Copying a device (former: Legend text -- discrepancy ...)

On Thu, 15 Jan 2004, Prof Brian Ripley wrote:

The short answer is not to copy the device, but to replot on the new 
device.  That is the advice given in MASS, for example.

But these device-copy functions could have been quite handy -- 
especially after a long sequence of plotting commands that are 
done interactively.

When you copy a device, you replay the device list and hence the lines and 
text are placed at the positions calculated using the font metrics of the 
first device and not the second.  dev.copy2eps does not try to adjust the 
pointsize of the postscript device, and provided the fonts match you 
should just be able to adjust the pointsize in this case.

OK, so I tried various things and the best I could come up with 
is replacing [My X11() width and height defaults are 7]
	dev.copy2eps(file="test.eps", paper="letter")
with
	dev.copy2eps(file="test.eps", paper="letter",
			width=8, height=8)

Let's see if I understood what I did above:
the physical size of the X11 and PS fonts is different. 
Therefore, instead of changing the fontsize we re-scale the plot.
(This is what is implied by the various printouts I have made.)
If so, is there a way to reduce the font size, instead of  
increasing the plot size?

(To avoid the plot extending beyond the physical page, for 
example.)

If the only way is to change the EPS device dimensions how could 
I do it in a more robust way?
Is, e.g., 'width=some.factor*par("din")[1]' a sensible way?
Is there a better way?
Could I pre-determine some.factor?

You do need to be suspicious of on-screen viewers and indeed of 
ghostscript, for they are often not pixel-perfect and ghostscript does 
font substitution (it does not have Helvetica).  I would always test by 
printing on a postscript printer.

Screen and print rendering were the same.