Re: [Axiom-developer] Aldor / Axiom

[Martin Rubey]

C Y writes:

 > I should note that I don't use Aldor, just basic Axiom - I tend to stick to
 > open source stuff, so Aldor specific syntax won't be terribly useful if it's
 > too different from Axiom's.  Is Aldor's syntax a subset of Axiom's?

Aldor and Spad are quite similar. Ralf already pointed out one of the
differences. In Aldor you may choose between "pile" syntax, which is the syntax
used by Spad, i.e., blocks are defined by indentation and C like syntax, where
blocks are defined by pairs of braces. In Spad you have to use "pile" syntax.

You can already write your stuff in Aldor syntax and use it for Axiom. In this
case your file must have the extension .as and you have to have Aldor
installed. Furthermore, you probably need to follow the instructions from the
last comment (by myself) in

http://page.axiom-developer.org/zope/mathaction/37CoAldorFileDoesNotWork

However, Peter Broadbery seems to have updated his patches and distributed them
this week on the list. I did not try them yet.

One big difference between Aldor and Spad is, that Spad only implements a
subset of Aldors grammar. In my opinion, the two most important points are

* dependent types in full generality, i.e., in Axiom the following signature is
  currently illegal:

  test(n: Integer) -> IntegerMod(n)

  In some cases, this is a show stopper, I'd say. Although William Sit has
  argued that it can be emulated in Axiom, but I'm not really convinced. If you
  look for Marcus Better on the mailing list axiom-math and axiom-developer,
  you will find that his question was in the end not answered. (And I still
  don't know how to do it...)

* the extend keyword is missing. This is explained in detail by the Aldor
  documentation.

Another big difference between Aldor and Axiom are the libraries. Very
unfortunately, the libraries are quite incompatible, as far as I know. So, if
you want to use your program in Axiom, you will need to use the Axiom library,
which is no problem, fortunately.

My sincere hope is that we will somehow be able to make the libraries
converge. This probably means to spend a lot of time thinking about a good
design. I'm quite sure that the Aldor library are better designed, but on the
other hand, a lot of things are missing. So I think we have to look which
domains from Aldor can be integrated in the Axiom type hierarchy. A big
project, which is probably feasible, are the polynomial categories and domains.

However, before this can be done, there is a problem with licensing. We would
need to convince Stephen Watt to free Aldor. I know from private Communication
that Manuel Bronstein was very much in favour of this step, but Stephen Watt
has not been very responsive yet...

If -- alas -- it is not possible to free Aldor, I think the best thing to do
would be to hire a lisp guru who implements the Aldor extension for us. I don't
think that this would be too difficult, and in fact, it would make sense in any
case, whether Aldor becomes free or not.

I suggest that we write a signed petition to Stephen Watt when we come to it.

Martin