Re: [Texmacs-dev] TeXmacs interface to R 2.0.0 is broken

[Joris van der Hoeven]

Hi Ed,

> Yeah, I found that. I think what needs to be done is this:
>
> R has a separate package management interface. So TeXmacs needs to carry the R
> plugin as a "standard" R source package tarball. During the configure step,
> TeXmacs should check for the existence of R. There are two possibilities:
>
> 1. R is present. In that case, the TeXmacs install should execute the *R 
> package
> installation* on the source package. It should leave the R package behind in a
> documented place. Users with an older version of R who upgrade to R 2.0.0 will
> need to rebuild the package. That's what happened to the original poster and 
> to
> me.
>
> 2. R is absent. In that case, the TeXmacs install should leave the R package 
> on
> the system in source tarball form and give the user instructions for 
> installing
> it when they install R.

This is typically what we do *not* want to do. For optimal user confort,
our policy is that a plug-in should show up when it is present and
not show up if it is not installed. Moreover, many people do not compile
TeXmacs from source and still might want to use R.

Whenever TeXmacs detects R, then it should either tell R at runtime
where to search for additional (TeXmacs-related) packages, or perform
some on-the-fly installation of a tarball for R (if that is really what
the new version of R requires).

> Since it doesn't seem to be a very volatile piece of code, you might want to
> contribute the R TeXmacs interface library to the CRAN repository, so someone
> with R and TeXmacs can get the package that way.

Of course, another possibility would be to ship the plug-in along with R.
This might actually be reasonable, because it would increase the concern
of the R developers about this plug-in.

Best wishes, Joris