[Rd] Sweave support added to rgl package

Duncan Murdoch murdoch.duncan at gmail.com
Thu Apr 21 18:48:30 CEST 2011


I've just committed some changes to rgl that will probably detect 
bitmaps that are specified to be too large.  As well, Brian Ripley 
backported some of the R-devel additions to the RweaveLatex driver, so 
now R 2.13.0-patched (revision 55572 or newer) should
work as well as R-devel.

Duncan Murdoch

On 20/04/2011 8:56 PM, Duncan Murdoch wrote:
> On 20/04/2011 7:10 PM, Dominick Samperi wrote:
> >  On Wed, Apr 20, 2011 at 2:29 PM, Duncan Murdoch
> >  <murdoch.duncan at gmail.com>   wrote:
> >>  On 20/04/2011 1:52 PM, Dominick Samperi wrote:
> >>>
> >>>  On Tue, Apr 19, 2011 at 2:18 PM, Duncan Murdoch
> >>>  <murdoch.duncan at gmail.com>     wrote:
> >>>>    I have just committed some code to the rgl package on
> >>>>    https://r-forge.r-project.org/projects/rgl/ to allow rgl images to be
> >>>>    inserted into Sweave documents.  (This is not in the CRAN version yet.)
> >>>>    It
> >>>>    makes use of the custom graphics driver support added by Brian Ripley.
> >>>>
> >>>>    In R-devel (which will become R 2.14.0 next spring in New Zealand, next
> >>>>  fall
> >>>>    in most other places), usage is quite straightforward.  For
> >>>>    example, code like this in a Sweave document:
> >>>>
> >>>>    <<fig=true, grdevice=rgl.Sweave, pdf=false, stayopen=TRUE>>=
> >>>>    x<- rnorm(100); y<- rnorm(100); z<- rnorm(100)
> >>>>    plot3d(x, y, z)
> >>>>    @
> >>>>
> >>>>    will insert a .png snapshot of the figure.  Because that chunk has
> >>>>    "stayopen=TRUE", it can be followed by another chunk to add
> >>>>    to the figure, e.g.
> >>>>
> >>>>    <<fig=true, grdevice=rgl.Sweave, pdf=false>>=
> >>>>    lines3d(x[1:10], y[1:10], z[1:10], col="red")
> >>>>    @
> >>>>
> >>>>    All of this is possible in R 2.13.0, but it takes more work:  see the
> >>>>    ?rgl.Sweave help page.
> >>>>
> >>>>    I will eventually add postscript and PDF output options as well, and
> >>>>  perhaps
> >>>>    some support for the LaTeX movie15 package, but those are not there
> >>>>  yet.
> >>>>      Comments or bug reports are welcome.
> >>>>
> >>>>    Duncan Murdoch
> >>>
> >>>  I inserted your example into testrgl.Rnw under R 2.13.0, with
> >>>  Sweave.snapshot()
> >>>  at the end of both chunks, but things did not work as expected.
> >>>
> >>>  I used:
> >>>  $ R CMD Sweave testrgl.Rnw
> >>>  $ pdflatex tesetrgl
> >>>  (view testrgl.pdf)
> >>>
> >>>  When R CMD Sweave is run the graphics is displayed interactively.
> >>
> >>  That's unavoidable as far as I know.  I don't think there's a general
> >>  purpose way to tell OpenGL to render in the background, so it works by
> >>  rendering on screen, then copying a bitmap to the .png file.
> >>>
> >>>  There is no graphics in the PDF file, even though both .png files
> >>>  are read when pdflatex is run.
> >>
> >>  Do they look okay?  One possible problem is that you may have asked for a
> >>  bitmap too big for your hardware to render, in which case those png files
> >>  will end up with junk (probably blank).  Setting resolution=100 in the chunk
> >>  headers will do it more coarsely.  (The default is 300 dpi.)  The same
> >>  effect comes from width=1, height=1  (or some other small numbers).
> >>
> >>  Duncan Murdoch
> >>
> >
> >  The resolution=100 tip fixed the problem, thanks.
>
> I'll see if I am skipping over some error message in there.  It would be
> much better for Sweave to fail with an error than generate empty images.
>
>
>   >  Now I see the snapshots
> >  in the PDF file. Using this in a package will certainly change the
> >  user experience,
> >  but it moves away from the traditional batch-oriented R package
> >  processing, it seems to me.
>
> I don't follow that.
>
> >  The idea of adding support for movies and 3D graphics to Sweave/PDF files
> >  sounds very interesting and revolutionary.
>
> Movies will likely be pretty slow.  I think you'll want caching of some
> sort if you want to produce those.
>
> Duncan Murdoch
>
> >
> >  Dominick
>



More information about the R-devel mailing list