>>>>> "Jon" == Jon B <3ujg84402@sneakemail.com> writes:
>> Can you give an example of this?
Jon> plot2d((1/(f))/((1/(f))+1),[f,.0001,1000],[gnuplot_preamble,"set log
Jon> xy"]);
Jon> The line gets "thicker" as it moves to the right, the result of the
Jon> extra points "bunching up", and it looks like there's only two or
Jon> three samples for the entire left half.
>> I think I know why this is. plot2d doesn't really know about log
>> plots and assumes, more or less, uniform sampling of the function. So
>> there won't be as many points in 1-10 as there are in 10-100.
>>
>> I guess we would need to tell plot2d that we're doing a log plot and
>> have it sample logarithmically to get uniform sampling of the plot.
Jon> That's what I figured. Does anyone know how to get it to send
Jon> logarithmically spaced samples to gnuplot? Something involving
Jon> "transform_xy"?
That might work, but I don't know how to use that. I thought it was
for 3D plotting.
I have a little hack now that seems to work reasonably well for
adaptive 2D plotting.
It modifies (internally) the function to be plotted to be f(exp(x))
instead of f(x). The sampling is then done on log(xlow) to
log(xhigh), so we get uniformly spaced log samples.
At the end, the log samples are converted back before being sent to
the plotting program.
This seems to work, but, of course, the routines need to be informed
that log plotting is to be done.
I haven't done that yet. Some additional option needs to be
specified? Any suggestions for the name?
Also, since 5.9.2 is supposed to be soon, we'll probably have to wait
for this.
Ray