[Rd] Reproducible Rterm crash. (PR#7072)

[Deepayan Sarkar]

On Friday 09 July 2004 22:14, Richard M. Heiberger wrote:
> Deepayan Sarkar:
> > This will give you non-empty panels, but not different angles of
> > rotation. This is because panel.wireframe doesn't handle screen,
> > the top-level function (wireframe) does. There's no good reason for
> > this behavior, so I'll fix it. But since this would be an API
> > change, I don't think I should do it in a stable update. I can send
> > you a patched version privately if you wish.
>
> I named the arguments as you suggested and now the panels are filled.
>  As you noted, they all show the same rotation.
>
> Please send me the revised wireframe and panel.wireframe with the
> rotation moved down into the panel.  Thank you.

An updated source package is up at

http://www.stat.wisc.edu/~deepayan/R/lattice-versions/lattice_0.9-16.001.tar.gz

> Can you clarify the statement "stable update".  I think this means (I
> hope that is what it means) that you will make the change in 1.9.2,
> but not in any intermediate release.

By stable, I mean the R 1.9.x series. I don't know if there will be a 
1.9.2, but even if there is, I think it would be inappropriate to make 
changes that could potentially break currently working code. While 
that's extremely unlikely to happen with these changes, it's possible, 
and I can't really assume that no one would be affected. For example, 
it would be possible to get different rotations in different panels 
using the current lattice (in a non-S-compatible way), by using the 
'rot.mat' argument in panel.wireframe instead of 'screen', and that 
code would no longer work after the changes. 

On the other hand, such changes would be acceptable for a major upgrade. 
The next one's going to be 2.0.0, sometime around October. Till then, 
you can get the updated version of lattice with the pre-release 
unstable builds, but not with the 'released' 1.9.x versions.

Deepayan