Robert,
thanks for your work.
I simplified the testcase by ignoring the charge (charge(x):=0$)
and then it does run with the current maxima.
I guess the glue&smoothMin&smoothMax used inside the charge function are creating all the difficulties.
Peter
--- On Mon, 12/17/12, Robert Dodier <robert.dodier at gmail.com> wrote:
> From: Robert Dodier <robert.dodier at gmail.com>
> Subject: Re: taylor series
> To: "Maxima Mailing List" <maxima at math.utexas.edu>
> Cc: "Peter Foelsche" <foelsche at sbcglobal.net>
> Date: Monday, December 17, 2012, 9:58 AM
> OK, I got some results, although I'm
> not sure if they're correct: all of
> the 2nd order partial derivatives are zero or very close to
> zero. (Same
> with the 1st order partials, but that's plausibly OK -- e.g.
> we are
> looking at a local minimum of some function. Just guessing
> of course.)
>
> I reorganized the computations to construct a formal Taylor
> series and
> work on that before substituting in the actual definitions.
> Note that
> some of the intermediate results are still very large --
> circa 50,000
> terms. Dunno if that's to be expected or strange. It may
> well be that
> there is still another, better way to go about it.
>
> Released versions of Maxima cannot run the attached script
> -- there is a
> bug in PARTITION which is tickled by expressions containing
> dummy
> variables ("at" in this case). I fixed the bug, patch
> attached for reference
> (I pushed the commit already). You'll have to recompile in
> order to run
> the script. Or take a different approach which doesn't
> trigger the bug.
>
> Well, I don't know if I've accomplished anything but anyway
> here it is.
> It was, as they say, a learning experience.
>
> best
>
> Robert Dodier
>