gnuplot-py-users Mailing List for Gnuplot.py

Michael Haggerty wrote:
> John Parejko wrote:

>>Along those lines: it appears noone else took up the offer to become the
>>Gnuplot.py maintainer.  I sent a private message to Michael expressing
>>my interest (and noting that it will be a bit before I can really take
>>over), but I guess I'll post my interest here.  I'm using Gnuplot pretty
>>heavily, and would like its development to continue (and get it ported
>>to numpy).

It's probably worth noting that the numpy port should be nearly trivial, given 
that Gnuplot.py doesn't really make extensive use of any sophisticated/weird 
features of Numeric.

On another note, if you (John) continue to maintain Gnuplot.py, you may be 
interested in merging some of the code in IPython.Gnuplot2.py:

http://projects.scipy.org/ipython/ipython/browser/ipython/trunk/IPython/Gnuplot2.py

I think some of the enhancements there are worthwhile, and sent them to 
Michael a while ago but I know his priorities have moved from Gnuplot.py 
development.  I also have stopped using this code at all (I'm using matplotlib 
these days), but I'd be happy to see this code continue to be useful to 
gnuplot users.

The following utility module may also be of use, that's the one I actually 
imported in all my top-level code:

http://projects.scipy.org/ipython/ipython/browser/ipython/trunk/IPython/GnuplotRuntime.py

Cheers,

f

John Parejko wrote:
> float_array(m) will downcast sometimes.  I'm not sure of the exact
> reasons, but I had points disappearing from my plots that were ~> 3e38,
> which is the single-precision float size.  tuples were going in that
> were correct, but whatever was coming out wasn't...

That is very likely.  I think I reasoned that floating point precision
was enough for creating plots that will be plotted at most at a few
hundred dpi.  But of course floating point has limited exponent range,
too, and you are apparently reaching that limit.

IIRC, you can pass this routine Numeric arrays of higher-than-float
precision, in which case they will not be downcasted.

But also be sure you have checked that gnuplot itself is not truncating
your data.  I don't know if it uses float or double internally.  gnuplot
also has some setting for "minimum number not considered to be zero"
that is sometimes confusing (though not in your case).

It shouldn't be too big a deal to change the default Gnuplot.py casting
to "double" for your plots.  (Whether that should be a new universal
default could also be discussed.)

> Along those lines: it appears noone else took up the offer to become the
> Gnuplot.py maintainer.  I sent a private message to Michael expressing
> my interest (and noting that it will be a bit before I can really take
> over), but I guess I'll post my interest here.  I'm using Gnuplot pretty
> heavily, and would like its development to continue (and get it ported
> to numpy).
> 
> Was there any interest on the scipy mailing lists?

I didn't get around to asking because (a) I don't have any real
connection to that community and (b) I don't have time to look into it.
 This would be an excellent task for a volunteer :-)

Good luck with your orals!

Michael