Re: pcolor bugs

[Ben Abbott]

On Jan 31, 2012, at 4:16 PM, Ismael Núñez-Riboni wrote:

> Thanks for the answer. None of the bug reports seems to describe what I found 
> with pcolor...
> 
>> If you're not sure it is a bug, you can describe what you found here.
> 
> 1) NaNs are displayed as the lowest value of the color scale and not as 
> blanks.
> 
> 2) Pcolors with "shading faceted" or "shading flat" shadow anything else in 
> the figure (like plot(x,y) or contours), i.e., only the pcolor is visible and 
> everything else in the figure seems to be "behind" the pcolor. It does not 
> happen with "shading interp" (here you can combine pcolors and other plots in 
> the same figure).
> 
> 3) I have to check this one deeper, but I think I get pcolor cells which seem 
> to be above the maximum color code (they look red in the pcolor), even though 
> all values are inside the color range. This seems to happen only in 
> combination with "clf" and "hold on" (i.e., when I make a video of pcolors in 
> one figure, cleared each time with clf).
> 
> Do these features sound familiar? They definetely look like bugs to me, 
> unless I'm misunderstanding something.
> 
> Cheers, Ismael.

They do sound like bugs, but there isn't a lot that can be done when using 
gnuplot for the backend.

Octave aspires for Matlab compatibility, but gnuplot does not.

#2 is likely due to gnuplot's approach to faking hidden line removal. The 
result isn't so much a bug, but a limitation.

If you are using Octave 3.4.x, you might try the experimental OpenGL backend. 
You can do that by ....

        close all
        graphics_toolkit fltk
        pcolor (peaks)

In any event, can you provide a simple example that we can reproduce ?

Ben