recipes

Improvements

• bake() gained a stop_at argument, which takes a step number or step id and applies only the steps up to and including that step. This is useful for diagnostics such as visualizing the effect of each step. (#1551)
• The following steps are now dramatically faster when baking data with many columns, as they no longer copy the whole data set once per column: check_new_values(), check_range(), step_bin2factor(), step_BoxCox(), step_bs(), step_center(), step_cut(), step_discretize(), step_dummy(), step_factor2string(), step_harmonic(), step_holiday(), step_hyperbolic(), step_impute_bag(), step_impute_knn(), step_impute_linear(), step_impute_lower(), step_impute_mean(), step_impute_median(), step_impute_mode(), step_impute_roll(), step_integer(), step_inverse(), step_invlogit(), step_lag(), step_log(), step_logit(), step_normalize(), step_novel(), step_ns(), step_num2factor(), step_ordinalscore(), step_other(), step_percentile(), step_range(), step_relevel(), step_scale(), step_shuffle(), step_sqrt(), step_string2factor(), step_unknown(), step_unorder(), step_window(), and step_YeoJohnson(). (#1543)
• prep() is now faster on recipes with many columns or many steps. (#1543)
• step_count() and step_regex() now accepts selectors that resolve to more than one column. When more than one column is selected, the new columns are named {column}_{result}. (#1384)

Bug Fixes

• step_impute_bag(), step_impute_knn(), and step_impute_linear() no longer error with "C stack usage is too close to the limit" when the data contains a survival outcome column. (#1517, #1542)
• step_num2factor() no longer errors when more than one column is selected. (#1543)

Developer

• New recipes_map_cols() helper for use in bake() methods of steps that transform columns in place. It applies a function to the selected columns and assigns the results in bulk, avoiding the quadratic cost of assigning one column at a time. (#1543)