On Mon, Aug 24, 2009 at 5:36 PM, Peter Dalgaard <[email protected]>wrote:
>
> The documentation has:
>
> >>
> The internal equivalent of the default method of as.character is performed
> on x (so there is no method dispatch). If you want to operate on non-vector
> objects passing them through deparse first will be required.
> <<
>
> and (notice the no method dispatch bit)
>
> > as.character(unclass(factor(LETTERS)))
> [1] "1" "2" "3" "4" "5" "6" "7" "8" "9" "10" "11" "12" "13" "14"
> "15"
> [16] "16" "17" "18" "19" "20" "21" "22" "23" "24" "25" "26"
>
> So, the documentation does appear to be in sync with the code (useful or
> not).
May I suggest that the documentation be made more explicit? For example, "x
is converted to a character vector using as.character(unclass(...))". That
eliminates the reference to internals, which is inappropriate in user
documentation.
-s
PS What *is* the rationale for this? Other functions which expect character
inputs seem to do as.character, not as.character/unclass for factors:
grep( 'x', factor('x')) => 1
paste(factor('x')) => "x"
tolower(factor("x")) => "x"
charmatch('x',factor('x')) => 1
factor('x',levels=factor('x')) => x / Levels: x
format(factor('x')) => 'x'
[[alternative HTML version deleted]]
______________________________________________
[email protected] mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
