[R] animation without intermediate files?

[Paul Murrell]

Hi


Martin Maechler wrote:
>>>>>>"MM" == Martin Maechler <maechler at stat.math.ethz.ch>
>>>>>>    on Tue, 25 Jan 2005 09:59:03 +0100 writes:
>>>>>
> 
>>>>>>"Paul" == Paul Murrell <p.murrell at auckland.ac.nz>
>>>>>>    on Tue, 25 Jan 2005 13:40:15 +1300 writes:
>>>>>
> 
>     Paul> Hi
>     Paul> Cari G Kaufman wrote:
>     >>> Hello, 
>     >>> 
>     >>> Does anyone know how to make "movies" in R by making a
>     >>> sequence of plots?  I'd like to animate a long
>     >>> trajectory for exploratory purposes only, without
>     >>> creating a bunch of image files and then using another
>     >>> program to string them together.  In Splus I would do
>     >>> this using double.buffer() to eliminate the flickering
>     >>> caused by replotting. For instance, with a 2-D
>     >>> trajectory in vectors x and y I would use the following:
>     >>> 
>     >>> motif()
>     >>> double.buffer("back")
>     >>> for (i in 1:length(x)) {
>     >>>   plot(x[i], y[i], xlim=range(x), ylim=range(y))
>     >>>   double.buffer("copy")
>     >>> }
>     >>> double.buffer("front")
>     >>> 
>     >>> I haven't found an equivalent function to double.buffer in R.  I tried
>     >>> playing around with dev.set() and dev.copy() but so far with no success
>     >>> (still flickers).
> 
>      Paul> Double buffering is only currently an option on the Windows graphics 
>      Paul> device (and there it is "on" by default).  So something like ...
> 
>      Paul> x <- rnorm(100)
>      Paul> for (i in 1:100)
>      Paul>     plot(1:i, x[1:i], xlim=c(0, 100), ylim=c(-4, 4), pch=16, cex=2)
> 
>      Paul> is already "smooth"
> 
>     MM> well, sorry Paul, but not for my definition of "smooth"!
> 
>     MM> Instead, 
> 
>     MM> n <- 100
>     MM> plot(1,1, xlim=c(0,n), ylim=c(-4,4), type="n")
>     MM> x <- rnorm(n)
>     MM> for (i in 1:n) { points(i, x[i], pch=16, cex=2); Sys.sleep(0.02) }
> 
>     MM> comes much closer to my version of "smooth"  ;-)
> 
> I apologize to Paul, since  what I said seems to be quite
> platform dependent.  Here's my current "knowledge" on the matter:
> 
> o  Paul's  " for(..) plot(..) "
>     - flickers quite a bit for me {on Linux X11 with no
>       particularly fast graphics card}.
>     - seems quite smooth for at least two Windows users who have
>       relatively fast graphics cards.
> 
> o  My solution of
> 	   " for(..) { points(..) ; Sys.sleep(..) } "
>    doesn't redraw the coordinate system and so doesn't "flicker" 
>    (afaik, independently of platform)
> 
>    HOWEVER on windows; the graphics are somehow buffered and
>    points are not drawn one by one, but rather in batches --> "not smooth"


Thanks Martin; I wasn't very clear on my original message.  Double 
buffering has only been implemented on the Windows graphics device at 
this stage (thanks to Brian) and this implementation basically always 
writes to a buffer and updates the screen at fixed time intervals 
(quoting the source: "100ms after last plotting call or 500ms after last 
update") so there is no user control of when the off-screen buffer is 
swapped to the screen.

For "animating" a plot where only new output is added (i.e., no existing 
output is modified or removed), your suggestion should produce the 
smoothest result.

Paul
-- 
Dr Paul Murrell
Department of Statistics
The University of Auckland
Private Bag 92019
Auckland
New Zealand
64 9 3737599 x85392
paul at stat.auckland.ac.nz
http://www.stat.auckland.ac.nz/~paul/