[Axiom-developer] Re: [Aldor-l] exports and constants

[Christian Aistleitner]

Hello,


> > > You can interpret both kinds as "provides at least".
>
> No, I don't think that is true.

Can you provide an example where you cannot interprete it as "provides at  
least"?

> > >> Take for example
> > >>
> > >> A: Join(CatA, CatX) == add {...}
> > >> B: Join(CatB, CatX) == add {...}
> > >> X: CatX == if odd? random() then A else B;
>
> The reason of course is because the expression 'odd? random()'
> is not static - it does not have a well-defined value at the
> time that the source is compiled. It is only later when the
> code is run that we can know if the condition is true or false
> in a particular instance. The Aldor compiler is not able to
> compile *types* of this kind.

I see, that the compiler is not able to compile this. We all agree on  
that. But I cannot see mhy the "odd? random()" is the problem. From my  
point of view, the "if" is the problem. Ralf already provided an example,  
where

  X: CatX == if true then (A add) else (B add);

worked, while

  X: CatX == if true then A else B;

did not. In this example "true" takes the place of "odd? random()". "true"  
is static as could be and still the problem remains. Therefore I doubt  
that "odd? random()" is the problem. Type inference across the "if"  
statement is the problem. As I showed in
http://www.aldor.org/pipermail/aldor-l/2006-July/000228.html
, hinting the comiler by "@" works perfectly well.

--
Kind regards,
Christian