Rafael Garcia-Suarez writes:
> Luke Palmer wrote:
> > Also, if this is going to be an explanation rather than just a picture,
> > I suggest you go with Perl's usual versatile power, and store the
> > operators in a declarative data source.
> >
> > grammar RPN {
> > my @operator = << + - * / >>;
> >
> > rule input { <line>* }
> > rule line { <exp>? \n }
> > rule exp :w{ <NUM> | <@operator> }
> >
> > rule NUM { \d+ }
>
> This becomes too weak.
>
> rule exp :w{ <NUM> | <exp> <exp> <@binary_operator> }
>
> is stronger.
> But it involves left-recursion, so it's typically better adapted to
> yacc-like parser generators. Probably not to P6 rules...
Whoops, yeah, that's what I meant to write. That is, after I realized
we were working with an RPN caluclator and not an infix one. And even
then my brain was broken.
That left recursion won't do. I can't remember my transformation rules
well enough to know how to put that in a form suitable for a recursive
descent parser. To be honest, I've never seen an RPN calculator modeled
with a grammar.
To be sure, a prefix calculator should be the easiest:
rule exp :w { <@operator> <exp> <exp> | <NUM> }
Luke
