[Rd] Best way for rgl's .onLoad to fail?

[Duncan Murdoch]

On 01/02/2016 4:26 AM, Martin Maechler wrote:
>>>>>> "BH" == Bryan Hanson <hanson at depauw.edu>
>>>>>>      on Sun, 31 Jan 2016 09:50:46 -0500 writes:
>
>      BH> I think the 2nd option will be more palatable to
>      BH> inexperienced users, but both do state the important
>      BH> detail.  Bryan
>
>      >> On Jan 30, 2016, at 4:11 PM, Duncan Murdoch <murdoch.duncan at gmail.com> wrote:
>      >>
>      >> On OSX and Linux, the rgl package currently requires X11
>      >> libs to be available for linking.  Recent versions of OSX
>      >> don't include them by default, so I'd like rgl to fail
>      >> nicely.
>      >>
>      >> Ideally, it will load a library that doesn't need to link
>      >> to the X11 libs but will still allow WebGL code to work,
>      >> but that's complicated, so I'd like a stopgap.
>      >>
>      >> I can detect that the failure is about to happen, and
>      >> call stop() in the .onLoad hook, but that gives an ugly
>      >> message:
>      >>
>      >> > library(rgl)
>      >> Error : .onLoad failed in loadNamespace() for 'rgl', details:
>      >> call: NULL
>      >> error: X11 not found; XQuartz (from www.xquartz.org) is required to run rgl.
>      >> Error: package or namespace load failed for ‘rgl’
>
> I agree that the error message is a bit messy or even ugly,
> however, other than Bryan, I would want  library(.)  to signal
> an error when it cannot provide a working package, loaded and
> attached to search().
>
> Other functions, such as  require(.)  do rely on this behavior of library(.),
> e.g., the much used idiom
>
>    if(require(<some package>)) {
>        ....
>        ....
>        ....
>    }
>
> needs library() to signal an error on  non-success.

Yes, that's a good point.  That's what the version on R-forge currently 
does.

Duncan Murdoch

>
> Martin
>
>      >> Alternatively, I can just give a warning and not attempt to load the rgl lib:
>      >>
>      >> > library(rgl)
>      >> Warning message:
>      >> X11 not found; XQuartz (from www.xquartz.org) is required to run rgl.
>      >>
>      >> rgl is now loaded, but it doesn't work; just about any function call will give an error, e.g.
>      >>
>      >> > plot3d(1,2,3)
>      >> Error in rgl.cur() : object 'rgl_dev_getcurrent' not found
>      >>
>      >> Do people have opinions about this?
>      >>
>      >> One comparable package is RGtk2:  if Gtk2 isn't install, it offers to install it.  I could probably do that for XQuartz.  If the user says no, RGtk2 gives really ugly error messages.  rgl can work without XQuartz, but as I already mentioned, making this work is complicated, so I'd like something simple for now.
>      >>
>      >> Duncan Murdoch
>      >>
>      >> ______________________________________________
>      >> R-devel at r-project.org mailing list
>      >> https://stat.ethz.ch/mailman/listinfo/r-devel
>
>      > ______________________________________________
>      > R-devel at r-project.org mailing list
>      > https://stat.ethz.ch/mailman/listinfo/r-devel
>