Re: defining functions in basin scripts

> Or, as a very bad possibility, we could define ALL functions as a taylor
> series expansion.

If you look in the <math> source, you will see that many standard
functions are basically Taylor expansions (but not the ones you'd write
down first). However, I don't think we should start re-implementing the
math libraries. If we have a parser that can identify standard
functions, then there is no reason not to use the built-ins, is there?

> Bruce, this is clearly a solved problem with Maple, so what would you
> suggest? I think that we will probably limit the complexity of the
> functions to combinations of polynomials and trigonometric functions.

Agreed. Let's not get carried away.

> Anything else (e.g. Fourier convolutions, integrals, etc.) can be done
> with a series of loops in python or direct calls to the FFT library.

Yup.

Of course, we could also compile arbitrary functions on the fly into a
run-time module library, and then reference them by name. NEMO does
that, for example, although our applications are a bit more complex than
those in NEMO.

Steve
Received on Fri Aug 19 14:06:52 2005