Re: proposed changes to legend()

[Rik]

On 07/06/2013 05:18 AM, Ben Abbott wrote:
> Rik,
>
> I thought I'd address this to you as the you and I have made most of recent 
> changes to legend(), and I'd like to hear from someone familiar with legend.m 
> before proceeding.
>
> Since Octave has no way to reliably determine the tightinset for gnuplot, I 
> am proposing implementing a different behavior for gnuplot and opengl.
>
> In the case of gnuplot, my thought is to let gnuplot handle the location of 
> the keybox and the plotbox (this requires that legend() set the 
> activepositionproperty of the plotbox to "outerposition" when the keybox is 
> outside the plotbox).
>
> For opengl, the goal is full Matlab compatibility.
>
> I have three changesets prepared (1) new demos which highlight the problems, 
> (2) the new gnuplot behavior.  Which implements a behavior close to what was 
> in the 3.2.x series. (3) fixes for opengl.  I've placed them on the patch 
> tracker, and will follow up with pdf's of all the demos for both gnuplot and 
> opengl shortly.
7/6/13

Ben,

I think what you are proposing is fine.  Supporting two graphics backends
is difficult for such a small group of volunteers, and we know that the
future for Octave lies with OpenGL and Qt.  Given that, implementing a few
kluges for gnuplot, which are eventually going to be ripped out, seems like
an expedient tactic.

--Rik