R versions and PostScript files

This is due to the introduction of sRGB. Since this actually does
something (Google for sRGB and you will be approximately as wise as
me...), I don't think it is likely to be taken out. You can, however,
always edit .ps.prolog. (I would expect that the line

/setrgb { setrgbcolor } def

instead of what is already there would reinstate the old behavior, but
no guarantees.
)


On Jul 21, 2011, at 17:26 , (Ted Harding) wrote:

On 21-Jul-11 13:24:32, Duncan Murdoch wrote:

On 11-07-21 3:23 AM, pilchat wrote:

Dear R users,

I have a desktop computer and a laptop, both of them
with Ubuntu Lucid. The former has R2.10 installed from
Ubuntu repositories (this is the most recent version
in the repositories), while the latter has R2.13 from
the CRAN repositories.

I noticed that postscript files generated with R2.10
are "better" than files generated with the latest release
of R, in particular for plots with colored areas, such
as the output of image or persp. The thing is that my ps
viewer (e.g. gv or evince) is very slow in opening ps
files from R2.13, while it smoothly displays ps files
from R2.10, regardless of "encapsulation".

I think this is related to differences in the way the
ps file is generated by the two versions of R, but I
don't know how to go deeper in the matter.

Postscript files are mostly text, so you can compare the
two files and > see what the differences are. The NEWS
file shows a number of changes since 2.10.0, but I can't
see any that would cause problems for viewers.

Duncan Murdoch

Is there anyone experiencing the same issue? Is there
any solution?

Thank you in advance

Cheers
Gaetano

Gaetano has now sent me two files, generated (as he posted
just now on R-help) by the same commands:

 setEPS()
 postscript (file="volc.eps",width=5,height=4)
 image(volcano)
 dev.off()

on his two machines:

volc2.10.eps generated using R-2.10 on his desktop
 (the EPS file with fast rendering)

volc2.13.eps generated using R-2.13 on his laptop
 (the EPS file with slow rendering)

I have viewed both files on the same machine, and the
result indeed is that while volc2.10.eps renders very
quickly, volc2.13.eps does render very slowly (painting
in by vertical strips which move jerkily from left
to right). I estimate that 'gv volc2.10.eps' does the
rendering in less than 1 second, while 'gv volc2.13.eps'
takes about 18 seconds.

Comparing the two files, I think I have found the reason.

A 'diff' on the two files shows a basic difference in
definitions of a function used in the plotting:

[A] In file volc2.10.eps (the fast one):

 /rgb { setrgbcolor } def

[B] In file volc2.13.eps (the slow one):

 /srgb { [ /CIEBasedABC
           << /DecodeLMN
                [ { dup 0.03928 le
                         {12.92321 div}
                         {0.055 add 1.055 div 2.4 exp }
                      ifelse
                  } bind dup dup
                ]
              /MatrixLMN [0.412457 0.212673 0.019334
                          0.357576 0.715152 0.119192
                          0.180437 0.072175 0.950301]
              /WhitePoint [0.9505 1.0 1.0890]

          ] setcolorspace } def
 /setrgb { srgb setcolor } def


Then [A] volc2.10.eps (the fast one) uses commands like:

 /bg { 1 0 0 rgb } def

while [B] volc2.13.eps (the slow one) uses commands like:

 /bg { 1 0 0 setrgb } def

in each case for exactly the same purpose. Thus [B] the
slow one uses repeatedly (1157 times) a function setrgb
which has much higher overheads (see definition above)
than the function rgb used (1156 times) by [A] the fast one.

So the difference in performance is *definitely* down to
a specific difference in how R-2.13 implements 'postscript()'
compared with R-2.10.

Hoping this is useful!
Ted.

--------------------------------------------------------------------
E-Mail: (Ted Harding) <ted.harding at wlandres.net>
Fax-to-email: +44 (0)870 094 0861
Date: 21-Jul-11                                       Time: 16:26:33
------------------------------ XFMail ------------------------------

-- 
Peter Dalgaard
Center for Statistics, Copenhagen Business School
Solbjerg Plads 3, 2000 Frederiksberg, Denmark
Phone: (+45)38153501
Email: pd.mes at cbs.dk  Priv: PDalgd at gmail.com