[Rd] PR#6614

[Peter Dalgaard]

Bill Dunlap <bill at insightful.com> writes:

> On Wed, 23 Feb 2006, Peter Dalgaard wrote:
> 
> > "Gabor Grothendieck" <ggrothendieck at gmail.com> writes:
> >
> > > Try this:
> > >
> > >  subset(iris, select = - Species)
> >
> > Or, canonically,
> >
> > nm <- names(iris)
> > iris[, nm != "Species" ]
> >
> > iris[, -match("Species", nm)]
> >
> > >
> > > On 2/22/06, davidhughjones at gmail.com <davidhughjones at gmail.com> wrote:
> > > > I agree with the submitter that this needs some kind of solution.
> > > > Although data.frame[,-12] works, how do I drop a named column (the
> > > > most common use case)? (I found this bug while searching for an
> > > > answer.)
> 
> That code with match only works if "Species" is actually
> a column of iris.  If not, your result depends on whether
> you have a data.frame (an error) or a matrix (a column
> of NA's).
> 
> I found some mail in a 15 year old sent-mail file that suggested
> allowing the tag 'except=' on any of the arguments to "[" to
> replace/extend the limited negative integer convention.  The idea was
> that
>     iris[ except=c(10:20), except=c("Petal.Width","Petal.Length")) ]
> would return all rows except 10:20 and all columns except
> the ones named.
>     iris[ except=integer(0), ]
> would return all rows of iris, while iris[-integer(0), ] returns
> no rows of iris.
> 
> This abuses the tag= notation, but the "[" function doesn't
> really support the i= and j= tags that some people expect.
> 
> This would take care of the problem that subset(data.frame,select=-name)
> only lets you omit columns.
> 
> The mail had a version of [.data.frame (for Splus 2.1?) that
> implemented this, although, if it is to be used it should be
> implemented in the most primitive [ code so all methods use it.

>From the language viewpoint, I think that using the same tag multiple
times and mixing up tagged and positional argument matching would be
quite nasty. I actually suspect that the argument matching rules will
make it impossible in R (except possibly from C and .Primitive). 

However, it could be possible to have an omit() function, as in

iris[ omit(c(10:20)), omit(c("Petal.Width","Petal.Length")) ]  

all this needs to do is to tack an attribute onto the index which the
indexing code can look for. 

-- 
   O__  ---- Peter Dalgaard             Øster Farimagsgade 5, Entr.B
  c/ /'_ --- Dept. of Biostatistics     PO Box 2099, 1014 Cph. K
 (*) \(*) -- University of Copenhagen   Denmark          Ph:  (+45) 35327918
~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk)                  FAX: (+45) 35327907