Re: Computer Algebra Algorithms lisp vs. C.

bernard parisse wrote:
> Printing does not require too much time, but parsing does.

That depends entirely on what you're parsing, of course. If you use a simple
format then parsing will take very little time.

> There is an interpreter in giac, mainly for interactive
> user session, each function on the user side calls a
> C++ function of the library that a C++ programmer can call directly.
> That's why I said there is no kernel/user language approach,
> since CAS like maple or mupad have many built-in functions
> programmed in the user language (and kernel function access is not
> provided, at least in the standard distributions).

Can I just clarify: Is it or is it not possible to write arbitrary programs
in giac's language?

> From my experience, it is not much harder to implement CAS algorithms
> directly in C++ than in say maple,

If you could design and implement your own language, specifically created to
let you describe the specialised set of algorithms that you're
implementing, do you not think it would be much better for representing
those algorithms than C++?

> and I find debugging much easier.

I can understand that, although I'd say that code in C++ is more error-prone
than in Mathematica (and, I assume, Maple). OCaml is much more robust than
both C++ and Mathematica and it also has better debugging facilities.

> And the resulting function is sometimes significantly faster.

If you don't have control over the language, then yes. But if you're also
developer the interpreter then you can simply improve its optimiser. If you
do this, you're likely to end up with faster code overall.

Dr Jon D Harrop, Flying Frog Consultancy