Re: [Axiom-developer] Axiom meeting at ISSAC

[Stephen Wilson]

Didier,

The idea of improving the current Axiom compiler is itself not new
(Aldor is an example), but this particular effort is only a few months
old.  Consequently, it is not well developed at this point.  In
concrete terms, we have a new parser which can handle a good subset
(say 90%) of the current SPAD language. Current activity is centered
around developing new resources which need to exist before further
work continues -- literate programming tools, a build environment
which integrates seamlessly with all components, and a port of Axiom
as a whole to run on an ANSI common lisp (gcl-2.7.0).  This
`foundational' setup is almost complete.

This effort does not have the same goals as Aldor, nor does it aim to
be compatible with Aldor.  However, a lot of the lessons and insights
which users of Aldor can bring to this implementation will certainly
guide the effort.  Many people want dependent types, for example.  We
will support them, but it will likely be in a non-compatible way.  

For example, a type which depends on a runtime expression like "1 + n"
might not be allowed, since deciding equality over syntax trees does
not fit well in an imperative language.  On the other hand, such types
could rely on runtime support to enforce correctness.  This is an
example of how the semantics of the new SPAD are not tied to Aldor's
rules, and how we are still in the initial stages of designing the
system.

Another example is Aldors generators.  It is planned that these will
be supported in the new SPAD as a special case of continuations.
Thus, generators in SPAD will become an algebra item, not a builtin.


There is also the question of being able to provide the compiler
components in the form of a library which can be used by other aspects
Axiom.  For example, one would be able to use the new parser and
(forthcoming) type checker to support an IDE.  One of the main goals
aside from providing a better compiler is to provide a set of tools
which can be used to enrich the Axiom system as a whole.  The main
point here is one of design.  The new system will be flexible and
easily `plugged into' at all levels.


There is a wiki page which attempts to give some basic links for those
who wish to develop the system:

     http://wiki.axiom-developer.org/AxispBranch

The current work is only suitable for developers but there will
certainly be `public' releases made incrementally to keep `on-track'
for an eventual merge with the canonical Axiom sources.


Take care,
Steve