R/range_check.R

In recipes: Preprocessing and Feature Engineering Steps for Modeling

#' Check Range Consistency
#'
#' `check_range` creates a *specification* of a recipe
#'  check that will check if the range of a numeric
#'  variable changed in the new data.
#'
#' @inheritParams check_missing
#' @param slack_prop The allowed slack as a proportion of the range
#'   of the variable in the train set.
#' @param warn If `TRUE` the check will throw a warning instead
#'   of an error when failing.
#' @param lower A named numeric vector of minimum values in the train set.
#'   This is `NULL` until computed by [prep()].
#' @param upper A named numeric vector of maximum values in the train set.
#'   This is `NULL` until computed by [prep()].
#' @template check-return
#' @family checks
#' @export
#' @details
#'   The amount of slack that is allowed is determined by the
#'   `slack_prop`. This is a numeric of length one or two. If
#'   of length one, the same proportion will be used at both ends
#'   of the train set range. If of length two, its first value
#'   is used to compute the allowed slack at the lower end,
#'   the second to compute the allowed slack at the upper end.
#'
#'  # Tidying
#'
#'  When you [`tidy()`][tidy.recipe()] this check, a tibble with columns
#'  `terms` (the selectors or variables selected) and `value` (the means)
#'  is returned.
#'
#' @examples
#' slack_df <- data_frame(x = 0:100)
#' slack_new_data <- data_frame(x = -10:110)
#'
#' # this will fail the check both ends
#' \dontrun{
#' recipe(slack_df) %>%
#'   check_range(x) %>%
#'   prep() %>%
#'   bake(slack_new_data)
#' }
#'
#' # this will fail the check only at the upper end
#' \dontrun{
#' recipe(slack_df) %>%
#'   check_range(x, slack_prop = c(0.1, 0.05)) %>%
#'   prep() %>%
#'   bake(slack_new_data)
#' }
#'
#' # give a warning instead of an error
#' \dontrun{
#' recipe(slack_df) %>%
#'   check_range(x, warn = TRUE) %>%
#'   prep() %>%
#'   bake(slack_new_data)
#' }
check_range <-
  function(recipe,
           ...,
           role = NA,
           skip = FALSE,
           trained = FALSE,
           slack_prop = 0.05,
           warn = FALSE,
           lower = NULL,
           upper = NULL,
           id = rand_id("range_check_")) {
    add_check(
      recipe,
      check_range_new(
        terms = enquos(...),
        role = role,
        skip = skip,
        trained = trained,
        warn = warn,
        lower = lower,
        upper = upper,
        slack_prop = slack_prop,
        id = id
      )
    )
  }

## Initializes a new object
check_range_new <-
  function(terms, role, skip, trained, slack_prop, warn, lower, upper, id) {
    check(
      subclass = "range",
      terms = terms,
      role = role,
      skip = skip,
      trained = trained,
      warn = warn,
      lower = lower,
      upper = upper,
      slack_prop = slack_prop,
      id = id
    )
  }

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

  ## TODO add informative error for nonnumerics

  lower_vals <- vapply(training[, col_names], min, c(min = 1),
    na.rm = TRUE
  )
  upper_vals <- vapply(training[, col_names], max, c(max = 1),
    na.rm = TRUE
  )
  check_range_new(
    terms      = x$terms,
    role       = x$role,
    trained    = TRUE,
    skip       = x$skip,
    warn       = x$warn,
    lower      = lower_vals,
    upper      = upper_vals,
    slack_prop = x$slack_prop,
    id         = x$id
  )
}

range_check_func <- function(x,
                             lower,
                             upper,
                             slack_prop = 0.05,
                             warn = FALSE,
                             colname = "x") {
  stopifnot(
    is.numeric(slack_prop),
    is.numeric(x)
  )
  min_x <- min(x)
  max_x <- max(x)
  msg <- NULL
  if (length(slack_prop) == 1) {
    lower_allowed <- lower - ((upper - lower) * slack_prop)
    upper_allowed <- upper + ((upper - lower) * slack_prop)
  } else if (length(slack_prop) == 2) {
    lower_allowed <- lower - ((upper - lower) * slack_prop[1])
    upper_allowed <- upper + ((upper - lower) * slack_prop[2])
  } else {
    rlang::abort("slack_prop should be of length 1 or of length 2")
  }

  if (min_x < lower_allowed & max_x > upper_allowed) {
    msg <- paste0(
      "min ", colname, " is ", min_x, ", lower bound is ",
      lower_allowed, ", max x is ", max_x, ", upper bound is ",
      upper_allowed
    )
  } else if (min_x < lower_allowed) {
    msg <- paste0(
      "min ", colname, " is ", min_x, ", lower bound is ",
      lower_allowed
    )
  } else if (max_x > upper_allowed) {
    msg <- paste0(
      "max ", colname, " is ", max_x, ", upper bound is ",
      upper_allowed
    )
  }
  if (warn & !is.null(msg)) {
    rlang::warn(msg)
  } else if (!is.null(msg)) {
    rlang::abort(msg)
  }
}

#' @export
bake.check_range <- function(object, new_data, ...) {
  col_names <- names(object$lower)
  check_new_data(col_names, object, new_data)

  for (col_name in col_names) {
    range_check_func(
      new_data[[col_name]],
      object$lower[col_name],
      object$upper[col_name],
      object$slack_prop,
      object$warn,
      col_name
    )
  }
  new_data
}

print.check_range <-
  function(x, width = max(20, options()$width - 30), ...) {
    title <- "Checking range of "
    print_step(names(x$lower), x$terms, x$trained, title, width)
    invisible(x)
  }


#' @rdname tidy.recipe
#' @export
tidy.check_range <- function(x, ...) {
  if (is_trained(x)) {
    res <- tibble(terms = names(x$lower))
  } else {
    res <- tibble(terms = sel2char(x$terms))
  }
  res$id <- x$id
  res
}