Bison语法中无用的规则和非终结符

Question

I have the following Bison file describing my scripting language grammar:

%error-verbose
%{
#include "node.h"
NBlock *programBlock;

#define YYDEBUG 1

extern int yylex();
void yyerror(const char *) { printf("Error: %s\n", s); }
%}

%union {
    Node *node;
    NBlock *block;
    NBody *body;
    NHeader *header;
    NExpression *expression;
    NStatement *statement;
    NIdentifier *identifier;
    NVariableDeclaration *variableDeclaration;
    NDoWhileStatement *doWhileStatement;
    NWhileStatement *whileStatement;
    NIfStatement *ifStatement;
    NForStatement *forStatement;
    std::vector<NVariableDeclaration*> *variableDeclarations;
    std::vector<NExpression*> *expressions;
    std::vector<NStatement*> *statements;
    std:string *string;
    int token;
}


%token <string> TIDENTIFIER TINTEGER TDOUBLE
%token <token> TCEQ TCNE TCLT TCLE TCGT TCGE TEQUAL
%token <token> TLPAREN TRPAREN TLBRACE TRBRACE TCOMMA TDOT
%token <token> TPLUS TMINUS TMUL TDIV TDO TDOUBLE_TYPE TINT_TYPE
%token <token> TELSE TFOR TIF TSEMICOLON TTHEN TWHILE

%type <expression> expression term factor
%type <block> program body header tail statements
%type <statement> statement forStatement ifStatement doWhileStatement whileStatement variableDeclaration
%type <token> comparison
%type <string> identifier_type


%left TPLUS TMINUS
%left TMUL TDIV

%start program

%%

program:                TLBRACE body TRBRACE { printf("Reduce body to program\n"); }
                        ;

body:                   header TLBRACE block TRBRACE tail { printf("Reduce header block tail to body\n"); }
                        ;

header:                 TLBRACE variableDeclarations TRBRACE { printf("Reduce variableDeclarations to header\n"); }
                        | TLBRACE TRBRACE { printf("Reduce empty to header\n"); }
                        ;

variableDeclarations:   variableDeclaration TSEMICOLON { printf("Reduce variable declaration to header\n"); }
                        | variableDeclarations variableDeclaration TSEMICOLON { printf("Reduce variable declaration list to header\n"); }
                        ;

tail:                   TLBRACE statements TRBRACE { printf("reduce statement list to tail\n"); }
                        | TLBRACE TRBRACE { printf("Reduce empty to tal\n"); }
                        ;                   

statements:             statement TSEMICOLON { printf("Reduce statement to statement list\n"); }
                        | statements statement TSEMICOLON { printf("Reduce statement list to statement list\n"); }
                        ;

statement:              ifStatement { printf("Reduce if to statement\n"); };
                        | forStatement { printf("Reduce for to statement\n"); };
                        | doWhileStatement { printf("Reduce doWhile to statement\n"); };
                        | whileStatement { printf("reduce while to statement\n"); }
                        | expression { printf("Reduce expression to statement\n"); };
                        ;

forStatement:           TFOR TLPAREN expression TSEMICOLON expression TSEMICOLON expression TRPAREN block { printf("Reduce for to for statement\n"); }
                        ;


ifStatement:            TIF expression TTHEN block { printf("Reduce if to if statement\n"); }
                        | TIF expression block TELSE block { printf("Reduce ifelse to if statement\n"); }
                        ;

doWhileStatement:       TDO block TWHILE expression { printf("reduce dowhile to while statement\n"); }
                        ;

whileStatement:         TWHILE block expression { printf("Reduce while to while statement\n"); }
                        ;

block:                  TLBRACE statements TRBRACE { printf("Reduce statement list to block\n"); }
                        | TLBRACE TRBRACE { printf("Reduce empty to block\n"); }
                        ;

variableDeclaration:    identifier_type TIDENTIFIER { printf("reduce uninitialized identifier to variable declaration\n"); }
                        | identifier_type TIDENTIFIER TEQUAL expression { printf("Reduce initialized identifier to variable declaration\n"); }
                        ;

