[Rd] data frame method for as.table()

[Ernest Adrogué]

Hello,

Currently it's possible to convert an object of class table to a data frame
with as.data.frame.table(), but there's no ready-made function, AFAIK, to do
the reverse operation, i.e. conversion of a data frame to a table.

Do you think it would be a good idea to add a data.frame method to
as.table(), to allow such conversions?

The idea is that if `x' is a table and `y <- as.data.frame(x)', then the
object returned by `as.table(y)' should be equal to `x'.

Below is a proof of concept

as.table.data.frame <- function(x, ..., response) {
    if (missing(response))
        resp <- ncol(x)
    else {
        resp <- match(response[1L], names(x))
        if (is.na(resp))
            stop('not found: ', response[1L])
    }
    x[-resp] <- lapply(x[-resp], as.factor)
    if (any(do.call(table, x[-resp]) > 1L))
        stop('repeated frequency value')
    dn <- lapply(x[-resp], function(y) {
        if (any(is.na(y))) c(levels(y), NA) else levels(y)
    })
    ind <- mapply(function(val, lev) match(val, lev), x[-resp], dn)
    out <- array(dim=unlist(lapply(dn, length)), dimnames=dn)
    out[ind] <- x[[resp]]
    as.table(out)
}

and a simple usage example:

> (y <- table(foo=c('a','a',NA,'b','b'), useNA='always'))
foo
   a    b <NA> 
   2    2    1 
> (yy <- as.data.frame(y))
   foo Freq
1    a    2
2    b    2
3 <NA>    1
> as.table(yy)
foo
   a    b <NA> 
   2    2    1 
> 

Any thoughts?