The inconcsisteny of matlab contourgroups

[David Bateman]

I find matlab's behavior wrt contour groups bizarre and inconsistent and 
so am asking for help in how people think that this feature should be 
implemented. The issue is that at the moment matlab uses two different 
fashions to represent contours depending on what function created them. 
The first representation is a set of patch much as Octave currently 
uses, and was the only means of representing contours in matlab V6. This 
format is still used by all 3D contouring like contour3, meshc, surfc 
even in matlab 2008a. The second means is a "contourgroup" hggroup that 
encapsulates the patches in a single object with callbacks to alter them.

I see no reason not to have everything as a contourgroup, and the only 
change to a contourgroup to support 3D contouring would be an additional 
pair of properties like

addproperty ("zlevel", hg, "double", 0);
addproperty ("zlevelmode", hg, "radio", "{none}|level|manual", "none");

where if zlevelmode is "none" we have a 2D plot, if its "level" the 
contours are plotted as the Z value of the contour itself an if 
"manual" then "zlevel" can be used to force a particular contour Z value.

I hesitate to make a change like this as I imagine that the reason the 
3D contouring is not a contourgroup in matlab is purely that is was a 
dropped feature in the rush to a release and hasn't since been addressed 
by mathworks. So I fully believe that contour3, meshc and surfc will 
eventually also use contour groups instead of a set of patches in a 
future version of matlab. Therefore any changes we make here are 
certainly going to be matlab incompatible, and so that makes me hesitate 
to do this.

However, if we choose to keep the 3D contours in Octave to be compatible 
with current versions of matlab there is likely to be significant code 
duplication in any changes I make.

So, what do people think we should go for

1) Full matlab compatibility, accept the inconsistency of 2D vs 3D 
contouring and the code duplication in Octave, or

2) A matlab incompatible change that allows contourgroups to also work 
for 3D contours?

I have a preference for 2) as it'll probably be simpler in the end..

Cheers
David


--
David Bateman                                address@hidden
35 rue Gambetta                              +33 1 46 04 02 18 (Home)
92100 Boulogne-Billancourt FRANCE            +33 6 72 01 06 33 (Mob)