Skip to content

calling R from a shell script and have it display graphics

5 messages · Christophe Pallier, Don MacQueen, Seth Falcon +2 more

#
Hello,

I am running R under Linux/x11.

I would like to call R from a shell script and have it display a series 
of graphics.
The graphics should remain visible until the user clicks or presses a key.

I first tried R BATCH, but it does not load the x11 module, making it 
impossible to open x11 or png devices.

Then, I tried to call R with a 'here' script:

R --vanilla --quiet --args text.txt <<'EOF'
file=commandArgs()[5]
cat('processing ',file,'\n')
...
x11()
plot(f2,log='xy',type='b',las=1,cex=.5,xlab='rang',ylab='freq')
Sys.sleep(10)
q()
EOF

The problem with this approach is that the script cannot interact with 
the user.
par(ask=T) will fail because it reads input from the script rather than 
from the keyboard.

While I am writing this, a solution comes to my mind: I could save all 
the graphics in png format (using R <script.R), and when it is finished, 
call ImageMagick's display to show all the png (or use any other 
diaporama system). However, I find it a dirty hack.

Is there a simpler and cleaner way to achieve this?

Christophe Pallier
www.pallier.org
#
I don't know about the "simpler" part, but you could use the tcltk 
package to put up a window that prompts the user to continue.

-Don
At 11:57 AM +0100 2/12/04, Christophe Pallier wrote:

  
    
#
How about saving to png and writing a small html file and then launching
a browser?
On Thu, Feb 12, 2004 at 07:45:42AM -0800, Don MacQueen wrote:
#
> I don't know about the "simpler" part, but you could use
    > the tcltk package to put up a window that prompts the user
    > to continue.

Here's a function that does that.  I use to prompt the user to
choose among printing the current device with dev.print, saving
it with dev.copy2eps or continuing without doing anything.  I
write my codes to use this function and then, depending on
whether or not the code is running from a script or not, I
redefine it with the tk version.

Mike

##===================================================================================
## give the user some options at the command prompt
print.or.not <- function() {
  print("<return> to continue, `p' to print, `s' to save to R.ps...")
  result <- readline("[<ret>ps] ")
  if ( result == "p" ) { dev.print() }
  if ( result == "s" ) { dev.copy2eps(file="R.ps") }
}

##===================================================================================
## give the user some options in a tk window...
print.or.not <- function() {
  tt.print <- tktoplevel()
  tktitle(tt.print) <- 'Print or save plot?'
  b.print <- tkbutton(tt.print,
                      text='print',
                      command=function(...){ dev.print()
                                             tkdestroy(tt.print)
                                           }
                      )
  b.save <- tkbutton(tt.print,
                     text='save',
                     command=function(...){ file.obj <- tkgetSaveFile()
                                            file.name <- tclvalue(file.obj)
                                            dev.copy2eps(file=file.name)
                                            tkdestroy(tt.print)
                                          }
                     )
  b.continue <- tkbutton(tt.print,
                         text='continue',
                         command=function(...){ tkdestroy(tt.print)
                                              }
                         )
  tkpack(b.print, b.save, b.continue,fill='both')
  tkwait.window(tt.print)
}
#
Christophe Pallier <pallier at lscp.ehess.fr> wrote:
One trick is to use locator(1), which waits for a mouse click on a plot.
Here's a minimal Perl script that runs R, displays a plot, and exits when the
click is detected.


eval 'exec /usr/local/bin/perl -s -S $0 "$@"'
  if 0;
open(SCRIPT, "| R --vanilla --slave");
print SCRIPT <<'EOF';

x11(width=5, height=3.5)
plot(1:10, 1:10)
z <- locator(1)
q()

EOF
close(SCRIPT);