[antlr-interest] Lexer rule with alternatives (Newbie question)

[Johannes Luber]

Петров Александр schrieb:
> Hello,
> why such grammar doesn't work properly:
> --------------------------------------------------------------------------------------
> expr	:
>         operand ((oper) operand)*
>       	;
> //Correct operands
> operand	:
> 	INT
> 	;
> OPER: ('+'|'-'|'*'|'/')
> 	;
> INT	:	
> 	('0'..'9')+
> 	;
> --------------------------------------------------------------------------------------
>> t+1
> line 0:3 mismatched input '+' expecting EOF
> 
> But if I replace the Lexer rule by Parser rule:
> 
> oper: ('+'|'-'|'*'|'/');
> 
> All work as it should. Why I should't use Lexer rule with alternatives ?

The problem with your example is that in the expr rule you use oper, but
define OPER as rule. Did you overlook this while posting or does even

expr	:
         operand (OPER operand)*
       	;

not work? Nonetheless, the usual way to define operands is to create a
single rule for each operand and reference those lexer rules in a single
parser rule for convenience. As "oper: ('+'|'-'|'*'|'/');" creates
implicit lexer rules doing it this way simulates the described approach,
but there can be problems as '+' used in to different rules will create
two distinct token types which will cause a MismatchedTokenTypeException
later.

Johannes
> 
> Thank you
> 
> List: http://www.antlr.org/mailman/listinfo/antlr-interest
> Unsubscribe: http://www.antlr.org/mailman/options/antlr-interest/your-email-address
>