[R] R versions and PostScript files

[Prof Brian Ripley]

On Fri, 22 Jul 2011, James Cloos wrote:

Please do note the posting guide: the context you omitted (when 
explicitly asked not to) is in the thread starting:

https://stat.ethz.ch/pipermail/r-help/2011-July/284352.html

> I suspect that a few s/def/bind def/ on that prologue might improve
> things a bit.

No discernable difference on the viewers I tested.

> The ps R creates freely switches between the sRGB ABC space and
> setgray.  That is a good thing; it helps ensure that black will
> be just black when sent to a CMYK device.  Without that you can
> get CMY black from some devices.  But it does complicate things.
>
> (CMYK devices often use CMY black rather than K black when the
> source is RGB black because it is better when printing contone
> images, such as photographs.  But it just wastes ink and toner
> when printing charts and similar graphics.)

I've not seen a printer driver which does that for many years.  But 
remember that R's postscript() has origins in  the mid-1990s.

R's pdf() is newer (R 1.3.0, 2001) and even by then the optimization 
could safely be left to the driver.

> The best improvement would be to have R keep track of the last
> colorspace and only run the srgb procedure when switching from
> DeviceGray, rather than every time it changes the RGB colour.

Actually no, as it can do that rather often.  Consider the pairs plots 
of the iris data in demo(graphics): that does (1200 times)

switch to red, blue or green
fill circle
switch to black
draw circle outline.

Some viewers are rather slow in running that procedure, and on those 
viewers this example is slow.

Workarounds:

1) Use pdf(), where the viewers seem to cache colorspaces.
2) In R-patched, use colormodel="rgb-nogray", which never switches 
colorspace ("srgb-nogray" in R-devel).
3) In R-devel, use colormodel="rgb" (rather than "srgb") and give up 
the colour accuracy (if your viewer is not already set for sRGB) for 
speed.

> The SetColor() function in R/src/library/grDevices/src/devPS.c
> looks to be the place to do that.  The Invalidate() function
> and the PostScriptDesc struct there also need to track any
> such changes.

Hmm, you seem not to understand how the code works.  The postscript 
macro /bg switches colour, and the macros /p3 and similar use 
different colorspaces during their execution.  It isn't just SetColor 
as its setting may not be used immediately.

> -JimC
> -- 
> James Cloos <cloos at jhcloos.com>         OpenPGP: 1024D/ED7DAEA6

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595