On Mon, 26 Apr 2010, Stuart Andrews wrote:
Thanks Charles, for clarifying.My statement holds for matrices, which are 2 dimensional. And, as you mentioned, a single index implies vector indexing where the drop argument doesn't make sense. I am somewhat relieved, given this new understanding.But I am still puzzled as to why R doesn't complain about the unused "drop=F" argument. Since this argument is nonsensical, R should tell me this, no?
I take your point that user who was intending to type a vector subscript would not also bother to type 'drop=FALSE', and from the POV of the user typing "a[1:3,drop=F]" at the keyboard, it makes good sense to me to have R tip him/her off to a potential goof.
But subscripts get used a lot down deep in the code, so there are other considerations, I suspect. If you are interested in this from a 'code development' POV, you might repost to R-devel to ask what might have lead to this 'feature'. I have two guesses:
1) subscripting is a pretty low-level operation and checking for this particular case is not important enough to justify the added code and perhaps slowing things down.
2) a construction like
do.call("[", c(list(x), args, list(drop = FALSE)))
will work without comment as long as the 'args' argument delivers a valid
subscript or set of subscripts (valid for 'x') and the drop = FALSE
will always be innocuous even if it is superfluous.
For example, when I add an unrecognized argument to the ls() function I get the following:ls(nonsense="42")Error in ls(nonsense = "42") : unused argument(s) (nonsense = "42")
Priorities! Failure of argument matching is a critical error. Having a valid but superfluous argument is not, and there are loads of settings in which no error is raised like
y <- rnorm(3)
x <- 1:3
lm(y~x,data=list("blah"))
which arguably reveals a mistake by the user of the kind that having the unneeded 'drop=FALSE' suggested.
HTH, Chuck
Cheers, - Stu On Apr 26, 2010, at 9:40 PM, Charles C. Berry wrote:On Mon, 26 Apr 2010, Stu wrote: > Hi all,> > One subtlety is that the drop argument only works if you specify 2 or> more indices e.g. [i, j, ..., drop=F]; but not for a single index e.g > [i, drop=F]. Wrong. > a <- structure(1:5,dim=5) > dim(a) [1] 5 > dim(a[2:3,drop=F]) # don't drop regardless [1] 2 > dim(a[2,drop=F]) # dont' drop regardless [1] 1 > dim(a[2:3,drop=T]) # no extent of length 1 [1] 2 > dim(a[2,drop=T]) # drop, extent of length 1 NULL> > Why doesn't R complain about the unused "drop=F" argument in the> single index case?In the example you give (one index for a two-dimension array), vector indexing is assumed. For vector indexing, drop is irrelevant.HTH, Chuck> > Cheers,> - Stu> > a = matrix(1:10, nrow=1)> b = matrix(10:1, ncol=1)> > # a1 is an vector w/o dim attribute (i.e. drop=F is ignored silently)> (a1 = a[2:5, drop=F]) > dim(a1)> > # a2 is an vector WITH dim attribute: a row matrix (drop=F works)> (a2 = a[, 2:5, drop=F]) > dim(a2)> > # b1 is an vector w/o dim attribute (i.e. drop=F is ignored silently)> (b1 = b[2:5, drop=F]) > dim(b1)> > # b2 is an vector WITH dim attribute: a column matrix (drop=F works)> (b2 = b[2:5, , drop=F]) > dim(b2)> > > On Mar 30, 4:08 pm, lith <minil...@gmail.com> wrote: > > > Reframe the problem. Rethink why you need to keep dimensions. I never > > > ever had to use drop. > > > > The problem is that the type of the return value changes if you happen> > to forget to use drop = FALSE, which can easily turn into a nightmare:> > > > m <-matrix(1:20, ncol=4)> > for (i in seq(3, 1, -1)) { > > print(class(m[1:i, ]))}> > > > [1] "matrix"> > [1] "matrix" > > [1] "integer"> > > > ______________________________________________ > > r-h...@r-project.org mailing > > listhttps://stat.ethz.ch/mailman/listinfo/r-help > > PLEASE do read the posting > > guidehttp://www.R-project.org/posting-guide.html> > and provide commented, minimal, self-contained, reproducible code.> > ______________________________________________> R-help@r-project.org mailing list > https://stat.ethz.ch/mailman/listinfo/r-help> PLEASE do read the posting guide > http://www.R-project.org/posting-guide.html> and provide commented, minimal, self-contained, reproducible code.>Charles C. Berry (858) 534-2098Dept of Family/Preventive MedicineE mailto:cbe...@tajo.ucsd.edu UC San Diego http://famprevmed.ucsd.edu/faculty/cberry/ La Jolla, San Diego 92093-0901
Charles C. Berry (858) 534-2098
Dept of Family/Preventive Medicine
E mailto:cbe...@tajo.ucsd.edu UC San Diego
http://famprevmed.ucsd.edu/faculty/cberry/ La Jolla, San Diego 92093-0901
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
