R-alpha: Re: lwd, cex and par(.) issues [was "... windows quirks .."]

Martin Maechler Martin Maechler <maechler@stat.math.ethz.ch>
Mon, 8 Sep 1997 16:21:56 +0200

[[yet another discussion brought from  R-core  to  R-devel]]

>>>>> "RG" == Robert Gentleman <rgentlem@stat.auckland.ac.nz> writes:

    RG> Peter, there is some code for line types and widths in all the
    RG> drivers. With windows there is the well known limitation that you
    RG> can't have patterened lines with lwd larger than 1 (unless you
    RG> write your own routines ....  I just haven't had time). I do agree
    RG> that there is something very weird going on with our code though!
    RG>    I think that in plot(1:4,lwd=2) the lwd should have absolutely
    RG> no effect. I think it's a bug in Splus that it does. For example,
    RG> plot(1:4,col="red")

I agree, and I also have been hating this (Splus) behavior.

I do prefer R's behavior have different 'col', 'cex', 'font' for
axis, labs, and points/lines.

After reading PD's reply to this message, I agree that
"lwd" is somewhat different:
	If your device has too thin lines in general, it is nice if you can
	increase line width  GLOBALLY.
The same for "cex": 
	There are situations where ALL text and plotting characters should
	just be made bigger with the same zoom factor.

Below, you (RG) mention the use of   par(lwd=...)
{and correspondingly would be    par(cex=...)}.

Is this going to be ``the standard'':

	For global changes of  linewidth or character size,
	one must use  par(lwd=..) or  par(cex=..) 

    RG>     gives you everything in red; axes, labels, lines and
    RG> points. That's wrong.  In plot(1:4,type="l",lwd=2) lwd should
    RG> affect the width of the line joining the points but not the width
    RG> of the lines for the bounding box (IMHO).

I agree, but again, it might be nice if there was something like a 
    RG>     I think that if you want to change the line widths on the
    RG> bounding box you should use par(lwd=2) and then all lines are
    RG> changed. This seems to work (but I don't doubt there are bugs; we
    RG> simply have never had the chance to exercise everything).

    RG>     As for character expansion- par cex seems to be the only route
    RG> (which makes me like my suggestion even more).
    RG>     Opinions?  It would be easy enough to mimic S but I do think
    RG> it's wrong.

We should not mimic S here, sometimes.
I agree it's wrong in some places.

But if you use a different scheme, it should be
	- well documented
	- very consequently implemented   ==>  easily remembered by users

-- Martin

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