Merge pull request #277 from tidymodels/more-error-calls

Sending the call to `check_data_frame_or_matrix()`
tidymodels · Jan 23, 2025 · 7aaa32f · 7aaa32f
2 parents b1fe750 + 5e47856
commit 7aaa32f
Show file tree

Hide file tree

Showing 22 changed files with 195 additions and 93 deletions.
diff --git a/R/blueprint-formula-default.R b/R/blueprint-formula-default.R
@@ -537,7 +537,7 @@ mold_formula_default_process_predictors <- function(blueprint, data, ..., call =
   original_names <- get_all_predictors(formula, data, call = call)
   data <- data[original_names]
 
-  ptype <- extract_ptype(data)
+  ptype <- extract_ptype(data, call = call)
 
   if (identical(blueprint$indicators, "traditional") ||
       identical(blueprint$indicators, "one_hot")) {
@@ -566,18 +566,20 @@ mold_formula_default_process_predictors <- function(blueprint, data, ..., call =
     data <- mask_factorish_in_data(data, factorish_names)
   }
 
-  framed <- model_frame(formula, data)
+  framed <- model_frame(formula, data, call = call)
   offset <- extract_offset(framed$terms, framed$data, call = call)
 
   if (identical(blueprint$indicators, "one_hot")) {
     predictors <- model_matrix_one_hot(
       terms = framed$terms,
-      data = framed$data
+      data = framed$data,
+      call = call
     )
   } else {
     predictors <- model_matrix(
       terms = framed$terms,
-      data = framed$data
+      data = framed$data,
+      call = call
     )
   }
 
@@ -609,14 +611,14 @@ mold_formula_default_process_outcomes <- function(blueprint, data, ..., call = c
   original_names <- get_all_outcomes(formula, data, call = call)
   data <- data[original_names]
 
-  ptype <- extract_ptype(data)
+  ptype <- extract_ptype(data, call = call)
 
   formula <- get_outcomes_formula(formula)
 
   # used on the `~ LHS` formula
   check_no_interactions(formula, error_call = call)
 
-  framed <- model_frame(formula, data)
+  framed <- model_frame(formula, data, call = call)
 
   outcomes <- flatten_embedded_columns(framed$data)
 
@@ -670,7 +672,7 @@ run_forge.default_formula_blueprint <- function(blueprint,
 
 forge_formula_default_clean <- function(blueprint, new_data, outcomes, ..., call = caller_env()) {
   check_dots_empty0(...)
-  check_data_frame_or_matrix(new_data)
+  check_data_frame_or_matrix(new_data, call = call)
   new_data <- coerce_to_tibble(new_data)
   check_unique_column_names(new_data)
   check_bool(outcomes)
@@ -691,13 +693,14 @@ forge_formula_default_clean <- function(blueprint, new_data, outcomes, ..., call
   predictors <- scream(
     predictors,
     predictors_ptype,
-    allow_novel_levels = blueprint$allow_novel_levels
+    allow_novel_levels = blueprint$allow_novel_levels,
+    call = call
   )
 
   if (outcomes) {
     outcomes <- shrink(new_data, blueprint$ptypes$outcomes, call = call)
     # Never allow novel levels for outcomes
-    outcomes <- scream(outcomes, blueprint$ptypes$outcomes)
+    outcomes <- scream(outcomes, blueprint$ptypes$outcomes, call = call)
   } else {
     outcomes <- NULL
   }
@@ -722,7 +725,8 @@ forge_formula_default_process <- function(blueprint, predictors, outcomes, extra
 
   processed <- forge_formula_default_process_outcomes(
     blueprint = blueprint,
-    outcomes = outcomes
+    outcomes = outcomes,
+    call = call
   )
 
   blueprint <- processed$blueprint
@@ -749,17 +753,19 @@ forge_formula_default_process_predictors <- function(blueprint, predictors, ...,
     predictors <- mask_factorish_in_data(predictors, factorish_names)
   }
 
-  framed <- model_frame(terms, predictors)
+  framed <- model_frame(terms, predictors, call = call)
 
   if (identical(blueprint$indicators, "one_hot")) {
     data <- model_matrix_one_hot(
       terms = framed$terms,
-      data = framed$data
+      data = framed$data,
+      call = call
     )
   } else {
     data <- model_matrix(
       terms = framed$terms,
-      data = framed$data
+      data = framed$data,
+      call = call
     )
   }
 
@@ -780,7 +786,8 @@ forge_formula_default_process_predictors <- function(blueprint, predictors, ...,
   )
 }
 
-forge_formula_default_process_outcomes <- function(blueprint, outcomes) {
+forge_formula_default_process_outcomes <- function(blueprint, outcomes, ..., call = caller_env()) {
+  check_dots_empty0(...)
 
