R help - Nov 2011 - Select columns of a data.frame by name OR index in a function

Dear all,

Sometimes I have the situation where a function takes a data.frame and
an additional argument describing come columns. For greater flexibility
I want to allow for either column names or column indices. What I
usually do then is something like the following:

-------------8<-------------
f <- function(datf, cols) {
  nc <- seq_along(datf)
  cn <- colnames(datf)
  colOK <- (cols %in% nc) | (cols %in% cn)
  if (!all(colOK)) {
    badc <- paste(sQuote(cols[!colOK]), collapse = ", ")
    msg <- sprintf(ngettext(sum(!colOK),
                            "%s is not a valid column selector",
                            "%s are not valid column selectors"),
                   badc)
    stop(msg)
  }
  which((nc %in% cols) | (cn %in% cols)) # with this set of indices I
would work in the rest of the code
}

dd <- data.frame(a=1, b=1, c=1)
f(dd, 2:3) # [1] 2 3
f(dd, 1:4)  # Error in f(dd, 1:4) : '4' is not a valid column selector
f(dd, "a") # [1] 1
f(dd, c("a", "d", "e")) # Error in f(dd,
c("a", "d", "e")) : 'd', 'e'
are not valid column selectors
------------->8-------------

So my question is, whether there are smarter/better/easier/more R-like
ways of doing that?

Any input appreciated.


KR,

-Thorn

Thaler, Thorn, LAUSANNE, Applied Mathematics wrote on 11/03/2011 08:48:26 
AM:
> 
> Dear all,
> 
> Sometimes I have the situation where a function takes a data.frame and
> an additional argument describing come columns. For greater flexibility
> I want to allow for either column names or column indices. What I
> usually do then is something like the following:
> 
> -------------8<-------------
> f <- function(datf, cols) {
>   nc <- seq_along(datf)
>   cn <- colnames(datf)
>   colOK <- (cols %in% nc) | (cols %in% cn)
>   if (!all(colOK)) {
>     badc <- paste(sQuote(cols[!colOK]), collapse = ", ")
>     msg <- sprintf(ngettext(sum(!colOK),
>                             "%s is not a valid column selector",
>                             "%s are not valid column selectors"),
>                    badc)
>     stop(msg)
>   }
>   which((nc %in% cols) | (cn %in% cols)) # with this set of indices I
> would work in the rest of the code
> }
> 
> dd <- data.frame(a=1, b=1, c=1)
> f(dd, 2:3) # [1] 2 3
> f(dd, 1:4)  # Error in f(dd, 1:4) : '4' is not a valid column
selector
> f(dd, "a") # [1] 1
> f(dd, c("a", "d", "e")) # Error in f(dd,
c("a", "d", "e")) : 'd', 'e'
> are not valid column selectors
> ------------->8-------------
> 
> So my question is, whether there are smarter/better/easier/more R-like
> ways of doing that?
> 
> Any input appreciated.
> 
> 
> KR,
> 
> -Thorn
> 
Since the extract function, [, already works with either column names or 
indices, you might as well take advantage of this.  If you really want the 
indices for the selected columns, the code below will work.  If you just 
wanted the subsetted data frame, you can just output x, rather than 
matching the names.

f <- function(df, cols) {
        x <- df[, cols, drop=FALSE]
        match(names(x), names(df))
        }

Jean

	[[alternative HTML version deleted]]