On 5/19/07, Kostas Oikonomou <ko at research.att.com> wrote:
> Thanks very much for the fix. It works very nicely!
Terrific, I'm glad to hear that.
> Are these messages suggesting that some improvements can be
> made to the code?
Well, the bulk of the lbfgs code is an automatic translation from
Fortran. So I am hesitant to silence the warnings by modifying
the translated code. Probably the right thing to do is to modify
the translator program (f2cl) instead. I am too lazy to do that,
but if you want to work on it, go right ahead.
> 2. Can we have augmented_lagrangian load lbfgs, since it
> can't be used without it?
I guess so -- see below. But better still would be a mechanism
to load a file only if it hasn't been loaded. E.g. require(lbfgs);
or import(lbfgs); or something like that.
> I tried adding a simple-minded
>
> if not member($lbfgs, map(op,functions)) then
> load(lbfgs);
>
> but the "$" produces a parsing error.
Oh -- $lbfgs in Lisp is just lbfgs in Maxima (no $ prefix).
Consider this slightly more general formulation:
/* fboundp detects various kinds of Maxima functions */
fboundp (a) := ?fboundp (a) # false or ?mget (a, ?mexpr) # false or
?get (a, ?mfexpr\*) # false or ?mget (a, ?mmacro) # false;
if not fboundp (lbfgs) then load (lbfgs);
HTH
Robert