Re: [Aldor-l] [Axiom-developer] "has" and "with" (was curious algebrafailure)

[Ralf Hemmecke]

On 08/14/2007 11:10 PM, Bill Page wrote:
> On 8/14/07, William Sit wrote:
> > ...
> > For me, translating Spad code accepted by the Axiom interpreter into Spad code
> > accepted by the compiler is a pain. Either require explicit coercions and calls 
> > in
> > both the compiler and interpreter, or provide the SAME verifiable/selectable
> > assistance in both cases.

William, you should really try to experience how it is to program in 
Aldor. Sure you will feel some pain in the beginning, but that will not 
last long. You will love the clarity.

> I am inclined to agree. To me this means both, making coercions is
> Spad more explicit (Like Aldor with only a few well know and
> frequently used automatic coercions), and creating a version of the
> Axiom interpreter that is equally strict. That said, of course a
> strick interpreter might be viewed by some as more painful for the
> user, so then there would be even more reason to create an alternate
> single-typed (not type-free!) user interface language like BNatural.

Right, that would be my path, too. Interpreter and compiler language are 
identical and identically strict. No auto-coercions.

But what is now called "Axiom Interpreter" is not that what I meant 
here. I rather used "interpreter" in the sense of PASCAL interpreter vs. 
PASCAL compiler.

The Axiom Interpreter should be replaced by something like BNatural. But 
here again, it will happen that programs that are written using 
BNatural, cannot easily be translated into pure library code suitable 
for the compiler, since BNatural would be designed to do a lot of 
guessing in order to assist the user. If in BNatural you use something 
like "Expression Integer" you would not want to use that for an actual 
library code. Library code should be clean and make the experiments and 
experiences that are gained with BNatural into clear algorithms over 
well defined domains.

Note that when you switch from "experiment" (=BNatural framework) to 
writing it down in a library, you have to come up with a good 
documentation anyway. Of course, I would not like to see a documentation 
of *your experimental code*, I would only accept if you tell me the 
theorems, definitions, motivations, etc. And if you do that, sure you 
rewrite your original BNatural experiment. You cannot expect that it 
will be painless to implement library code from fuzzy experimental code. 
(Experimental) Code that "just works" is not enough.

Ralf