[om-list] Re: Language Power

[Mark Butler]

Long, long ago, LISP had a normal readable syntax, but apparently the
inventors decided to change to emphasize that LISP code was just another form
of LISP data.  Unfortunately this decision was fatal for its prospects of
general acceptance.

I believe a reasonable compromise would be to reverse course and develop a
readable high level language with a similar internal representation as LISP.  
You could then support run time code generation in two forms: by using code
that generates LISP style operation trees directly or by using code that
generated high level language that was translated to operation trees as a
separate step.

This capability is particular relevant in our system for modeling a wide
variety of natural phenomena.  Any attribute that is distributed across space
and time can only be efficiently represented and operated on using functional
representations, even if only to interpolate between various samples.  

The only way to actually do this in languages like C++ or Java is to develop a
LISP-like symbolic math engine.  For serious applications, it would also be
very helpful to be able to compile a mathematical expression into native
machine code.  The current way to do this is to translate it into regular
source code, write it to a file, use a compiler to compile it, and then load
it back as a shared library or Java class file.  

 - Mark