Merge pull request #548 from tidymodels/checks-vfold

Update input checks for `vfold_cv.R`

R/clustering.R

@@ -57,7 +57,7 @@ clustering_cv <- function(data,
                           distance_function = "dist",
                           cluster_function = c("kmeans", "hclust"),
                           ...) {
-  check_repeats(repeats)
+  check_number_whole(repeats, min = 1)
 
   if (!rlang::is_function(cluster_function)) {
     cluster_function <- rlang::arg_match(cluster_function)

R/vfold.R

@@ -72,7 +72,7 @@ vfold_cv <- function(data, v = 10, repeats = 1,
   }
 
   check_strata(strata, data)
-  check_repeats(repeats)
+  check_number_whole(repeats, min = 1)
 
   if (repeats == 1) {
     split_objs <- vfold_splits(
@@ -213,7 +213,7 @@ vfold_splits <- function(data, v = 10, strata = NULL, breaks = 4, pool = 0.1, pr
 #' @export
 group_vfold_cv <- function(data, group = NULL, v = NULL, repeats = 1, balance = c("groups", "observations"), ..., strata = NULL, pool = 0.1) {
   check_dots_empty()
-  check_repeats(repeats)
+  check_number_whole(repeats, min = 1)
   group <- validate_group({{ group }}, data)
   balance <- rlang::arg_match(balance)
 
@@ -331,23 +331,24 @@ add_vfolds <- function(x, v) {
 }
 
 check_v <- function(v, max_v, rows = "rows", prevent_loo = TRUE, call = rlang::caller_env()) {
-  if (!is.numeric(v) || length(v) != 1 || v < 2) {
-    cli_abort("{.arg v} must be a single positive integer greater than 1.", call = call)
-  } else if (v > max_v) {
+  check_number_whole(v, min = 2, call = call)
+
+  if (v > max_v) {
     cli_abort(
       "The number of {rows} is less than {.arg v} = {.val {v}}.",
       call = call
     )
-  } else if (prevent_loo && isTRUE(v == max_v)) {
+  } 
+  if (prevent_loo && isTRUE(v == max_v)) {
     cli_abort(c(
       "Leave-one-out cross-validation is not supported by this function.",
-      "x" = "You set `v` to `nrow(data)`, which would result in a leave-one-out cross-validation.",
-      "i" = "Use `loo_cv()` in this case."
+      "x" = "You set {.arg v} to {.code nrow(data)}, which would result in a leave-one-out cross-validation.",
+      "i" = "Use {.fn loo_cv} in this case."
     ), call = call)
   }
 }
 
-check_grouped_strata <- function(group, strata, pool, data) {
+check_grouped_strata <- function(group, strata, pool, data, call = caller_env()) {
 
   strata <- tidyselect::vars_select(names(data), !!enquo(strata))
 
@@ -363,14 +364,11 @@ check_grouped_strata <- function(group, strata, pool, data) {
 
   if (nrow(vctrs::vec_unique(grouped_table)) !=
       nrow(vctrs::vec_unique(grouped_table["group"]))) {
-    cli_abort("{.arg strata} must be constant across all members of each {.arg group}.")
+    cli_abort(
+      "{.field strata} must be constant across all members of each {.field group}.",
+      call = call
+    )
   }
 
   strata
 }
-
-check_repeats <- function(repeats, call = rlang::caller_env()) {
-  if (!is.numeric(repeats) || length(repeats) != 1 || repeats < 1) {
-    cli_abort("{.arg repeats} must be a single positive integer.", call = call)
-  }
-}

tests/testthat/_snaps/clustering.md

@@ -12,7 +12,7 @@
       clustering_cv(iris, Sepal.Length, v = -500)
     Condition
       Error in `clustering_cv()`:
-      ! `v` must be a single positive integer greater than 1.
+      ! `v` must be a whole number larger than or equal to 2, not the number -500.
 
 ---
 
@@ -36,23 +36,23 @@
       clustering_cv(Orange, v = 1, vars = "Tree")
     Condition
       Error in `clustering_cv()`:
-      ! `v` must be a single positive integer greater than 1.
+      ! `v` must be a whole number larger than or equal to 2, not the number 1.
 
 ---
 
     Code
       clustering_cv(Orange, repeats = 0)
     Condition
       Error in `clustering_cv()`:
-      ! `repeats` must be a single positive integer.
+      ! `repeats` must be a whole number larger than or equal to 1, not the number 0.
 
 ---
 
     Code
       clustering_cv(Orange, repeats = NULL)
     Condition
       Error in `clustering_cv()`:
-      ! `repeats` must be a single positive integer.
+      ! `repeats` must be a whole number, not `NULL`.
 
 ---
 

tests/testthat/_snaps/vfold.md

@@ -41,29 +41,29 @@
       ! strata cannot be a <Surv> object.
       i Use the time or event variable directly.
 
-# bad args
+# v arg is checked
 
     Code
       vfold_cv(iris, v = -500)
     Condition
       Error in `vfold_cv()`:
-      ! `v` must be a single positive integer greater than 1.
+      ! `v` must be a whole number larger than or equal to 2, not the number -500.
 
