[antlr-interest] Which one is quicker?

[Johannes Luber]

chain one schrieb:
> 1.
> INT
> :   '-'? DIGIT+                  
> ;
> FLOAT
> 
> :   ('-'|'+')? DIGIT+  '.' DIGIT* (('e' | 'E') ('+' | '-')? (DIGIT)+)?
>      ;
> 
> 2.
> INT_OR_FLOAT
>   : '-'? DIGIT+ FLOAT?
>  ;
> 
> fragment
> FLOAT
> 
> :   '.' DIGIT* EXPONENT?
>      ;
>     
> fragment
> EXPONENT: ('e' | 'E') ('+' | '-')? (DIGIT)+;
> 
> 
> Which one is quicker?
> I think the 2th is quicker. Because in the 1st lex rule, A DFA is needed
> to predict if it is a INT or FLOAT.
> Am I right?
> 
> 
> Thanks
> Best Regards,
> Young

Did you benchmark the code? I've heard that people don't see the true
bottlenecks just by looking at code alone. Run the recognition a million
times and you should see a difference.

Johannes