> This is also a bug to be fixed in 'gpdraw' (ok, ok, let's call it 'draw'
> in the near future :), since it also calls 'coerce-float-fun'.
>
> When the plotting program finds a pole in function f(x), say in x0, my
> present idea is to re-evaluate f in some point in the neighborhood of
> x0, somewhere in (x0-eps/k,x0+eps/k)-{x0}, for some k>1. Andrej, in
> private e-mail (I must say he is an excellent bug-hunter), suggested to
> compute some average.
>
> In explicit functions in 2d, we could also remove this conflictive
> point, but in 3d the problem is harder, since points must be sent to
> gnuplot in groups of equal size separated by blank lines.
>
> Any other ideas?
>
Ideally, the algorithm will attempt to evaluate the limit as x -> x0.
(In some numerical fashion, of course. I am no numerical expert, so exactly
how this should be done is not clear to me.)
If it exists, as in the case of sin(x)/x, then the limit (or the approximation
to it) will be taken as the value of f(x0). Otherwise the plot will diverge to
infinity around x0.
In saying all this, I am taking my cues from the way Mathematica does these
things.
Kostas