R flatten out list hierarchy to matrix or data.frame

Question

I would like to flatten out a list hierarchy (similar to JSON) to a matrix or data frame. Let's say that I create the following list:

a <- list(
  b1 = list(
     c1 = list(
       d1 = data.frame()
     ),
     c2 = data.frame()
  ),
  b2 = data.frame()
)

Where each letter is another level or step down the hierarchy. Then I want a function, e.g. listToMatrix(mylist = a, steps = 2), that generates the following:

    [,1] [,2]
[1,] "b1" "c1"
[2,] "b1" "c2"
[3,] "b2" "b2"

Observe that the function's argument steps = 2 imply that it should only go 2 steps down the hierarchy. Also, if there aren't enough levels available in one direction, see b2, then it should keep the previous list name in the matrix.

Any suggestions? :)

Answer 1

Here is a solution. So it reads easily here, I have broken the code into two parts. Later, you can easily merge the two parts into a single function.

First, a function that gets a matrix of all the names, using recursion:

anames <- function(x) {

   require(plyr)
   if (is.data.frame(x)) return(NA)

   y <- do.call(rbind.fill.matrix,
                mapply(cbind, names(x), lapply(x, anames),
                       SIMPLIFY = FALSE))
   colnames(y) <- NULL

   return(y)
}

anames(a)
#      [,1] [,2] [,3] [,4]
# [1,] "b1" "c1" "d1" NA  
# [2,] "b1" "c2" NA   NA  
# [3,] "b2" NA   NA   NA

Then, a function that applies the given steps input, and fills the NAs like you requested:

listToMatrix <- function(myList, steps = Inf) {

   a <- anames(myList)
   steps <- min(steps, ncol(a) - 1)
   cols.idx <- seq_len(steps)
   a <- a[, cols.idx]
   for (j in tail(cols.idx, -1))
      a[, j] <- ifelse(is.na(a[, j]), a[, j - 1], a[, j])

   return(a)
}

listToMatrix(a, 2)
#      [,1] [,2]
# [1,] "b1" "c1"
# [2,] "b1" "c2"
# [3,] "b2" "b2"