R devel - May 2004 - windows(record=T) loses last plot (PR#3663)

The audit trail indicates that PR#3663 is not reproducible and/or fixed,
but I believe windows(record=T) is still not working as documented:

 windows(record=T)
 for(i in 1:5) plot(0, 0, cex=4, pch=as.character(i))

After cycling through the plots with PgUp and PgDn, the user finds out
that plot 5 is lost. Indeed,

 .SavedPlots

reports that only 4 plots were saved. I'm running the example in a fresh R
session, after rm(list=ls(all=T)), quitting, and restarting. I have
reproduced this bug on two machines, one with a fresh R installation.

Thanks,
Arni

OS:       Windows XP
R:        1.9.0
graphics: 1.9.0

On Sun,  9 May 2004 00:30:14 +0200 (CEST), arnima@u.washington.edu
wrote :
>The audit trail indicates that PR#3663 is not reproducible and/or fixed,
>but I believe windows(record=T) is still not working as documented:
>
> windows(record=T)
> for(i in 1:5) plot(0, 0, cex=4, pch=as.character(i))
>
>After cycling through the plots with PgUp and PgDn, the user finds out
>that plot 5 is lost. Indeed,
>
> .SavedPlots
>
>reports that only 4 plots were saved. I'm running the example in a fresh
R
>session, after rm(list=ls(all=T)), quitting, and restarting. I have
>reproduced this bug on two machines, one with a fresh R installation.
I vaguely recall that the issue was that history saving doesn't happen
until drawing to the graph is done.  Since that never happened for the
last one, it doesn't get saved.

A way to fix this would be to implicitly save the graph when
responding to a request to look at a different page.  Another would be
to save a graph as soon as it is created; graphs on the history list
can be modified, and the modifications stick.

Any chance you could track down what patch is necessary?  I'm not
likely to get to it in the next few weeks, and it would be nice to
have this fixed in 1.9.1.

Duncan Murdoch

The claim in the original report is that it loses the 5,9th etc plot if it 
is the last, but not otherwise.  At one time the example in the report 
worked (but that was a long time ago and there have been extensive 
changes since then.).

The problem was that growing the list changed vDL, and Guido's macros hid 
that.  A simple fix once found ....

On Mon, 10 May 2004, Duncan Murdoch wrote:
> On Sun,  9 May 2004 00:30:14 +0200 (CEST), arnima@u.washington.edu
> wrote :
> 
> >The audit trail indicates that PR#3663 is not reproducible and/or
fixed,
> >but I believe windows(record=T) is still not working as documented:
> >
> > windows(record=T)
> > for(i in 1:5) plot(0, 0, cex=4, pch=as.character(i))
> >
> >After cycling through the plots with PgUp and PgDn, the user finds out
> >that plot 5 is lost. Indeed,
> >
> > .SavedPlots
> >
> >reports that only 4 plots were saved. I'm running the example in a
fresh R
> >session, after rm(list=ls(all=T)), quitting, and restarting. I have
> >reproduced this bug on two machines, one with a fresh R installation.
> 
> I vaguely recall that the issue was that history saving doesn't happen
> until drawing to the graph is done.  Since that never happened for the
> last one, it doesn't get saved.
That was originally true, but was changed.
> A way to fix this would be to implicitly save the graph when
> responding to a request to look at a different page.  Another would be
> to save a graph as soon as it is created; graphs on the history list
> can be modified, and the modifications stick.
> 
> Any chance you could track down what patch is necessary?  I'm not
> likely to get to it in the next few weeks, and it would be nice to
> have this fixed in 1.9.1.
-- 
Brian D. Ripley,                  ripley@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