the associativity of R.
  71
  72    RULES[R].dprec -- the dynamic precedence level of R (for GLR
  73    parsing).
  74
  75    RULES[R].merger -- index of merging function for R (for GLR
  76    parsing).
  77
  78    RULES[R].line -- the line where R was defined.
  79
  80    RULES[R].useful -- true iff the rule is used (i.e., false if thrown
  81    away by reduce).
  82
  83    The right hand side is stored as symbol numbers in a portion of
  84    RITEM.
  85
  86    The length of the portion is one greater than the number of symbols
  87    in the rule's right hand side.  The last element in the portion
  88    contains minus R, which identifies it as the end of a portion and
  89    says which rule it is for.
  90
  91    The portions of RITEM come in order of increasing rule number.
  92    NRITEMS is the total length of RITEM.  Each element of RITEM is
  93    called an "item" and its index in RITEM is an item number.
  94
  95    Item numbers are used in the finite state machine to represent
  96    places that parsing can get to.
  97
  98    SYMBOLS[I]->prec records the precedence level of each symbol.
  99
 100    Precedence levels are assigned in increasing order starting with 1
 101    so that numerically higher precedence values mean tighter binding
 102    as they ought to.  Zero as a symbol or rule's precedence means none
 103    is assigned.
 104
 105    Associativities are recorded similarly in SYMBOLS[I]->assoc.  */
 106
 107 # include "location.h"
 108 # include "symtab.h"
 109
 110 # define ISTOKEN(s)     ((s) < ntokens)
 111 # define ISVAR(s)       ((s) >= ntokens)
 112
 113 extern int nsyms;
 114 extern int ntokens;
 115 extern int nvars;
 116
 117 typedef int item_number;
 118 extern item_number *ritem;
 119 extern unsigned int nritems;
 120
 121 /* There is weird relationship between OT1H item_number and OTOH
 122    symbol_number and rule_number: we store the latter in
 123    item_number.  symbol_number values are stored as-is, while
 124    the negation of (rule_number + 1) is stored.
 125
 126    Therefore, a symbol_number must be a valid item_number, and we
 127    sometimes have to perform the converse transformation.  */
 128
 129 static inline item_number
 130 symbol_number_as_item_number (symbol_number s)
 131 {
 132   return s;
 133 }
 134
 135 static inline symbol_number
 136 item_number_as_symbol_number (item_number i)
 137 {
 138   return i;
 139 }
 140
 141 extern symbol_number start_symbol;
 142
 143 /* Rule numbers.  */
 144 typedef short rule_number;
 145 extern rule_number nrules;
 146
 147 static inline item_number
 148 rule_number_as_item_number (rule_number r)
 149 {
 150   return -1 - r;
 151 }
 152
 153 static inline rule_number
 154 item_number_as_rule_number (item_number i)
 155 {
 156   return -1 - i;
 157 }
 158
 159
 160 /*--------.
 161 | Rules.  |
 162 `--------*/
 163
 164 typedef struct
 165 {
 166   /* The number of the rule in the source.  It is usually the index in
 167      RULES too, except if there are useless rules.  */
 168   rule_number user_number;
 169
 170   /* The index in RULES.  Usually the rule number in the source,
 171      except if some rules are useless.  */
 172   rule_number number;
 173
 174   symbol *lhs;
 175   item_number *rhs;
 176
 177   /* This symbol provides both the associativity, and the precedence. */
 178   symbol *prec;
 179
 180   short dprec;
 181   short merger;
 182
 183   /* This symbol was attached to the rule via %prec. */
 184   symbol *precsym;
 185
 186   location location;
 187   bool useful;
 188
 189   const char *action;
 190   location action_location;
 191 } rule;
 192
 193 extern rule *rules;
 194
 195 /* A function that selects a rule.  */
 196 typedef bool (*rule_filter) (rule *);
 197
 198 /* Return true IFF the rule has a `number' smaller than NRULES.  */
 199 bool rule_useful_p (rule *r);
 200
 201 /* Return true IFF the rule has a `number' higher than NRULES.  */
 202 bool rule_useless_p (rule *r);
 203
 204 /* Return true IFF the rule is not flagged as useful *and* is useful.
 205    In other words, it was discarded because of conflicts.  */
 206 bool rule_never_reduced_p (rule *r);
 207
 208 /* Print this rule's number and lhs on OUT.  If a PREVIOUS_LHS was
 209    already displayed (by a previous call for another rule), avoid
 210    useless repetitions.  */
 211 void rule_lhs_print (rule *r, symbol *previous_lhs, FILE *out);
 212
 213 /* Return the length of the RHS.  */
 214 int rule_rhs_length (rule *r);
 215
 216 /* Print this rule's RHS on OUT.  */
 217 void rule_rhs_print (rule *r, FILE *out);
 218
 219 /* Print this rule on OUT.  */
 220 void rule_print (rule *r, FILE *out);
 221
 222
 223
 224
 225 /* Table of the symbols, indexed by the symbol number. */
 226 extern symbol **symbols;
 227
 228 /* TOKEN_TRANSLATION -- a table indexed by a token number as returned
 229    by the user's yylex routine, it yields the internal token number
 230    used by the parser and throughout bison.  */
 231 extern symbol_number *token_translations;
 232 extern int max_user_token_number;
 233
 234
 235
 236 /* GLR_PARSER is nonzero if the input file says to use the GLR
 237    (Generalized LR) parser, and to output some additional
 238    information used by the GLR algorithm. */
 239
 240 extern int glr_parser;
 241
 242 /* PURE_PARSER is nonzero if should generate a parser that is all pure
 243    and reentrant.  */
 244
 245 extern int pure_parser;
 246
 247 /* Dump RITEM for traces. */
 248 void ritem_print (FILE *out);
 249
 250 /* Return the size of the longest rule RHS.  */
 251 size_t ritem_longest_rhs (void);
 252
 253 /* Print the grammar's rules numbers from BEGIN (inclusive) to END
 254    (exclusive) on OUT under TITLE.  */
 255 void grammar_rules_partial_print (FILE *out, const char *title,
 256                                   rule_filter filter);
 257
 258 /* Print the grammar's rules on OUT.  */
 259 void grammar_rules_print (FILE *out);
 260
 261 /* Dump the grammar. */
 262 void grammar_dump (FILE *out, const char *title);
 263
 264 /* Report on STDERR the rules that are not flagged USEFUL, using the
 265    MESSAGE (which can be `useless rule' when invoked after grammar
 266    reduction, or `never reduced' after conflicts were taken into
 267    account).  */
 268 void grammar_rules_never_reduced_report (const char *message);
 269
 270 /* Free the packed grammar. */
 271 void grammar_free (void);
 272
 273 #endif /* !GRAM_H_ */