recordPlot in 1.5.x, replayPlot in 1.6.0 problem

I have plots that were stored using recordPlot() in a fairly recent
but pre-1.6.0 version of R that will not replayPlot() in R 1.6.0. The
error message is
    Error in replayPlot(x) : invalid hex digit in color or lty

The plots were stored in June (2002-6-14), each as an element of a
list named 'rpl'. The version of R that was current then is what I
used. At that time I was able to re-display them.

Today, in R 1.6.0, when I try to re-display them, I get an error message:

 rpl[[1]]

Error in replayPlot(x) : invalid hex digit in color or lty

The plots did have lines with col='blue' and lty=2 in them. This
little snippet of code executed today works

  > plot(1:5)
  > lines(1:5,col='blue',lty=2)
  > tmp <- recordPlot()
  > tmp

so there doesn't seem to be a problem if I work entirely within
v1.6.0. Which suggests that the problem is in going from one version
to the next.

 version

          _
platform sparc-sun-solaris2.7
arch     sparc
os       solaris2.7
system   sparc, solaris2.7
status   Patched
major    1
minor    6.0
year     2002
month    10
day      13
language R

Don MacQueen wrote:

 I have plots that were stored using recordPlot() in a fairly recent
 but pre-1.6.0 version of R that will not replayPlot() in R 1.6.0. The
 error message is
     Error in replayPlot(x) : invalid hex digit in color or lty

 The plots were stored in June (2002-6-14), each as an element of a
 list named 'rpl'. The version of R that was current then is what I
 used. At that time I was able to re-display them.

 Today, in R 1.6.0, when I try to re-display them, I get an error message:

 >  rpl[[1]]

 Error in replayPlot(x) : invalid hex digit in color or lty

 The plots did have lines with col='blue' and lty=2 in them. This
 little snippet of code executed today works

   > plot(1:5)
   > lines(1:5,col='blue',lty=2)
   > tmp <- recordPlot()
   > tmp

 so there doesn't seem to be a problem if I work entirely within
 v1.6.0. Which suggests that the problem is in going from one version
 to the next.

 >  version

           _
 platform sparc-sun-solaris2.7
 arch     sparc
 os       solaris2.7
 system   sparc, solaris2.7
 status   Patched
 major    1
 minor    6.0
 year     2002
 month    10
 day      13
 language R

I have tested your example with R-1.3.1 (!), R-1.5.1 (record) and
R-1.6.0, R-1.6.1 (replay) --- with success.
So it is really of interest
 a) to get a reproducable example and/or
 b) to know what "fairly recent" means.

Uwe Ligges

Thank you for responding.

I checked the dates, and "fairly recent" means R-1.5.0 in this case.
I will try to get a reproducible example.

Hi


Don MacQueen wrote:

 Thank you for responding.

 I checked the dates, and "fairly recent" means R-1.5.0 in this case.
 I will try to get a reproducible example.

I have a reproducible example.  Try this in R 1.5.0 ...

plot(1:10)
temp <- recordPlot()
save(temp, file="dl")

... then this in R 1.6.0 ...

load(temp)
x11()
replayPlot(temp) 

... and the tick-marks on the axis are a bit funny (dotty).

<GORY DETAIL>
I think the problem is that the interface for the internal axis code has
changed.  In 1.5.0 we had ...

    /* axis(side, at, labels, tick,
     *	     line, pos, outer, font, vfont, ...) */

... in 1.6.0 we have ...

    /* axis(side, at, labels, tick, line, pos,
     *	    outer, font, vfont, lty, lwd, col, ...) */

What gets saved by recordPlot() is the display list which is a list of
internal C calls.  What you are seeing is the 1.6.0 code being called
with a 1.5.0 argument list and the lty, lwd, and col arguments are
missing so goodness knows where they are being taken from.
</GORY DETAIL>

In simple terms, this is a danger of trying to use display lists across
different versions of R -- the information that is stored in the display
list *may* be version-specific.  Ideally we would detect these changes
automatically -- sorting out display lists is a todo item.

In case it is of any use, I have a very basic discussion of saving and
reusing graphics output at
http://www.stat.auckland.ac.nz/~paul/grid/saveload.pdf
This is aimed at using grid, but some of it is relevant for base
graphics too.

Paul