   # no outcomes to process
   if (is.null(outcomes)) {
@@ -794,7 +801,7 @@ forge_formula_default_process_outcomes <- function(blueprint, outcomes) {
   terms <- blueprint$terms$outcomes
   terms <- alter_terms_environment(terms)
 
-  framed <- model_frame(terms, outcomes)
+  framed <- model_frame(terms, outcomes, call = call)
 
   # Because model.matrix() does this for the RHS and we want
   # to be consistent even though we are only going through

diff --git a/R/blueprint-recipe-default.R b/R/blueprint-recipe-default.R
@@ -243,14 +243,14 @@ mold_recipe_default_process <- function(blueprint, data, ..., call = caller_env(
   predictors_ptype <- processed$ptype
   predictors_extras <- processed$extras
 
-  processed <- mold_recipe_default_process_outcomes(blueprint = blueprint, data = data)
+  processed <- mold_recipe_default_process_outcomes(blueprint = blueprint, data = data, call = call)
 
   blueprint <- processed$blueprint
   outcomes <- processed$data
   outcomes_ptype <- processed$ptype
   outcomes_extras <- processed$extras
 
-  processed <- mold_recipe_default_process_extras(blueprint, data)
+  processed <- mold_recipe_default_process_extras(blueprint, data, call = call)
 
   blueprint <- processed$blueprint
   extras <- processed$extras
@@ -277,11 +277,11 @@ mold_recipe_default_process_predictors <- function(blueprint, data, ..., call =
 
   predictors <- recipes::juice(blueprint$recipe, all_predictors())
 
-  predictors <- maybe_add_intercept_column(predictors, blueprint$intercept)
+  predictors <- maybe_add_intercept_column(predictors, blueprint$intercept, call = call)
 
   predictors <- recompose(predictors, composition = blueprint$composition, call = call)
 
-  ptype <- get_original_predictor_ptype(blueprint$recipe, data)
+  ptype <- get_original_predictor_ptype(blueprint$recipe, data, call = call)
 
   new_mold_process_terms(
     blueprint = blueprint,
@@ -290,12 +290,14 @@ mold_recipe_default_process_predictors <- function(blueprint, data, ..., call =
   )
 }
 
-mold_recipe_default_process_outcomes <- function(blueprint, data) {
+mold_recipe_default_process_outcomes <- function(blueprint, data, ..., call = caller_env()) {
+  check_dots_empty0(...)
+
   all_outcomes <- recipes::all_outcomes
 
   outcomes <- recipes::juice(blueprint$recipe, all_outcomes())
 
-  ptype <- get_original_outcome_ptype(blueprint$recipe, data)
+  ptype <- get_original_outcome_ptype(blueprint$recipe, data, call = call)
 
   new_mold_process_terms(
     blueprint = blueprint,
@@ -304,7 +306,8 @@ mold_recipe_default_process_outcomes <- function(blueprint, data) {
   )
 }
 
-mold_recipe_default_process_extras <- function(blueprint, data) {
+mold_recipe_default_process_extras <- function(blueprint, data, ..., call = caller_env()) {
+  check_dots_empty0(...)
 
   # Capture original non standard role columns that exist in `data` and are also
   # required by the `recipe$requirements$bake` requirement. These columns are
@@ -315,7 +318,7 @@ mold_recipe_default_process_extras <- function(blueprint, data) {
   )
 
   if (!is.null(original_extra_role_cols)) {
-    original_extra_role_ptypes <- lapply(original_extra_role_cols, extract_ptype)
+    original_extra_role_ptypes <- lapply(original_extra_role_cols, extract_ptype, call = call)
 
     blueprint <- update_blueprint0(
       blueprint,
@@ -376,7 +379,7 @@ run_forge.default_recipe_blueprint <- function(blueprint,
 
