Add map_*_matrix functions similar to vapply with FUN.VALUE of length > 1.

[AshesITR]

Sometimes I need to compute a matrix column-wise using a function.
Base-R code to achieve that would be vapply() with a FUN.VALUE of length greater than 1.

For a toy example with an easier solution consider this function:

rnorm_mat <- function(n, means = 0:5, sd = 1) {
  vapply(
    means,
    function(mu) rnorm(n, mean = mu, sd = sd),
    numeric(n)
  )
}

If n > 1, this works as expected and returns a matrix of normal random variates with true column means means and constant standard deviation sd. However, the return type is not stable for n = 1 in which case the value will be a length length(means) vector instead of a 1-row matrix.

I propose a family of functions map_*_matrix() which always return a matrix and behaviour akin to

map_dbl_matrix <- function(.x, .f, .n, ...) {
  .f <- purrr::as_mapper(.f, ...)
  res <- vapply(.x, .f, numeric(.n), ...)
  dim(res) <- c(.n, length(.x))
  res
}

Maybe additional convenience arguments such as .byrow = FALSE to decide whether to create the matrix rowwise or column-wise are necessary as well.

If there is interest in such a feature, I'll gladly try to get a PR started, so please let me know.

[hadley]

I think this is out of scope for purrr, since we don't currently have functions that return matrices and I don't think we want to increase the API size so much.