Skip to content

gpar fill and transparency on devices

4 messages · Baptiste Auguie, Paul Murrell

Original
Baptiste Auguie
# 
Dear list,

I'm puzzled by the graphical output in the following example,

library(grid)

foo <- function(){
  grid.rect(gp=gpar(fill="black"))
  print(get.gpar()$fill)
  grid.rect(width=0.2,height=0.2)
}

png("test.png", bg = "transparent")
foo()
dev.off()

png("test1.png", bg = "white")
foo()
dev.off()


It seems that the default value of gpar()$fill is set according to the
device background. I couldn't find this documented in ?gpar or in
?png, and it caused a rather puzzling bug in my code (the pdf() output
was OK, whilst the png output (default bg to white) was seemingly
empty because covered by a white rectangle.)

Best regards,

baptiste
1 day later
Paul Murrell
# 
Hi

The help page for "Working with Viewports" (e.g., pushViewport()) has a 
brief mention when talking about the ROOT viewport ...

"The viewport tree always has a single root viewport (created by the 
system) which corresponds to the entire device (and default graphical 
parameter settings)."

... which is a reasonable place for it because this is a feature of the 
gpar of the ROOT viewport, not of gpars in general.  That mention might 
be a bit hard to find, but a very similar statement is also made in the 
Section on Viewports in the R Graphics book.  That in turn might be hard 
to find if you don't have the book, but that chapter is also available 
online (http://www.stat.auckland.ac.nz/~paul/RGraphics/chapter5.pdf)

It might be worth adding something more explicit about this sort of 
gotcha ...

"Some devices have different default graphics parameter settings, so it 
is not safe to assume that the ROOT viewport will be identical on 
different devices."

... ?

Paul
On 5/08/2010 8:14 a.m., baptiste auguie wrote:

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Baptiste Auguie
    

    

      
      #
    

  


  

      
Hi,

Thanks for pointing this out. I do have the book, unfortunately I left
it abroad this year. I would think that such brief mention in ?gpar
could be useful (because that's where one first looks for gpar()
defaults --- which are not listed).

Best regards,

baptiste

        
On 6 August 2010 00:54, Paul Murrell <p.murrell at auckland.ac.nz> wrote:

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

  
    

      
      
    

  


  


      

    

    
2 days later

    

      
      

        


  

        

      Paul Murrell
    

    

      
      #
    

  


  

      
Hi

        
On 6/08/2010 5:34 p.m., baptiste auguie wrote:

            

      
      
      
    

        
Righto.  I've added a mention in ?gpar

Paul