[R] Legend problem when exporting a plot to PDF

Joerg van den Hoff j.van_den_hoff at fzd.de
Tue Apr 29 15:12:29 CEST 2008

On Tue, Apr 29, 2008 at 01:49:01PM +0200, Philipp Pagel wrote:
> > When exporting to PDF a graph with a legend, in the final PDF, the
> > text  is going beyond the legend box.
> > > dev2bitmap("test.pdf", type="pdfwrite", h=6, w=6)
> > The legend looks OK on the screen.  I noticed that the size of the
> > legend box depends on the size of the  screen window,
> As far as I remember, te problem has to do with different font handling
> in different devices. I'm sure someone more familiar wiht the internals
> will comment on this.
> The fix is easy: don't use dev2bitmap but open the desired target device
> before plotting. In your case: 
> pdf(file='test.pdf', width=6, height=6)
> plot(...)
> legend(...)
> dev.off()
> cu
> 	Philipp

I'm not sure, whether this is the way to go. at least until recently the `pdf'
device of R had a few rough edges which surfaced by an then. In my experience
using `dev2bitmap(type="pdfwrite", ...)' -- and thus ghostscript for pdf
generation resulted in cleaner/better pdf output (actually, `dev2bitmap' sort
of postprocessess output from the `pdf' device a bit).

concerning the original question: yes, the problem is there. I believe it
is related simply to the fact that the same absolute font size used in the graphic window
is used in the pdf output, too, which makes its _relative_ size  dependent on the
chosen size (widht/height) of the pdf output. 

my workaround is to ensure that the graphic windows size is exactly the same
as that used in the `dev2bitmap` call. e.g. the X11() device has defaults of
width = 7, height = 7 (but you can enforce other values by open a new one with,
e.g., X11(width = 6, height = 9). thus, first plot to graphic window, make sure
that no interactive resizing is necessary to get acceptable display on the
monitor (otherwise close it and open new X11() with better width/height values). 
then call dev2bitmap() with the same width/height settings. that should lead essentially
to WYSIWYG (modulo font substitution, maybe).


More information about the R-help mailing list