quantile -> quantile_levels for #1203

NEWS.md

@@ -11,6 +11,9 @@
 
 * New `extract_fit_time()` method has been added that returns the time it took to train the model (#853).
 
+## Breaking Change
+
+* For quantile prediction, the `predict()` argument has been changed from `quantile` to `quantile_levels` for consistency. This does not affect models with mode `"quantile regression"`. 
 
 # parsnip 1.2.1
 

R/predict.R

@@ -344,7 +344,7 @@ check_pred_type_dots <- function(object, type, ..., call = rlang::caller_env())
 
  # ----------------------------------------------------------------------------
 
- other_args <- c("interval", "level", "std_error", "quantile",
+ other_args <- c("interval", "level", "std_error", "quantile_levels",
  "time", "eval_time", "increasing")
  is_pred_arg <- names(the_dots) %in% other_args
  if (any(!is_pred_arg)) {

R/predict_quantile.R

@@ -1,14 +1,13 @@
 #' @keywords internal
 #' @rdname other_predict
-#' @param quantile A vector of numbers between 0 and 1 for the quantile being
-#' predicted.
+#' @param quantile_levels A vector of values between zero and one.
 #' @inheritParams predict.model_fit
 #' @method predict_quantile model_fit
 #' @export predict_quantile.model_fit
 #' @export
 predict_quantile.model_fit <- function(object,
  new_data,
- quantile = (1:9)/10,
+ quantile_levels = NULL,
  interval = "none",
  level = 0.95,
  ...) {
@@ -20,15 +19,27 @@ predict_quantile.model_fit <- function(object,
  return(NULL)
  }
 
+ if (object$spec$mode != "quantile regression") {
+ if (is.null(quantile_levels)) {
+ quantile_levels <- (1:9)/10
+ }
+ hardhat::check_quantile_levels(quantile_levels)
+ # Pass some extra arguments to be used in post-processor
+ object$quantile_levels <- quantile_levels
+ } else {
+ if (!is.null(quantile_levels)) {
+ cli::cli_abort("{.arg quantile_levels} are specified by {.fn set_mode}
+ when the mode is {.val quantile regression}.")
+ }
+ }
+
  new_data <- prepare_data(object, new_data)
 
  # preprocess data
  if (!is.null(object$spec$method$pred$quantile$pre)) {
  new_data <- object$spec$method$pred$quantile$pre(new_data, object)
  }
 
- # Pass some extra arguments to be used in post-processor
- object$spec$method$pred$quantile$args$p <- quantile
  pred_call <- make_pred_call(object$spec$method$pred$quantile)
 
  res <- eval_tidy(pred_call)

tests/testthat/test-linear_reg_quantreg.R

@@ -83,6 +83,11 @@ test_that('linear quantile regression via quantreg - multiple quantiles', {
  expect_named(ten_quant_df, c(".pred_quantile", ".quantile_levels", ".row"))
  expect_true(nrow(ten_quant_df) == nrow(sac_test) * 10)
 
+ expect_snapshot(
+ ten_quant_pred <- predict(ten_quant, new_data = sac_test),
+ error = TRUE
+ )
+
  ###
 
  ten_quant_one_row <- predict(ten_quant, new_data = sac_test[1,])