Re: [manual] suggested addition to pcolor description

[Ben A. <bpa...@ma...>]

On Feb 4, 2009, at 4:03 AM, Petr Mikulik wrote:

>>> Yes, please try to patch both gnuplot and Octave and tell me  
>>> whether it
>>> gives the desired image.
>>>
>>> Note that
>>>  set pm3d interpolate 0,0
>>> can always be issued by Octave because current versions of gnuplot
>>> silently ignore non-positive numbers.
>>
>> Ok, I've made a local change to __go_draw_axes__.m and produce some  
>> images
>> with both Octave+Gnuplot as well as Matlab.
>>
>> Those interested can see them at the link below.
>>
>> 	http://www.picturehosting.com/gallery.php?u=bpabbott&g=pm3d
>>
>> Those who would like to exercise this functionality will need to  
>> apply Petr's
>> patch to the developer's sources for gnuplot, and apply the  
>> attached changeset
>> to the developers sources for Octave.
>
> The patch should be:
>    if (doing_interp_color)
>      interp_str = "interpolate 0,0";
>    else
>
> If you find this resolution being (very) different from Matlab, then  
> use
>      interp_str = "interpolate -300,-300";
> or some other negative numbers. Currently the default "0,0" is  
> equivalent to
> "-200,-200".

opps ... I should have been more thorough with my testing. I've  
updated the figures at the link to include those produced with (0,0)

	http://www.picturehosting.com/gallery.php?u=bpabbott&g=pm3d

>> 	surf(peaks)
>> 	colormap(jet(200))
>> 	shading interp
>>
>> After  an unpleasant delay a nice image was produced. Checking my  
>> Mac OSX
>> activity monitor it appears that gnuplot requires 750MB of real  
>> memory and
>> 2.5GB of virtual memory to produce this image. It is due to the  
>> excessive
>> time and memory requiried that I decided to limit the levels of  
>> interpolation
>> that occur across a particular quadrangle.
>
> I don't wonder if the current changeset produced something like
> 	interpolate 90,90
> i.e. drawing image with (200*90)^2 = 18000^2 points.

Using "interpolate 0,0" the result renders very rapidly.

Petr, Thanks for pointing out my err, the result looks very good to  
me. The corrected changeset is attached ... attributed to you of  
course ;-)

I don't have a copy of an old gnuplot available. Petr are your running  
Octave built from the developer's sources? If so, can you verify that  
this change does not break anything?

Ben