R devel - Dec 2000 - some (may be related) problems with windows(rescale=) (PR#795)

If these things bother you, could you please supply fixes?  For some reason
you omitted the part of the help page dealing with one of these (which is
as documented), and I also think you are confusing `plot' and `device'
sizes. The latter will be the displayed device surface, I believe.

Nothing on the help page suggested that the fine details of the R read-only
parameters will be as you expected after re-scaling the plot, and it never
occurred to me that anyone would expect so.  As they are reported by the R
graphics engine (which is used unchanged), I expect they are correct in one
of the multiple interpretations.

On Thu, 28 Dec 2000 joehl@web.de wrote:
>
###############################################################################
> 
> Before reporting 4 problems with windows(rescale=) I want to congrat on
R1.2 and to thank r-developers for quickly adding the rescale workaround to the
windows version.
> 
> Happy New Year
> 
> 
> Jens Oehlschlaegel
> 
> 
>
###############################################################################
> 
> 
> 
>
###############################################################################
> ## windows(rescale="fixed") does not keep original device size 
> ## (probably device initialization error)
>
###############################################################################
> 
> dev.show <- function(){
> 	din <- par("din")
> 	c(width=din[1], height=din[2], aspectratio=din[1]/din[2])
> }
> width <- 42/2.54
> height <- 29.7/2.54
> 
> > width
> [1] 16.53543
> > height
> [1] 11.69291
> > width/height
> [1] 1.414141
> 
> 
> > windows(width=width, height=height, rescale="R")
> > dev.show()
>       width      height aspectratio 
>    9.229167    6.385417    1.445351 
> > dev.off()
> null device 
>           1 
> > windows(width=width, height=height, rescale="fixed")
> > dev.show()
>       width      height aspectratio 
>    9.229167    6.385417    1.445351 
> > dev.off()
> null device 
>           1 
> > windows(width=width, height=height, rescale="fit")
> > dev.show()
>       width      height aspectratio 
>    9.229167    6.385417    1.445351 
> 
> 
> ## Shouldn't at least rescale="fixed" give the desired device
size and aspect ratio?
> ## The Help to windows() says:
> ## Using `"fixed"' will leave
> ##     the plot size unchanged, adding scrollbars if part of the plot is
> ##     obscured. This setting also affects the initial re-scaling if the
> ##     specified window size is too big.   
I don't think so.  It says the `plot' size, not the `device' size
which is
what par("din") reports.
>
###############################################################################
> ## windows(rescale="fit") gives strange character sizes 
> ## (looks like device initialization error)
>
###############################################################################
> 
> 
> ## In the following examples the created devices seem to be equally
initialized with respect to par(din)
> ## However, with windows(rescale="fit") I observe different
par(pin) and
different character sizes in text() and strwidth()
> 
> width <- 16.53543
> height <- 11.69291
> 
> windows(width=width, height=height, rescale="fixed")
> par("cex")
> plot(0:2,0:2, type="n")
> text(1,1,"abc\ndef", cex=2)
> text(1,0.5,"fixed", cex=2)
> temp <- t(sapply(par(c("din", "fin",
"pin")),c))
> colnames(temp) <- c("x", "y")
> temp <- rbind(temp, text=c(strwidth("abc\ndef", cex=2),
strheight("abc\ndef", cex=2)))
> fixed <- temp
> 
> windows(width=width, height=height, rescale="fit")
> par("cex")
> plot(0:2,0:2, type="n")
> text(1,1,"abc\ndef", cex=2)
> text(1,0.5,"fit", cex=2)
> temp <- t(sapply(par(c("din", "fin",
"pin")),c))
> colnames(temp) <- c("x", "y")
> temp <- rbind(temp, text=c(strwidth("abc\ndef", cex=2),
strheight("abc\ndef", cex=2)))
> fit <- temp
> 
> windows(width=width, height=height, rescale="R")
> par("cex")
> plot(0:2,0:2, type="n")
> text(1,1,"abc\ndef", cex=2)
> text(1,0.5,"R", cex=2)
> temp <- t(sapply(par(c("din", "fin",
"pin")),c))
> colnames(temp) <- c("x", "y")
> temp <- rbind(temp, text=c(strwidth("abc\ndef", cex=2),
strheight("abc\ndef", cex=2)))
> R <- temp
> 
> # now compare
> fixed
> fit
> R
> 
> > fixed
>              x         y
> din  9.2291667 6.3854167
> fin  9.2291667 6.3854167
> pin  8.0666667 4.6604167
> text 0.1394628 0.2896737
> > fit
>               x         y
> din  9.22916667 6.3854167
> fin  9.22916667 6.3854167
> pin  8.79000000 5.7337500
> text 0.07935154 0.1122302
> > R
>              x         y
> din  9.2291667 6.3854167
> fin  9.2291667 6.3854167
> pin  8.0666667 4.6604167
> text 0.1394628 0.2896737
> 
> 
> > ## It looks like character size of windows(rescale="fit")
refers to originally intended device size of
> > ## width=16.53543, height=11.69291 as 
> > fit["text", "x"]/R["din",
"x"]*width
> [1] 0.1415101
> > ## aprox. equal 
> > fixed["text", "x"]
> [1] 0.1405518
As documented:

