src/gram.h

/* Data definitions for internal representation of bison's input,
   Copyright (C) 1984, 1986, 1989, 1992, 2001, 2002
   Free Software Foundation, Inc.

   This file is part of Bison, the GNU Compiler Compiler.

   Bison is free software; you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
   the Free Software Foundation; either version 2, or (at your option)
   any later version.

   Bison is distributed in the hope that it will be useful,
   but WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
   GNU General Public License for more details.

   You should have received a copy of the GNU General Public License
   along with Bison; see the file COPYING.  If not, write to
   the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
   Boston, MA 02111-1307, USA.  */

#ifndef GRAM_H_
# define GRAM_H_

/* Representation of the grammar rules:

   NTOKENS is the number of tokens, and NVARS is the number of
   variables (nonterminals).  NSYMS is the total number, ntokens +
   nvars.

   Each symbol (either token or variable) receives a symbol number.
   Numbers 0 to NTOKENS - 1 are for tokens, and NTOKENS to NSYMS - 1
   are for variables.  Symbol number zero is the end-of-input token.
   This token is counted in ntokens.  The true number of token values
   assigned is NTOKENS reduced by one for each alias declaration.

   The rules receive rule numbers 1 to NRULES in the order they are
   written.  More precisely Bison augments the grammar with the
   initial rule, `$axiom: START-SYMBOL EOF', which is numbered 1, all
   the user rules are 2, 3 etc.  Each time a rule number is presented
   to the user, we subtract 1, so *displayed* rule numbers are 0, 1,
   2...

   Internally, we cannot use the number 0 for a rule because for
   instance RITEM stores both symbol (the RHS) and rule numbers: the
   symbols are shorts >= 0, and rule number are stored negative.
   Therefore 0 cannot be used, since it would be both the rule number
   0, and the token EOF).

   Actions are accessed via the rule number.

   The rules themselves are described by several arrays: amongst which
   RITEM, and RULES.

   RULES is an array of struct rule_s, which members are:

   RULES[R].lhs -- the symbol of the left hand side of rule R.

   RULES[R].rhs -- the index in RITEM of the beginning of the portion
   for rule R.

   RULES[R].prec -- the symbol providing the precedence level of R.

   RULES[R].precsym -- the symbol attached (via %prec) to give its
   precedence to R.  Of course, if set, it is equal to `prec', but we
   need to distinguish one from the other when reducing: a symbol used
   in a %prec is not useless.

   RULES[R].assoc -- the associativity of R.

   RULES[R].line -- the line where R was defined.

   RULES[R].useful -- TRUE iff the rule is used (i.e., FALSE if thrown
   away by reduce).

   The right hand side is stored as symbol numbers in a portion of
   RITEM.

   The length of the portion is one greater than the number of symbols
   in the rule's right hand side.  The last element in the portion
   contains minus R, which identifies it as the end of a portion and
   says which rule it is for.

   The portions of RITEM come in order of increasing rule number.
   NRITEMS is the total length of RITEM.  Each element of RITEM is
   called an "item" and its index in RITEM is an item number.

   Item numbers are used in the finite state machine to represent
   places that parsing can get to.

   SYMBOLS[I]->prec records the precedence level of each symbol.

   Precedence levels are assigned in increasing order starting with 1
   so that numerically higher precedence values mean tighter binding
   as they ought to.  Zero as a symbol or rule's precedence means none
   is assigned.

   Associativities are recorded similarly in SYMBOLS[I]->assoc.  */

#include "symtab.h"

#define	ISTOKEN(s)	((s) < ntokens)
#define	ISVAR(s)	((s) >= ntokens)

extern int nrules;
extern int nsyms;
extern int ntokens;
extern int nvars;

#define ITEM_NUMBER_MAX INT_MAX
typedef int item_number_t;
extern item_number_t *ritem;
extern unsigned int nritems;

/* There is weird relationship between item_number_t and
   token_number_t: we store token_number_t in item_number_t, but in
   the latter we also store, as negative numbers, the rule numbers.

   Therefore, an token_number_t must be a valid item_number_t, and we
   sometimes have to perform the converse transformation.  */
#define token_number_as_item_number(Tok) ((item_number_t) (Tok))
#define item_number_as_token_number(Ite) ((token_number_t) (Ite))

extern token_number_t start_symbol;


typedef struct rule_s
{
  /* The number of the rule in the source.  It is usually the index in
     RULES too, except if there are useless rules.  */
  short user_number;

  /* The index in RULES.  Usually the rule number in the source,
     except if some rules are useless.  */
  short number;

  symbol_t *lhs;
  item_number_t *rhs;

  /* This symbol provides both the associativity, and the precedence. */
  symbol_t *prec;