 ---
 
     Code
       vfold_cv(iris, v = 1)
     Condition
       Error in `vfold_cv()`:
-      ! `v` must be a single positive integer greater than 1.
+      ! `v` must be a whole number larger than or equal to 2, not the number 1.
 
 ---
 
     Code
       vfold_cv(iris, v = NULL)
     Condition
       Error in `vfold_cv()`:
-      ! `v` must be a single positive integer greater than 1.
+      ! `v` must be a whole number, not `NULL`.
 
 ---
 
@@ -76,36 +76,36 @@
 ---
 
     Code
-      vfold_cv(iris, v = 150, repeats = 2)
+      vfold_cv(mtcars, v = nrow(mtcars))
     Condition
       Error in `vfold_cv()`:
-      ! Repeated resampling when `v` is 150 would create identical resamples.
+      ! Leave-one-out cross-validation is not supported by this function.
+      x You set `v` to `nrow(data)`, which would result in a leave-one-out cross-validation.
+      i Use `loo_cv()` in this case.
 
----
+# repeats arg is checked
 
     Code
-      vfold_cv(Orange, repeats = 0)
+      vfold_cv(iris, v = 150, repeats = 2)
     Condition
       Error in `vfold_cv()`:
-      ! `repeats` must be a single positive integer.
+      ! Repeated resampling when `v` is 150 would create identical resamples.
 
 ---
 
     Code
-      vfold_cv(Orange, repeats = NULL)
+      vfold_cv(Orange, repeats = 0)
     Condition
       Error in `vfold_cv()`:
-      ! `repeats` must be a single positive integer.
+      ! `repeats` must be a whole number larger than or equal to 1, not the number 0.
 
 ---
 
     Code
-      vfold_cv(mtcars, v = nrow(mtcars))
+      vfold_cv(Orange, repeats = NULL)
     Condition
       Error in `vfold_cv()`:
-      ! Leave-one-out cross-validation is not supported by this function.
-      x You set `v` to `nrow(data)`, which would result in a leave-one-out cross-validation.
-      i Use `loo_cv()` in this case.
+      ! `repeats` must be a whole number, not `NULL`.
 
 # printing
 
@@ -191,7 +191,7 @@
       group_vfold_cv(Orange, v = 1, group = "Tree")
     Condition
       Error in `group_vfold_cv()`:
-      ! `v` must be a single positive integer greater than 1.
+      ! `v` must be a whole number larger than or equal to 2, not the number 1.
 
 # grouping -- other balance methods
 
@@ -286,6 +286,14 @@
       10 <split [96051/3949]> Resample10
       # i 20 more rows
 
+# grouping fails for strata not constant across group members
+
+    Code
+      group_vfold_cv(sample_data, group, v = 5, strata = outcome)
+    Condition
+      Error in `group_vfold_cv()`:
+      ! strata must be constant across all members of each group.
+
 # grouping -- printing
 
     Code

tests/testthat/test-vfold.R

@@ -104,7 +104,7 @@ test_that("strata arg is checked", {
   })
 })
 
-test_that("bad args", {
+test_that("v arg is checked", {
   expect_snapshot(error = TRUE, {
     vfold_cv(iris, v = -500)
   })
@@ -117,6 +117,12 @@ test_that("bad args", {
   expect_snapshot(error = TRUE, {
     vfold_cv(iris, v = 500)
   })
+  expect_snapshot(error = TRUE, {
+    vfold_cv(mtcars, v = nrow(mtcars))
+  })
+})
+
+test_that("repeats arg is checked", {
   expect_snapshot(error = TRUE, {
     vfold_cv(iris, v = 150, repeats = 2)
   })
@@ -126,9 +132,6 @@ test_that("bad args", {
   expect_snapshot(error = TRUE, {
     vfold_cv(Orange, repeats = NULL)
   })
-  expect_snapshot(error = TRUE, {
-    vfold_cv(mtcars, v = nrow(mtcars))
-  })
 })
 
 test_that("printing", {
@@ -403,6 +406,35 @@ test_that("grouping -- strata", {
   )
 })
 
+test_that("grouping fails for strata not constant across group members", {
+  set.seed(11)
+
+  n_common_class <- 70
+  n_rare_class <- 30
+
+  group_table <- tibble(
+    group = 1:100,
+    outcome = sample(c(rep(0, n_common_class), rep(1, n_rare_class)))
+  )
+  observation_table <- tibble(
+    group = sample(1:100, 1e5, replace = TRUE),
+    observation = 1:1e5
+  )
+  sample_data <- dplyr::full_join(
+    group_table,
+    observation_table,
+    by = "group",
+    multiple = "all"
+  )
+
+  # violate requirement
+  sample_data$outcome[1] <- ifelse(sample_data$outcome[1], 0, 1)
+
+  expect_snapshot(error = TRUE, {
+    group_vfold_cv(sample_data, group, v = 5, strata = outcome)
+  })
+})
+
 test_that("grouping -- repeated", {
   set.seed(11)
   rs2 <- group_vfold_cv(dat1, c, v = 3, repeats = 4)