  Using \code{\link{strwidth}} or
  \code{\link{strheight}} after a window has been rescaled gives
  dimensions in the original units, but only approximately as they are
  derived from the metrics of the rescaled fonts.
> ## If we now repeat this procedure with device sizes small enough to avoid
initial rescaling
> ## the phenomenon disappears
Right, as documented.
>
###############################################################################
> ## wrong row distances with windows(rescale="fit")
> ## (looks like device initialization error)
>
###############################################################################
> 
> 
> width <- 16.53543
> height <- 11.69291
> 
> ## as a reference
> windows(width=width, height=height, rescale="fixed")
> plot(0:2,0:2, type="n")
> text(1,1,"there is hardly anything i know for sure\nneither of the
future nor of the past\nnor should i know what to ask\nif the person answering
was you\n\n(Ernst Jandl, stanzen)"
> ,cex=2)
> 
> ## Here rows are too close together
> windows(width=width, height=height, rescale="fit")
> plot(0:2,0:2, type="n")
> text(1,1,"there is hardly anything i know for sure\nneither of the
future nor of the past\nnor should i know what to ask\nif the person answering
was you\n\n(Ernst Jandl, stanzen)"
> ,cex=3)
I can't reproduce that.  It may depend on the fonts available.
> ## Specifying smaller device size thus avoiding initial rescaling 
> ## makes the problem disappear
> 
> width <- 9
> height <- 6
> 
> windows(width=width, height=height, rescale="fit")
> plot(0:2,0:2, type="n")
> text(1,1,"there is hardly anything i know for sure\nneither of the
future nor of the past\nnor should i know what to ask\nif the person answering
was you\n\n(Ernst Jandl, stanzen)"
> ,cex=3)
> 
> ## Once correctly initialized further rescaling using the mouse does retain
correct distances
> 
> dev.off()
> dev.off()
> dev.off()
> 
> 
> 
>
###############################################################################
> ## windows(rescale="fit") destroys aspect ratio 
> ## (looks like device initialization error)
>
###############################################################################
> 
> 
> dev.show <- function(){
> 	din <- par("din")
> 	c(width=din[1], height=din[2], aspectratio=din[1]/din[2])
> }
> width <- 42/2.54
> height <- 29.7/2.54
> 
> > width
> [1] 16.53543
> > height
> [1] 11.69291
> > width/height
> [1] 1.414141
> 
> 
> ## create a device
> > windows(width=width, height=height, rescale="fit")
> > dev.show()
>       width      height aspectratio 
>    9.229167    6.385417    1.445351 
> 
> ## some resizing of the window BEFORE plot() is called
> # par("din") doesn't SEEM to change
> > dev.show()
>       width      height aspectratio 
>    9.229167    6.385417    1.445351 
>    
> # but in fact calling plot() and dev.show()
> > plot(0,0)
> > dev.show()
>       width      height aspectratio 
>    6.321506    5.562500    1.136451 
> 
> ## now some resizing of the window AFTER plot() is called
> > dev.show()
>       width      height aspectratio 
>    4.708333    4.500471    1.046187 
> > dev.show()
>       width      height aspectratio 
>    4.718750    3.297956    1.430811 
> > dev.show()
>       width      height aspectratio 
>   3.3255361   4.4375000   0.7494166 
> > dev.show()
>       width      height aspectratio 
>   1.4091170   5.1145833   0.2755096 
> > dev.show()
>       width      height aspectratio 
>   1.4062500   5.1145833   0.2749491 
> > dev.show()
>       width      height aspectratio 
>   0.4701945   3.3854167   0.1388882 
>   
> ## we got very differing aspect ratios
> ## One reason for that COULD be that the initially intended device sizes
are not kept,
> ## so that many rescalings cumulate the deviations.
> ## May be par(din) should not be changed at all (keep the representation
layer stable)
> ## and only apply rescaling/zooming/scrolling only to the graphical
presentation.
Possibly.  I don't think the device size is very interesting after
rescaling.  What are you using it for?

-- 
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 272860 (secr)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595


-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To:
r-devel-request@stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._