split_on_dim {listarrays}

Split an array along a dimension

Description

Split an array along a dimension

Usage

split_on_dim(
  X,
  which_dim,
  f = dimnames(X)[[which_dim]],
  drop = FALSE,
  depth = Inf
)

split_on_rows(X, f = rownames(X), drop = FALSE, depth = Inf)

split_on_cols(X, f = rownames(X), drop = FALSE, depth = Inf)

split_along_dim(X, which_dim, depth = Inf)

split_along_rows(X, depth = Inf)

split_along_cols(X, depth = Inf)

Arguments

X

an array, or list of arrays. An atomic vector without a dimension attribute is treated as a 1 dimensional array (Meaning, atomic vectors without a dim attribute are only accepted if which_dim is 1. Names of the passed list are preserved. If a list of arrays, all the arrays must have the same length of the dimension being split.

which_dim

a scalar string or integer, specifying which dimension to split along. Negative integers count from the back. If a string, it must refer to a named dimension (e.g, one of names(dimnames(X)).

f

Specify how to split the dimension.

character, integer, factor

passed on to base::split(). Must be the same length as the dimension being split.

a list of vectors

Passed on to base::interaction() then base::split(). Each vector in the list must be the same length as the dimension being split.

a scalar integer

used to split into that many groups of equal size

a numeric vector where all(f<0)

specifies the relative size proportions of the groups being split. sum(f) must be 1. For example c(0.2, 0.2, 0.6) will return approximately a 20\ split.

drop

passed on to [.

depth

Scalar number, how many levels to recurse down. Set this if you want to explicitly treat a list as a vector (that is, a one-dimensional array). (You can alternatively set dim attributes with ⁠dim<-⁠ on the list to prevent recursion)

split_along_dim(X, which_dim) is equivalent to split_on_dim(X, which_dim, seq_along_dim(X, which_dim)).

Value

A list of arrays, or if a list of arrays was passed in, then a list of lists of arrays.

Examples

X <- array(1:8, c(2,3,4))
X
split_along_dim(X, 2)

# specify f as a factor, akin to base::split()
split_on_dim(X, 2, c("a", "a", "b"), drop = FALSE)

d <- c(10, 3, 3)
X <- array(1:prod(d), d)
y <- letters[1:10]
Y <- onehot(y)

# specify `f`` as relative partition sizes
if(require(zeallot) && require(magrittr) && require(purrr)) {

c(train, validate, test) %<-% {
  list(X = X, Y = Y, y = y) %>%
    shuffle_rows() %>%
    split_on_rows(c(0.6, 0.2, 0.2)) %>%
    transpose()
}

str(test)
str(train)
str(validate)

}


# with with array data in a data frame by splitting row-wise
if(require(tibble))
  tibble(y, X = split_along_rows(X))