How can I fix my ambiguity in bison file?

Question

I m working on a simple interpreter and I m using bison and flex for my lexer and parser.I have a problem when it comes to the parser :

progr          :decl_classes decl_gvars  decl_functions block {printf("The programme is correct!\n");}
               ;/*1. classes 2.global vars 3.functii 4.enetry point*/

decl_classes   :decl_class ';' decl_classes
                   |{cout<<"No more classes\n";}
               ;
decl_class     :CLASS ID '{' decl_gvars decl_functions '}'  {cout<<"We have a class :"<<$2<<"\n";}
               ;    

decl_functions :{cout<<"No more functions\n";}
               |decl_function ';' decl_functions
               ;
decl_function  :TYPE ID '('list_param')' '{'list'}' 
               
               ;
decl_gvars     :decl_gvar ';' decl_gvars 
               |{cout<<"No more gvars\n";}
               ;
decl_gvar      :typedecl list_ID  {cout<<"We have vars "<<endl;}
               ;


typedecl : TYPE {cout<<"Try to make a var\n";}
           ;

list_ID : ID '[' expr ']' 
        | ID 
        | ID ',' list_ID 
        ;
list_param :param ',' list_param
            |/*none*/
            ;
            
param     :typedecl ID
          |typedecl ID '['expr']'
          ; 
      

block : BGIN list END  
     ;
     

list :  statement ';' 
     | list statement ';'
     |/*none*/
     ;

the parser will always chose to make decl_gvar instead of decl_function. Any help would be apriciated.

I tested if the parser will make my decl_function if it s isolated and it does.

Answer 1

The basic problem is that after seeing a TYPE when the next token is an ID it doesn't yet know if it is looking at a decl_function or a decl_gvar, but your grammar forces it to reduce a typedecl to recognize a decl_gvar or not reduce it to recognize a decl_function. This results in a shift/reduce conflict (which bison should report). Since it will by default do the shift in preference to the reduce, the parser will always try to parse this as a decl_function, even if it is in fact a decl_gvar (and will in that case get a syntax error.)

The easiest fix is to get rid of of the typedecl : TYPE rule (and replace typedecl with TYPE in the decl_gvar rule.) This way, you don't need the reduction and things will be deferred until after the ID token, so it will be able to see a ( to recognize a function or a , or ; to recognize a gvar.