A toolbox for working with R arrays in a functional programming style. Flexibly split, bind, reshape, modify, subset, and name arrays.

The package provides:

  • split_on_dim() and split_along_dim() which take an array and return a list.

  • bind_on_dim() and bind_as_dim() take a list and return an array.

  • modify_along_dim() takes an array, calls the passed function .f() on each subset of the specified dimension, and returns an array of the same shape. (think of this as a safer and sometimes faster alternative to base::apply() that is guaranteed to return an array of the same shape as it received)

  • extract_dim() a wrapper around [ that allows you to specify the dimension being subset as a function argument. For example, extract_dim(X, 1, idx) will extract idx on the first dimension, regardless how many dimensions are in the array X. Contrast this with the base alternative X[idx,,], where you have to match the number of commas , to the number of dimensions in X.

  • Many of the functions have two variants *_rows() and *_cols() for the two most common case of the first and last dimension. For example split_on_rows() which is equivalent to split_on_dim(X, 1) and split_on_cols() which is equivalent to split_on_dim(X, -1)

  • set_dim() and set_dimnames(), pipe-friendly and more flexible versions of dim<- and dimnames<-

  • dim2()<-, set_dim2(), array2(), which reshape or fills arrays using row-major (C-style) semantics

  • A handful of lower-level helpers that abstract out patterns commonly encountered while working with arrays, for example expand_dims() (the inverse of base::drop(), or seq_along_rows() (a combination of seq_along() and nrow()).

  • A set of functions that help encode atomic vectors as onehot() binary matrix’s and decode_onehot() back into atomic vectors. (for example if training a neural network with keras)

  • Many of the functions work recursively if provided a list of arrays.


You can install listarrays from CRAN with:


Or install the development version from github with: