X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/99013900121bb2de1197f5d0768749f9e50ccb8a..0c2d3f4cba35c4bbea660794ab3cf91c030aafd3:/src/gram.h diff --git a/src/gram.h b/src/gram.h index 61631640..0a92993c 100644 --- a/src/gram.h +++ b/src/gram.h @@ -1,5 +1,5 @@ /* Data definitions for internal representation of bison's input, - Copyright 1984, 1986, 1989, 1992, 2001, 2002 + Copyright (C) 1984, 1986, 1989, 1992, 2001, 2002 Free Software Foundation, Inc. This file is part of Bison, the GNU Compiler Compiler. @@ -47,30 +47,31 @@ Therefore 0 cannot be used, since it would be both the rule number 0, and the token EOF). - Actions and guards are accessed via the rule number. + 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 number of the left hand side of rule R. - If -1, the rule has been thrown out by reduce.c and should be - ignored. + 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 precedence level of R. + RULES[R].prec -- the symbol providing the precedence level of R. - RULES[R].precsym -- the symbol-number of the symbol in %prec for R - (if any). + 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. + 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. @@ -80,11 +81,9 @@ 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 and - are followed by an element which is zero to mark the end. nitems - is the total length of ritem, not counting the final zero. Each - element of RITEM is called an "item" and its index in RITEM is an - item number. + 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. @@ -98,74 +97,78 @@ Associativities are recorded similarly in SYMBOLS[I]->assoc. */ +#include "symtab.h" #define ISTOKEN(s) ((s) < ntokens) #define ISVAR(s) ((s) >= ntokens) -extern int nitems; extern int nrules; extern int nsyms; extern int ntokens; extern int nvars; -extern short *ritem; -extern int nritems; +#define ITEM_NUMBER_MAX INT_MAX +typedef int item_number_t; +extern item_number_t *ritem; +extern unsigned int nritems; -extern int start_symbol; +/* 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. -/* Associativity values for tokens and rules. */ -typedef enum -{ - right_assoc, - left_assoc, - non_assoc -} associativity; + 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 { - short lhs; - short *rhs; - short prec; - short precsym; - associativity assoc; - short line; + /* 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; - short action_line; - - const char *guard; - short guard_line; + int action_line; } rule_t; extern struct rule_s *rules; /* Table of the symbols, indexed by the symbol number. */ -extern struct bucket **symbols; - -/* token translation 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 symbol_t **symbols; -extern short *token_translations; +/* 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; -/* SEMANTIC_PARSER is nonzero if the input file says to use the hairy - parser that provides for semantic error recovery. If it is zero, - the yacc-compatible simplified parser is used. */ - -extern int semantic_parser; /* PURE_PARSER is nonzero if should generate a parser that is all pure and reentrant. */ extern int pure_parser; -/* ERROR_TOKEN_NUMBER is the token number of the error token. */ - -extern int error_token_number; - +/* Report the length of the RHS. */ +int rule_rhs_length PARAMS ((rule_t *rule)); /* Dump RITEM for traces. */ void ritem_print PARAMS ((FILE *out)); @@ -173,4 +176,10 @@ 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_ */