Testing a numerical implementation I want to access data, which are
suggested as maximal errors through plotting (with care) to examine
that in more detail (so plotting is considered just a help)
For univariate functions I am aware how to look into data pairs for
the command plot.
My function is bivariate and real valued.
P:= plot3d( f(x,y), ...) lets me save the result and
op(P);
op(1,P);
arr:=op(3, %);
arr; Arr:=convert(%, Matrix);
plots[matrixplot](Arr, axes=boxed);
let me even see and guess: but that are only the coordinates for the
result in terms of 'column and row' standing for 'x and y'. And the
actual values are foregotten (of course, actually it is 2 D).
How do I find (or at least guess) the underlying numerical values
for x and y?
Is it equidistant (seems so in my case), would I have to take the
single precision equivalents (guessing it is a DLL, single precison
and working with _classical_ interface)?
Or is there some way to catch the (x,y) input for plotting f even
before (which I would prefer of course)?
