[R] Building ragged dataframe (was Re: Data Gaps)

[Keith Jewell]

> "Dennis Murphy" <djmuser at gmail.com> wrote in message
> news:AANLkTinrAuFrh71J9WpNfZVg_8JAYACkw1e8CPn=cZ2g at mail.gmail.com...
> Hi:
>
> The essential problem is that after you append items, the result is a list
> with possibly unequal lengths. Trying to convert that into a data frame by
> the 'usual' methods (do.call(rbind, ...) or ldply() in plyr) didn't work 
> (as
> anticipated). One approach is to initialize a maximum size matrix with NAs
> and then replace the NAs by the contents of each component of the list. 
> The
> following function is far from elegant, but the idea is to output a data
> frame by 'NA filling' the shorter vectors in the list and doing the
> equivalent of do.call(rbind, list).
>
>  listNAfill <- function(l) {
>        # input argument l is a list with numeric component vectors
>        lengths <- sapply(l, length)
>        m <- matrix(NA, nrow = length(l), ncol = max(lengths))
>        for(i in seq_len(length(l))) m[i, ] <- replace(m[i, ], 
> 1:lengths[i],
> l[[i]])
>        as.data.frame(m)
>    }
<snip>

Leaping in on a detail, and apologising in advance if my comment isn't 
relevant, I had need to build up a data frame from different length columns 
(a ragged dataframe?). In case it helps, this function is what I ended up 
with...
---------------
 # utility function, combining different length objects into a dataframe
 # padding short columns with NA
 CDF <- function(x, y)
 { out <- merge(data.frame(x), data.frame(y), all = T, by = "row.names") # 
merge
  out$Row.names <- as.integer(out$Row.names) # make row names integer
  data.frame(out[order(out$Row.names), -1], row.names = 1:length(out[[1]])) 
# sort
 }
----------------

I guess it could reasonably easily be extended to more than two arguments. 
This met my needs, and I didn't look very hard for alternatives, so there 
may be better approaches.

HTH

Keith J