Re: [gforth] naive fibonacci

[Elko Tchernev]

Terrence Brannon wrote:
> I [found](http://bit.ly/ar7uDi) a couple of naive fibonacci programs
> by Bill Spight that work:
>
> : FIB ?DUP
>   CASE
>     0 of 1 endof
>     1 of 1- RECURSE 1 endof
>     1- RECURSE TUCK +
>   ENDCASE ;
>
> : FIB DUP 0= IF 1 ELSE 1- RECURSE TUCK + THEN ;
>
> But I would like to know why mine does not work:
>
>
> : fib { n -- fibn }
>   assert( n 0>= )
>   n CASE
>     0 OF 0 ENDOF
>     1 OF 1 ENDOF
>     2 OF 1 ENDOF        
>     ( otherwise ) n 1 - recurse n 2 - recurse +
>   ENDCASE ;

There is a non-explicit DROP before the ENDCASE. That's because the case 
parameter has to disappear at some point. Change your code in this way:

>     ( otherwise ) n 1 - recurse n 2 - recurse + SWAP

This way, the parameter will be dropped, not your computed Fib value, 
and it will work.

In general, CASE is not a very useful or used construct in Forth; this 
is not C. The less you use it, the better.