Re: graphics crossroads

[Søren Hauberg]

John W. Eaton skrev:
> | One possibility would be to remove plotting support from octave. Then 
> | have a package containing the current plotting functions, and another 
> | package containing the development graphics object based system. That 
> | way it would be easier to release 3.0 if the objects based system takes 
> | time to develop.
> | (I'm not sure if this is a good idea, but I thought I'd mention it...)
>
> For minimum surprise, I think we would have to go back a few weeks on
> the plotting functions.  But then we would not have a reliable legend
> command or plotting on top of images in 3.0.
That would really be a shame, but I see where you're coming from. Could 
we develop the objects based system without removing the current system? 
Either as a package or just with prefixed function names.

> | It's fine with me if the __gnuplot_raw__ calls fails. But if we really 
> | want people to stop using these commands it might be helpful to provide 
> | some functions to write your data to a file that gnuplot can read.
>
> In the current CVS, I have
>
>   FILE = __gnuplot_save_data__ (VAL, NDIM, PARAMETRIC)
Perhaps that function should be renamed to gnuplot_save_data and when 
people need more control of their plots then the highlevel system 
allows, we could tell them to use that function?

I'm not really good with graphics (and I don't speak gnuplot), but I 
guess I could lend a hand in the development of an objects based system.
So assuming all information about a plot has to be saved, what is all 
information?

 * The type of the plot (2d, 3d, mesh, surface, etc.)
 * The data (x, y (and possibly z) values)
 * Style information (color, line thickness, etc.)
 * Figure number?
 * Some subplot things?

Also, what would be the target version of gnuplot? Can we assume 4.2? 
Should we require 4.0? How about 3.2? (you get the idea...)

Søren