Plot tactics generally. was Re: [Maxima] pathological plot

Rich Drewes wrote:

> On Fri, 6 Sep 2002 willisb@unk.edu wrote:
>
> 
> Perhaps there should at least be an option to pass the function directly
> to gnuplot rather than attempting to compute a table of values to be
> passed to it?  This may sacrifice some flexibility in specifying 
> plotting options.  Just a thought.
> 

This means that you can only plot expressions that explicitly use
functions in the gnuplot vocabulary.  Anything that you have defined
separately as a maxima function is not available. A better solution
would be a plotting program that could be handed a lisp "call-back"
stub and would repeatedly call into the lisp for function evaluation
at an (x,y) point.  This would also be helpful for external
numerical integration programs that use (only) function evaluation.

Another alternative would be for Maxima to try to make a stand-alone
program and compile that.  Typically generating a little C-function
and compiling it.  This also forbids use of the "full power" of
Maxima.

Mathematica tried to do something via its "Compile" facility but
at least earlier versions tended to make blunders. For example, it only
did real floating point arithmetic.  But the function you are plotting
might have intermediate complex expressions.  This caused it to fail.
They have fixed this particular flaw, but there are other cases in
which the Mathematica semantics go one way in ordinary discourse, and
another way if the same expression is plotted.  Needless to say,
this can be troublesome.  The Maxima translation idea is to be
faithful to maxima semantics, however subject to the type declarations
that you might include.



> Rich
> 
> _______________________________________________
> Maxima mailing list
> Maxima@www.math.utexas.edu
> http://www.math.utexas.edu/mailman/listinfo/maxima
>