R/rename_at.R

In recipes: Preprocessing and Feature Engineering Steps for Modeling

#' Rename multiple columns using dplyr
#'
#' `step_rename_at()` creates a *specification* of a recipe step that will
#' rename the selected variables using a common function via
#' [dplyr::rename_at()].
#'
#' @inheritParams step_pca
#' @inheritParams step_center
#' @param fn A function `fun`, a quosure style lambda `~ fun(.)`` or a list of
#' either form (but containing only a single function, see [dplyr::rename_at()]).
#' **Note that this argument must be named**.
#' @param inputs A vector of column names populated by [prep()].
#' @template step-return
#' @details
#'
#' # Tidying
#'
#' When you [`tidy()`][tidy.recipe()] this step, a tibble with columns
#' `terms` which contains the columns being transformed is returned.
#'
#' @template case-weights-not-supported
#'
#' @family dplyr steps
#' @export
#' @examples
#' library(dplyr)
#' recipe(~., data = iris) %>%
#'   step_rename_at(everything(), fn = ~ gsub(".", "_", ., fixed = TRUE)) %>%
#'   prep() %>%
#'   bake(new_data = NULL) %>%
#'   slice(1:10)
#' @export
step_rename_at <- function(recipe, ...,
                           fn,
                           role = "predictor",
                           trained = FALSE,
                           inputs = NULL,
                           skip = FALSE,
                           id = rand_id("rename_at")) {
  add_step(
    recipe,
    step_rename_at_new(
      terms = enquos(...),
      fn = fn,
      trained = trained,
      role = role,
      inputs = inputs,
      skip = skip,
      id = id
    )
  )
}

step_rename_at_new <-
  function(terms, fn, role, trained, inputs, skip, id) {
    step(
      subclass = "rename_at",
      terms = terms,
      fn = fn,
      role = role,
      trained = trained,
      inputs = inputs,
      skip = skip,
      id = id
    )
  }

#' @export
prep.step_rename_at <- function(x, training, info = NULL, ...) {
  col_names <- recipes_eval_select(x$terms, training, info)

  step_rename_at_new(
    terms = x$terms,
    fn = x$fn,
    trained = TRUE,
    role = x$role,
    inputs = col_names,
    skip = x$skip,
    id = x$id
  )
}

#' @export
bake.step_rename_at <- function(object, new_data, ...) {
  dplyr::rename_at(new_data, .vars = object$inputs, .funs = object$fn)
}

print.step_rename_at <-
  function(x, width = max(20, options()$width - 35), ...) {
    title <- "Variable renaming for "
    print_step(x$inputs, x$terms, x$trained, title, width)
    invisible(x)
  }

#' @rdname tidy.recipe
#' @export
tidy.step_rename_at <- function(x, ...) {
  if (is_trained(x)) {
    res <- tibble(terms = unname(x$inputs))
  } else {
    term_names <- sel2char(x$terms)
    res <- tibble(terms = term_names)
  }
  res$id <- x$id
  res
}