 forge_recipe_default_clean <- function(blueprint, new_data, outcomes, ..., call = caller_env()) {
   check_dots_empty0(...)
-  check_data_frame_or_matrix(new_data)
+  check_data_frame_or_matrix(new_data, call = call)
   new_data <- coerce_to_tibble(new_data)
   check_unique_column_names(new_data)
   check_bool(outcomes)
@@ -386,13 +389,14 @@ forge_recipe_default_clean <- function(blueprint, new_data, outcomes, ..., call
   predictors <- scream(
     predictors,
     blueprint$ptypes$predictors,
-    allow_novel_levels = blueprint$allow_novel_levels
+    allow_novel_levels = blueprint$allow_novel_levels,
+    call = call
   )
 
   if (outcomes) {
     outcomes <- shrink(new_data, blueprint$ptypes$outcomes, call = call)
     # Never allow novel levels for outcomes
-    outcomes <- scream(outcomes, blueprint$ptypes$outcomes)
+    outcomes <- scream(outcomes, blueprint$ptypes$outcomes, call = call)
   } else {
     outcomes <- NULL
   }
@@ -420,7 +424,8 @@ forge_recipe_default_clean_extras <- function(blueprint, new_data, ..., call = c
     extra_role_cols,
     blueprint$extra_role_ptypes,
     scream,
-    allow_novel_levels = blueprint$allow_novel_levels
+    allow_novel_levels = blueprint$allow_novel_levels,
+    call = call
   )
 
   extras <- list(roles = extra_role_cols)
@@ -500,7 +505,7 @@ forge_recipe_default_process <- function(blueprint, predictors, outcomes, extras
 forge_recipe_default_process_predictors <- function(blueprint, predictors, ..., call = caller_env()) {
   check_dots_empty0(...)
 
-  predictors <- maybe_add_intercept_column(predictors, blueprint$intercept)
+  predictors <- maybe_add_intercept_column(predictors, blueprint$intercept, call = call)
 
   predictors <- recompose(predictors, composition = blueprint$composition, call = call)
 
@@ -553,7 +558,9 @@ forge_recipe_default_process_extras <- function(extras,
 
 # ------------------------------------------------------------------------------
 
-get_original_predictor_ptype <- function(rec, data) {
+get_original_predictor_ptype <- function(rec, data, ..., call = caller_env()) {
+  check_dots_empty0(...)
+
   roles <- rec$var_info$role
   roles <- chr_explicit_na(roles)
 
@@ -562,18 +569,20 @@ get_original_predictor_ptype <- function(rec, data) {
 
   data <- data[original_names]
 
-  extract_ptype(data)
+  extract_ptype(data, call = call)
 }
 
-get_original_outcome_ptype <- function(rec, data) {
+get_original_outcome_ptype <- function(rec, data, ..., call = caller_env()) {
+  check_dots_empty0(...)
+
   roles <- rec$var_info$role
   roles <- chr_explicit_na(roles)
 
   original_names <- rec$var_info$variable[roles == "outcome"]
 
   data <- data[original_names]
 
-  extract_ptype(data)
+  extract_ptype(data, call = call)
 }
 
 get_extra_role_columns_original <- function(rec, data) {

diff --git a/R/blueprint-xy-default.R b/R/blueprint-xy-default.R
@@ -180,7 +180,7 @@ refresh_blueprint.default_xy_blueprint <- function(blueprint) {
 run_mold.default_xy_blueprint <- function(blueprint, ..., x, y, call = caller_env()) {
   check_dots_empty0(...)
 
-  cleaned <- mold_xy_default_clean(blueprint = blueprint, x = x, y = y)
+  cleaned <- mold_xy_default_clean(blueprint = blueprint, x = x, y = y, call = call)
 
   blueprint <- cleaned$blueprint
   x <- cleaned$x
@@ -192,8 +192,10 @@ run_mold.default_xy_blueprint <- function(blueprint, ..., x, y, call = caller_en
 # ------------------------------------------------------------------------------
 # mold - xy - clean
 
-mold_xy_default_clean <- function(blueprint, x, y) {
-  cleaned <- mold_xy_default_clean_predictors(blueprint, x)
+mold_xy_default_clean <- function(blueprint, x, y, ..., call = caller_env()) {
+  check_dots_empty0(...)
+
+  cleaned <- mold_xy_default_clean_predictors(blueprint, x, call = call)
 
   blueprint <- cleaned$blueprint
   x <- cleaned$x
@@ -211,8 +213,9 @@ mold_xy_default_clean <- function(blueprint, x, y) {
   new_mold_clean_xy(blueprint, x, y)
 }
 
