R devel - Oct 2010 - strange interaction between rasterImage and Grid graphics

Dear all,

This may be specific to Mac, I haven't had a chance to test another
platform. Consider this,

plot(1,1,t="n")
rasterImage(matrix(1),1,1,1,1)
library(grid)
grid.rect(gp=gpar(fill="grey"))

The grid.rect covers the full device window as expected. However, when
I resize the window ever so slightly (interactive device) the rectGrob
is suddenly clipped to the previous plot window. I cannot understand
this behavior, and it doesn't happen if one removes the rasterImage()
call, so I suspect something iffy is going on with the display list or
something.

The only device I've tried is quartz(), x11() crashed with rasterImage,

 *** caught segfault ***
address 0x28, cause 'memory not mapped'

Traceback:
 1: rasterImage(matrix(1), 1, 1, 1, 1)

sessionInfo()
R version 2.11.1 (2010-05-31)
x86_64-apple-darwin9.8.0

locale:
[1] en_GB.UTF-8/en_GB.UTF-8/C/C/en_GB.UTF-8/en_GB.UTF-8

attached base packages:
[1] grid      stats     graphics  grDevices utils     datasets  methods   base

On Fri, Oct 1, 2010 at 12:17 AM, baptiste auguie
<baptiste.auguie at googlemail.com> wrote:
> Dear all,
>
> This may be specific to Mac, I haven't had a chance to test another
> platform. Consider this,
>
> plot(1,1,t="n")
> rasterImage(matrix(1),1,1,1,1)
> library(grid)
> grid.rect(gp=gpar(fill="grey"))
>
> The grid.rect covers the full device window as expected. However, when
> I resize the window ever so slightly (interactive device) the rectGrob
> is suddenly clipped to the previous plot window. I cannot understand
> this behavior, and it doesn't happen if one removes the rasterImage()
> call, so I suspect something iffy is going on with the display list or
> something.
I can reproduce the behaviour on GNU/Linux x11(type="cairo"), but this
is inappropriate mixing of base and grid graphics, so officially I
don't think you are allowed to expect anything at all.

-Deepayan

>
> The only device I've tried is quartz(), x11() crashed with rasterImage,
>
> ?*** caught segfault ***
> address 0x28, cause 'memory not mapped'
>
> Traceback:
> ?1: rasterImage(matrix(1), 1, 1, 1, 1)
>
> sessionInfo()
> R version 2.11.1 (2010-05-31)
> x86_64-apple-darwin9.8.0
>
> locale:
> [1] en_GB.UTF-8/en_GB.UTF-8/C/C/en_GB.UTF-8/en_GB.UTF-8
>
> attached base packages:
> [1] grid ? ? ?stats ? ? graphics ?grDevices utils ? ? datasets ?methods ?
base
>
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
>

Hi

baptiste auguie wrote:
> Dear all,
> 
> This may be specific to Mac, I haven't had a chance to test another
> platform. Consider this,
> 
> plot(1,1,t="n")
> rasterImage(matrix(1),1,1,1,1)
> library(grid)
> grid.rect(gp=gpar(fill="grey"))
 >
> The grid.rect covers the full device window as expected. However, when
> I resize the window ever so slightly (interactive device) the rectGrob
> is suddenly clipped to the previous plot window. I cannot understand
> this behavior, and it doesn't happen if one removes the rasterImage()
> call, so I suspect something iffy is going on with the display list or
> something.
It happens like this:

# 1. Clip to the device and draw axes and labels
plot(1,1,t="n")
# 2. Clip to the plot region and draw raster
rasterImage(matrix(1),1,1,1,1)
library(grid)
# Oooh!  This is the first time any grid drawing
# has occurred on the device, so initialize grid
# stuff, including the top-level viewport,
# *which clips to the device*
# 3. Draw a rectangle
grid.rect(gp=gpar(fill="grey"))

# Resize the window ...
# which triggers a redraw ...
# Oooh! There is grid output on this device so
# initialize grid stuff, including the top-level
# viewport *which clips to the device* ...
# 1. Clip to the device and draw axes and labels
# 2. Clip to the plot region and draw raster
# 3. Draw a rectangle

A workaround is to explicitly do a clip before the grid.rect(), i.e., ...

plot(1,1,t="n")
rasterImage(matrix(1),1,1,1,1)
library(grid)
grid.clip()
grid.rect(gp=gpar(fill="grey"))

... and I will add this example to the things I will look at when I am 
trying to clean up the grid code a bit.
> The only device I've tried is quartz(), x11() crashed with rasterImage,
That is more serious.  I have heard of a couple of others like this and 
I think the common thread may be 64-bit MacOS X.  I need to get access 
to such a beast to take a look.

Paul
>  *** caught segfault ***
> address 0x28, cause 'memory not mapped'
> 
> Traceback:
>  1: rasterImage(matrix(1), 1, 1, 1, 1)
> 
> sessionInfo()
> R version 2.11.1 (2010-05-31)
> x86_64-apple-darwin9.8.0
> 
> locale:
> [1] en_GB.UTF-8/en_GB.UTF-8/C/C/en_GB.UTF-8/en_GB.UTF-8
> 
> attached base packages:
> [1] grid      stats     graphics  grDevices utils     datasets  methods  
base
> 
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
-- 
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/