[Aldor-l] Interpreter bug

Martin Rubey martin.rubey at univie.ac.at
Tue Oct 31 02:20:25 EST 2006


Dear Ralf,

Ralf Hemmecke <ralf at hemmecke.de> writes:

> Hi Martin,
> 
> I've just found
> 
> grammar(p: List String): List Type == {
>          import from MachineInteger;
>          A: CombinatorialSpecies == EmptySetSpecies;
>          res: List Type := [A pretend Type for x in p];
>          E(i: MachineInteger): CombinatorialSpecies ==
>            evaluate(parse(p.i), res) add;
> 
>          for i in 1..#p repeat res.i := E(i);
>          res;
> }
> 
> in your (workaround) suggestion for the signature-experiments branch.
> 
> Before I allow that to be committed to the trunk of Aldor-Combinat I would
> rather stop using Aldor.

So, why are you using Aldor? Is it really because of type safety? Really, I
don't care much about type safety, what's important to me is expressiveness. I
doubt that something as nice as the species project (= Aldor Combinat) can be
done easily in another language. I'd think that in Lisp it would be easy at
first, but you loose all the maths we have because of libaxiom and libalgebra.

Type safety is something nice, but I really want to have the possibility to
tell the compiler: trust me, what I'm doing is ok, even if you think otherwise.

> I don't want to see such ugly
> 
>    res: List Type := [A pretend Type for x in p];
> 
> where
> 
>    res: List CombinatorialSpecies := [A for x in p];
> 
> should do.
> 
> THE COMPILER NEEDS TO BE FIXED NOT OUR CODE.

Yes, but I cannot fix the compiler. Distribute a patch to libaldor that does
the trick, maybe that is possible. Trust me, I don't like it either. And I'm
investing a fair bit of time into finding a conforming replacement for the
grammar code. (It is quite certainly non-conforming, currently.) Maybe
Christian has one, that would be really cool.


I strongly doubt that Aldor will become free within a time frame that's useful
for me. Note that very likely I will have to quit Academia with december 2008,
I won't hack on Axiom nor Aldor then anymore.

I don't see a better language for the species project. Do you?

Martin

> >> you should be more minimalistic. ;-)
> >>
> >> aldor -gloop
> >> (press Ctrl-C)
> >>
> >> That one is known since long. :-(

you mean that the message is about AxiomXL? I don't care about such bugs, in
fact.

Martin




More information about the Aldor-l mailing list