  /* This symbol was attached to the rule via %prec. */
  symbol_t *precsym;

  int line;
  bool useful;

  const char *action;
  int action_line;
} rule_t;

extern struct rule_s *rules;

/* Table of the symbols, indexed by the symbol number. */
extern symbol_t **symbols;

/* TOKEN_TRANSLATION -- a table indexed by a token number as returned
   by the user's yylex routine, it yields the internal token number
   used by the parser and throughout bison.  */
extern token_number_t *token_translations;
extern int max_user_token_number;


/* PURE_PARSER is nonzero if should generate a parser that is all pure
   and reentrant.  */

extern int pure_parser;

/* Report the length of the RHS. */
int rule_rhs_length PARAMS ((rule_t *rule));

/* Dump RITEM for traces. */
void ritem_print PARAMS ((FILE *out));

/* Return the size of the longest rule RHS.  */
size_t ritem_longest_rhs PARAMS ((void));

/* Dump the grammar. */
void grammar_dump PARAMS ((FILE *out, const char *title));

/* Free the packed grammar. */
void grammar_free PARAMS ((void));

#endif /* !GRAM_H_ */
Commit	Line	Data
	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].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.
	85	NRITEMS is the total length of RITEM. Each element of RITEM is
	86	called an "item" and its index in RITEM is an item number.
	87
	88	Item numbers are used in the finite state machine to represent
	89	places that parsing can get to.
	90
	91	SYMBOLS[I]->prec records the precedence level of each symbol.
	92
	93	Precedence levels are assigned in increasing order starting with 1
	94	so that numerically higher precedence values mean tighter binding
	95	as they ought to. Zero as a symbol or rule's precedence means none
	96	is assigned.
	97
	98	Associativities are recorded similarly in SYMBOLS[I]->assoc. */
	99
	100	#include "symtab.h"
	101
	102	#define ISTOKEN(s) ((s) < ntokens)
	103	#define ISVAR(s) ((s) >= ntokens)
	104
	105	extern int nrules;
	106	extern int nsyms;
	107	extern int ntokens;
	108	extern int nvars;
	109
	110	#define ITEM_NUMBER_MAX INT_MAX
	111	typedef int item_number_t;
	112	extern item_number_t *ritem;
	113	extern unsigned int nritems;
	114
	115	/* There is weird relationship between item_number_t and
	116	token_number_t: we store token_number_t in item_number_t, but in
	117	the latter we also store, as negative numbers, the rule numbers.
	118
	119	Therefore, an token_number_t must be a valid item_number_t, and we
	120	sometimes have to perform the converse transformation. */
	121	#define token_number_as_item_number(Tok) ((item_number_t) (Tok))
	122	#define item_number_as_token_number(Ite) ((token_number_t) (Ite))
	123
	124	extern token_number_t start_symbol;
	125
	126
	127	typedef struct rule_s
	128	{
	129	/* The number of the rule in the source. It is usually the index in
	130	RULES too, except if there are useless rules. */
	131	short user_number;
	132
	133	/* The index in RULES. Usually the rule number in the source,
	134	except if some rules are useless. */
	135	short number;
	136
	137	symbol_t *lhs;
	138	item_number_t *rhs;
	139
	140	/* This symbol provides both the associativity, and the precedence. */
	141	symbol_t *prec;
	142
	143	/* This symbol was attached to the rule via %prec. */
	144	symbol_t *precsym;
	145
	146	int line;
	147	bool useful;
	148
	149	const char *action;
	150	int action_line;
	151	} rule_t;
	152
	153	extern struct rule_s *rules;
	154
	155	/* Table of the symbols, indexed by the symbol number. */
	156	extern symbol_t **symbols;
	157
	158	/* TOKEN_TRANSLATION -- a table indexed by a token number as returned
	159	by the user's yylex routine, it yields the internal token number
	160	used by the parser and throughout bison. */
	161	extern token_number_t *token_translations;
	162	extern int max_user_token_number;
	163
	164
	165	/* PURE_PARSER is nonzero if should generate a parser that is all pure
	166	and reentrant. */
	167
	168	extern int pure_parser;
	169
	170	/* Report the length of the RHS. */
	171	int rule_rhs_length PARAMS ((rule_t *rule));
	172
	173	/* Dump RITEM for traces. */
	174	void ritem_print PARAMS ((FILE *out));
	175
	176	/* Return the size of the longest rule RHS. */
	177	size_t ritem_longest_rhs PARAMS ((void));
	178
	179	/* Dump the grammar. */
	180	void grammar_dump PARAMS ((FILE out, const char title));
	181
	182	/* Free the packed grammar. */
	183	void grammar_free PARAMS ((void));
	184
	185	#endif /* !GRAM_H_ */