Re: colorbars

[Daniel Oberhoff]

Ah, ok. Got it. And I see the problem now: Imagesc just rescales the  
data to [0:255] and hands it to image, so the colorbar will be in the  
scaled coordinates of course. Same goes for pcolor from octave-forge,  
which seems to just use image. This is different from matlab. I  
suppose it would need a different kind of object for (nonrgb) image  
data in a different range? As for the colorbar: since graphics are  
object oriented I suppose there would need to be a state flag for the  
colorbar in the object tree?

Ok, looking at how Matlab does it. It seems to be clever about the  
data type first up (i.e. range [0 1] for real data and [0 255] for  
integer data). Hmm, actually looking at the object inside Matlab  
reveals that the actual data stored with the object is unscaled using  
imagesc. Also the colorbar is a separate object. But that souldnt be  
very important.

Hmm, interesting, it just works by handing the data to image instead  
of imagesc :). That's probably the best solution. Since fully  
implementing imagesc the matlab way would require settable color  
range to image. Remains the colorbar issue...additional flag to the  
object tree?

Daniel

Am 09.08.2007 um 21:42 schrieb John W. Eaton:

> On  9-Aug-2007, Daniel Oberhoff wrote:
>
> | Regarding the colorbars in gnuplot: I have figured out the commands
> | to do image plots in gnuplot with properly scaled colorbar (or
> | without one). How would I go about hacking octave to use them? All I
> | could find regarding gnuplot was __gnuplot_raw__.cc, and that seems
> | to be flex generated. Would I need to learn flex first?
>
> The current graphics code for handling properties is in
> graphics.{h,cc} and the .m files in the scripts/plot subdirectory of
> the sources.
>
> jwe