Unit of legend() coordinates (was: Re: [R] lines and legend)

[Prof Brian Ripley]

On Wed, 2 Jul 2003, Paul Lemmens wrote:

> I would like to make a suggestion for an improvement (IMHO) of ?legend.
> 
> Please add a remark that the x,y positioning coordinates are in the units 
> of the plot() itself, *not* in pixels or anything the like.

That's true of all x-y coordinates of all R plotting functions.  We don't
say it in ?points, for example.

> It would have saved me a lot of time figuring out why my legend just 
> wouldn't appear if ?legend would have told me to set the coordinates in the 
> units of the x and y axis. I haven't found it anywhere on the list or 
> manuals (but I may have missed something). Now that I know this, it seems a 
> logic choice, however, refering to a 'graphics device' seems to imply a 
> coordinate system in pixels starting from the upper/lower left of the 
> device (the usual choice in graphical manipulations in programming 
> languages). This assumption is false and IMHO is nowhere in FAQ/manuals 
> contradicted.

There are several examples on the help page for legend which contradict
that assumption, so I don't see how you can have maintained it after 
reading them, let alone after trying them out.

The coordinate systems are explained in detail in `An Introduction to R':
The phrase `graphics device' does not appear on the help page for legend,
so you must have found that somewhere else.


Is it reasonable to expect the help system/manuals to contradict all
possible misconceptions of all users?  If we attempted to do that, the
help pages would get impossibly cluttered by even those misconceptions
which come to light.

-- 
Brian D. Ripley,                  ripley at 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 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595