-mold_xy_default_clean_predictors <- function(blueprint, x) {
-  check_data_frame_or_matrix(x)
+mold_xy_default_clean_predictors <- function(blueprint, x, ..., call = caller_env()) {
+  check_dots_empty0(...)
+  check_data_frame_or_matrix(x, call = call)
   x <- coerce_to_tibble(x)
   list(blueprint = blueprint, x = x)
 }
@@ -235,7 +238,7 @@ mold_xy_default_process <- function(blueprint, x, y, ..., call = caller_env()) {
   predictors_ptype <- processed$ptype
   predictors_extras <- processed$extras
 
-  processed <- mold_xy_default_process_outcomes(blueprint, y)
+  processed <- mold_xy_default_process_outcomes(blueprint, y, call = call)
 
   blueprint <- processed$blueprint
   outcomes <- processed$data
@@ -254,9 +257,9 @@ mold_xy_default_process_predictors <- function(blueprint, x, ..., call = caller_
   check_dots_empty0(...)
 
   # Important! Collect ptype before adding intercept!
-  ptype <- extract_ptype(x)
+  ptype <- extract_ptype(x, call = call)
 
-  x <- maybe_add_intercept_column(x, blueprint$intercept)
+  x <- maybe_add_intercept_column(x, blueprint$intercept, call = call)
 
   x <- recompose(x, composition = blueprint$composition, call = call)
 
@@ -267,8 +270,10 @@ mold_xy_default_process_predictors <- function(blueprint, x, ..., call = caller_
   )
 }
 
-mold_xy_default_process_outcomes <- function(blueprint, y) {
-  ptype <- extract_ptype(y)
+mold_xy_default_process_outcomes <- function(blueprint, y, ..., call = caller_env()) {
+  check_dots_empty0(...)
+
+  ptype <- extract_ptype(y, call = call)
 
   new_mold_process_terms(
     blueprint = blueprint,
@@ -314,7 +319,7 @@ run_forge.default_xy_blueprint <- function(blueprint,
 
 forge_xy_default_clean <- function(blueprint, new_data, outcomes, ..., call = caller_env()) {
   check_dots_empty0(...)
-  check_data_frame_or_matrix(new_data)
+  check_data_frame_or_matrix(new_data, call = call)
   new_data <- coerce_to_tibble(new_data)
   check_unique_column_names(new_data)
   check_bool(outcomes)
@@ -324,13 +329,14 @@ forge_xy_default_clean <- function(blueprint, new_data, outcomes, ..., call = ca
   predictors <- scream(
     predictors,
     blueprint$ptypes$predictors,
-    allow_novel_levels = blueprint$allow_novel_levels
+    allow_novel_levels = blueprint$allow_novel_levels,
+    call = call
   )
 
   if (outcomes) {
     outcomes <- shrink(new_data, blueprint$ptypes$outcomes, call = call)
     # Never allow novel levels for outcomes
-    outcomes <- scream(outcomes, blueprint$ptypes$outcomes)
+    outcomes <- scream(outcomes, blueprint$ptypes$outcomes, call = call)
   } else {
     outcomes <- NULL
   }
@@ -366,7 +372,7 @@ forge_xy_default_process <- function(blueprint, predictors, outcomes, extras, ..
 forge_xy_default_process_predictors <- function(blueprint, predictors, ..., call = caller_env()) {
   check_dots_empty0(...)
 
-  predictors <- maybe_add_intercept_column(predictors, blueprint$intercept)
+  predictors <- maybe_add_intercept_column(predictors, blueprint$intercept, call = call)
 
   predictors <- recompose(predictors, composition = blueprint$composition, call = call)
 

diff --git a/R/classes.R b/R/classes.R
@@ -5,6 +5,8 @@
 #' `get_data_classes()` extracts the classes from the original training data.
 #'
 #' @param data A data frame or matrix.
+#' 
+#' @inheritParams validate_column_names
 #'
 #' @return
 #'
@@ -24,8 +26,9 @@
 #'
 #' get_data_classes(data)
 #' @export
-get_data_classes <- function(data) {
-  data <- extract_ptype(data)
-  check_unique_column_names(data)
+get_data_classes <- function(data, ..., call = current_env()) {
+  check_dots_empty0(...)
+  data <- extract_ptype(data, call = call)
+  check_unique_column_names(data, call = call)
   lapply(data, class)
 }