Re: image limits

[Daniel J Sebald]

John W. Eaton wrote:
> On  8-Feb-2007, Daniel J Sebald wrote:
>
> | Here's something for you to think about.  I don't think there is any
> | rush to resolve this, as gnuplot may tweak behavior on images
> | slightly (hence labeled experimental in gnuplot's configuration).
> | But, if one tries (in CVS) image, the outermost pixels are
> | fractionally displayed by the automatic choice of limits.  To
> | exaggerate, try
> | 
> | image([4 5 6; 7 8 9; 0 1 2])
> | 
> | The array is 3x3 and the limits are chosen as [1:3], which only
> | covers the length of two pixels.
> | 
> | Like I say, gnuplot may be adjusted if there is a discussion about
> | any accepted conventions in the graphics community (i.e., start
> | image at zero, etc.).
>
> Try this with the current CVS Octave:
>
>   image
>   axis ([0, 10, 0, 20])
>
> I think you'll see that the axis limits change, but the full image
> remains displayed.  In Matlab, the axis limits correspond to pixel
> indices, so changing the axis limits zooms in on the image.  We should
> probably do the same.  Are we just not using gnuplot correctly here,
> or is this a feature of gnuplot that the axis limits are independent
> of the size of the image?

Not using correctly.  This is part of the patch I'm creating.  Recall I said 
that limits should be kept as part of the image as well.  In 
__uiobject_draw_axes__ the image routine uses the original limits, not the axis 
limits.  Should be taken care of.

Dan