identifier_type:        TINT_TYPE { printf("Reduce int to identifier type\n"); }
                        | TDOUBLE_TYPE { printf("Reduce double to identifier type\n"); }
                        | { printf("Reduce empty to identifier type\n"); }
                        ;



expression:             TIDENTIFIER TEQUAL expression { printf("Reduce assignment to expression\n"); }
                        | TIDENTIFIER { printf("reduce identifier to expression\n"); }
                        | expression comparison expression { printf("Reduce comparison to expression\n"); }
                        | TLPAREN expression TRPAREN { printf("Reduce nested expression to expression\n"); }
                        | expression TPLUS term { printf("Reduce addition to expression\n"); }
                        | expression  TMINUS term { printf("Reduce subtraction to expression\n"); }
                        | term { printf("Reduce term to expresson\n"); }
                        | factor { printf("Reduce factor to expression\n"); }
                        ;

term:                   term TMUL factor { printf("Reduce multiplication to term\n"); }
                        | term TDIV factor { printf("Reduce division to term\n"); }
                        ;

factor:                 TIDENTIFIER { printf("Reduce identifier to factor\n"); }
                        | TINTEGER { printf("Reduce integer to numeric\n"); }
                        | TDOUBLE { printf("Reduce double to numeric\n"); }
                        ;

comparison:             TCEQ { printf("Reduce eq to comparison\n"); }
                        | TCNE { printf("Reduce ne to comparison\n"); }
                        | TCLT { printf("Reduce lt to comparison\n"); }
                        | TCLE { printf("Reduce le to comparison\n"); }
                        | TCGT { printf("reduce gt to comparison\n"); }
                        | TCGE { printf("Reduce ge to comparison\n"); }
                        | TPLUS { printf("Reduce plus to comparison\n"); }
                        | TMINUS { printf("Reduce minus to comparison\n"); }
                        | TMUL { printf("Reduce mul to comparison\n"); }
                        | TDIV { printf("Reduce div to comparison\n"); }
                        ; 

I really don't understand why I am getting errors related to useless nonterminals and useless rules.

Running the command:

bison -d -o parser.cpp parser.y

I get the following errors:

$ bison -d -o parser.cpp parser.y
parser.y: attention: 1 nonterminal useless in grammar
parser.y: attention: 5 rules useless in grammar
parser.y:52.31-34: attention: nonterminal useless in grammar: term
parser.y:135.51-118: attention: rule useless in grammar: expression: expression TPLUS term
parser.y:136.51-123: attention: rule useless in grammar: expression: expression TMINUS term
parser.y:137.51-96: attention: rule useless in grammar: expression: term
parser.y:141.41-103: attention: rule useless in grammar: term: term TMUL factor
parser.y:142.51-107: attention: rule useless in grammar: term: term TDIV factor
parser.y: conflicts: 20 shift/reduce, 14 reduce/reduce
parser.y:145.41-96: attention: rule useless in parser due to conflicts: factor: TIDENTIFIER

Answer 1

I think the conflict is that the parser will not know which rule to chose when it finds a TIDENTIFIER. The reason is that

TIDENTIFIER can be reduced to expression
factor can be reduced to expression
TIDENTIFIER can be reduced to factor

So when the parser get an identifier it can go two ways to reach an expression.

TIDENTIFIER -> expression  
TIDENTIFIER -> factor -> expression

I thin in your case you can safely remove TIDENTIFIER from factor since factor is reduced to ONLY expression anyway.

However a more proper way (at least for most common languages) might be to keep TIDENTIFIER in factor and handle the rest in statement . For example TIDENTIFIER ASSIGN expression can be a valid statement .

BTW your statement includes expression . Which is proably not what you intended. statement should always be different from expression (after all that's why you need two rules).

I think you should check for cases like that. If you have a rule using another rule only (without any other tokens or other rules) then it is probably wrong. Wrong in a sense that it is not what you intended .