[R] Quasi-bug in boxplot().

[Martin Maechler]

>>>>> "PD" == Peter Dalgaard BSA <p.dalgaard at biostat.ku.dk>
>>>>>     on 01 Dec 2002 22:03:34 +0100 writes:

    PD> Rolf Turner <rolf at math.unb.ca> writes:
    >> >  It is defined *not* to work.  The call is
    >> >  
    >> >       boxplot(x, ..., range = 1.5, width = NULL, varwidth = FALSE,
    >> >               notch = FALSE, outline = TRUE, names, boxwex = 0.8, plot = TRUE,
    >> >               border = par("fg"), col = NULL, log = "", pars = NULL,
    >> >               horizontal = FALSE, add = FALSE, at = NULL)
    >> >  
    >> >  and argument after ... are not partially matched.
    >> >  
    >> >  See, e.g. `S Programming' p.40.
    >> 
    >> O.K. That's clear enough now.  But ***WHY***?  I.e. why structure the
    >> arguments of boxplot() in this way?  I.e. why put the ... argument
    >> before everything but x, so that partial matching cannot be used on
    >> the rest of the arguments?

    PD> Not quite sure about boxplot.default. 
I'm sure there:
How else should calls like

    boxplot(rnorm(100), rt(100,df=10), rt(110, df = 4), rcauchy(120))

work?  And these should work if only for S backcompatibility.
Martin

    PD> In boxplot.formula however,
    PD> there's a subset argument that will cause some grief if partially
    PD> matched. One other side effect of putting arguments after ... is that
    PD> it prevents positional matching, which might be the point in
    PD> boxplot.default -- avoid coding like boxplot(x, 1.5, NULL, TRUE, TRUE)
    PD> and the ensuing complaints if the argument order gets reshuffled.

Martin Maechler <maechler at stat.math.ethz.ch>	http://stat.ethz.ch/~maechler/
Seminar fuer Statistik, ETH-Zentrum  LEO C16	Leonhardstr. 27
ETH (Federal Inst. Technology)	8092 Zurich	SWITZERLAND
phone: x-41-1-632-3408		fax: ...-1228			<><