X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/aa7815f5c6e8e07a85e47df9cd7b579468969efb..bfdcc3a0b273c116e01d289ce52d400f22e4da14:/src/gram.h?ds=inline diff --git a/src/gram.h b/src/gram.h index e270db5c..3f75dd8f 100644 --- a/src/gram.h +++ b/src/gram.h @@ -1,133 +1,275 @@ -/* Data definitions for internal representation of bison's input, - Copyright 1984, 1986, 1989, 1992 Free Software Foundation, Inc. +/* Data definitions for internal representation of Bison's input. + + Copyright (C) 1984, 1986, 1989, 1992, 2001-2007, 2009-2012 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 + This program 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. + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. - Bison is distributed in the hope that it will be useful, + This program 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. */ + along with this program. If not, see . */ #ifndef GRAM_H_ # define GRAM_H_ -/* representation of the grammar rules: +/* 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 + + NTOKENS is the number of tokens, and NVARS is the number of + variables (nonterminals). NSYMS is the total number, ntokens + nvars. - (the true number of token values assigned is ntokens reduced by one - for each alias declaration) - 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. + 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, `$accept: START-SYMBOL $end', 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 $end). + + 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 rules, whose 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].dprec -- the dynamic precedence level of R (for GLR + parsing). - The rules receive rule numbers 1 to nrules in the order they are - written. Actions and guards are accessed via the rule number. + RULES[R].merger -- index of merging function for R (for GLR + parsing). - The rules themselves are described by three arrays: rrhs, rlhs and - ritem. rlhs[R] is the symbol number of the left hand side of rule - R. The right hand side is stored as symbol numbers in a portion of - ritem. rrhs[R] contains the index in ritem of the beginning of the - portion for rule R. + RULES[R].line -- the line where R was defined. - If rlhs[R] is -1, the rule has been thrown out by reduce.c and - should be ignored. + 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 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. - Precedence levels are recorded in the vectors sprec and rprec. - sprec records the precedence level of each symbol, rprec the - precedence level of each rule. rprecsym is the symbol-number of - the symbol in %prec for this rule (if any). + 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 rassoc and sassoc. */ + Associativities are recorded similarly in SYMBOLS[I]->assoc. */ +# include "location.h" +# include "symtab.h" -#define ISTOKEN(s) ((s) < ntokens) -#define ISVAR(s) ((s) >= ntokens) +# define ISTOKEN(i) ((i) < ntokens) +# define ISVAR(i) ((i) >= ntokens) -extern int nitems; -extern int nrules; extern int nsyms; extern int ntokens; extern int nvars; -extern short *ritem; -extern short *rlhs; -extern short *rrhs; -extern short *rprec; -extern short *rprecsym; -extern short *sprec; -extern short *rassoc; -extern short *sassoc; -extern short *rline; /* Source line number of each rule */ +typedef int item_number; +#define ITEM_NUMBER_MAX INT_MAX +extern item_number *ritem; +extern unsigned int nritems; + +/* There is weird relationship between OT1H item_number and OTOH + symbol_number and rule_number: we store the latter in + item_number. symbol_number values are stored as-is, while + the negation of (rule_number + 1) is stored. + + Therefore, a symbol_number must be a valid item_number, and we + sometimes have to perform the converse transformation. */ + +static inline item_number +symbol_number_as_item_number (symbol_number sym) +{ + return sym; +} + +static inline symbol_number +item_number_as_symbol_number (item_number i) +{ + return i; +} + +static inline bool +item_number_is_symbol_number (item_number i) +{ + return i >= 0; +} + +/* Rule numbers. */ +typedef int rule_number; +#define RULE_NUMBER_MAX INT_MAX +extern rule_number nrules; + +static inline item_number +rule_number_as_item_number (rule_number r) +{ + return -1 - r; +} + +static inline rule_number +item_number_as_rule_number (item_number i) +{ + return -1 - i; +} -extern int start_symbol; +static inline bool +item_number_is_rule_number (item_number i) +{ + return i < 0; +} +/*--------. +| Rules. | +`--------*/ -/* associativity values in elements of rassoc, sassoc. */ -typedef enum +typedef struct { - right_assoc, - left_assoc, - non_assoc -} associativity; + /* The number of the rule in the source. It is usually the index in + RULES too, except if there are useless rules. */ + rule_number user_number; + + /* The index in RULES. Usually the rule number in the source, + except if some rules are useless. */ + rule_number number; + + symbol *lhs; + item_number *rhs; + + /* This symbol provides both the associativity, and the precedence. */ + symbol *prec; + + int dprec; + int merger; + + /* This symbol was attached to the rule via %prec. */ + symbol *precsym; + + location location; + bool useful; + + const char *action; + location action_location; +} rule; + +extern rule *rules; + +/* A function that selects a rule. */ +typedef bool (*rule_filter) (rule *); + +/* Return true IFF the rule has a `number' smaller than NRULES. That is, it is + useful in the grammar. */ +bool rule_useful_in_grammar_p (rule *r); +/* Return true IFF the rule has a `number' higher than NRULES. That is, it is + useless in the grammar. */ +bool rule_useless_in_grammar_p (rule *r); -/* 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. If translations is zero, the - translation table is not used because the two kinds of token - numbers are the same. (It is noted in reader.c that "Nowadays - translations is always set to 1...") */ +/* Return true IFF the rule is not flagged as useful but is useful in the + grammar. In other words, it was discarded because of conflicts. */ +bool rule_useless_in_parser_p (rule *r); -extern short *token_translations; -extern int translations; +/* Print this rule's number and lhs on OUT. If a PREVIOUS_LHS was + already displayed (by a previous call for another rule), avoid + useless repetitions. */ +void rule_lhs_print (rule *r, symbol *previous_lhs, FILE *out); +void rule_lhs_print_xml (rule *r, FILE *out, int level); + +/* Return the length of the RHS. */ +int rule_rhs_length (rule *r); + +/* Print this rule's RHS on OUT. */ +void rule_rhs_print (rule *r, FILE *out); + +/* Print this rule on OUT. */ +void rule_print (rule *r, FILE *out); + + + + +/* Table of the symbols, indexed by the symbol number. */ +extern symbol **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 symbol_number *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. */ +/* Dump RITEM for traces. */ +void ritem_print (FILE *out); + +/* Return the size of the longest rule RHS. */ +size_t ritem_longest_rhs (void); + +/* Print the grammar's rules that match FILTER on OUT under TITLE. */ +void grammar_rules_partial_print (FILE *out, const char *title, + rule_filter filter); + +/* Print the grammar's useful rules on OUT. */ +void grammar_rules_print (FILE *out); +/* Print all of the grammar's rules with a "usefulness" attribute. */ +void grammar_rules_print_xml (FILE *out, int level); + +/* Dump the grammar. */ +void grammar_dump (FILE *out, const char *title); -extern int pure_parser; +/* Report on STDERR the rules that are not flagged USEFUL, using the + MESSAGE (which can be `rule useless in grammar' when invoked after grammar + reduction, or `rule useless in parser due to conflicts' after conflicts + were taken into account). */ +void grammar_rules_useless_report (const char *message); -/* ERROR_TOKEN_NUMBER is the token number of the error token. */ +/* Free the packed grammar. */ +void grammar_free (void); -extern int error_token_number; #endif /* !GRAM_H_ */