R/generic-compute-dim.R

In DavisVaughan/nodegraph: Infrastructure of node based matrix graphs

compute_common_dim <- function(operation, ...) {
  delay_args <- list(...)
  common_type <- reduce(delay_args, vec_type2)
  compute_dim_engine(common_type, operation, ...)
}

#' @export
compute_dim_engine <- function(type, operation, ...) {
  UseMethod("compute_dim_engine")
}

# Default uses R semantics

#' @export
compute_dim_engine.delay_array <- function(type, operation, ...) {
  delay_args <- list(...)
  nodes <- map(delay_args, get_node)
  dim_lst <- map_R6(nodes, get_dim)

  # the only one this doesn't apply for is %*% with base R
  # that requires dim2[2]==dim1[1] for matrices
  ok <- map2_lgl(dim_lst[1], dim_lst, identical)

  # Special case scalars
  scalar <- map_int(dim_lst, vec_dims) == 1L
  ok <- ok | scalar

  if (!all(ok)) {
    abort("All arguments must have the same dimensions or be scalars.")
  }

  dim_lst[[1]]
}