tests/testthat/test-tidymodels_bridge.R

In classbound: Visualization for Classification Decision Boundaries

test_that("tidymodels_bridge creates a multi-model grid for valid engines", {
  skip_if_not_installed("parsnip")
  skip_if_not_installed("workflowsets")
  skip_if_not_installed("rpart")

  data <- palmerpenguins::penguins
  data <- na.omit(data[, c("bill_length_mm", "bill_depth_mm", "species")])

  # Test valid engine names
  models <- c("rpart")
  res <- tidymodels_bridge(
    data = data,
    response = "species",
    models = models,
    feature_range = list(bill_length_mm = c(30, 60), bill_depth_mm = c(10, 25)),
    resolution = 10
  )

  expect_s3_class(res, "classbound")
  expect_s3_class(res, "classbound_multi")
  expect_true("model" %in% colnames(res$boundary_data))
})

test_that("tidymodels_bridge rejects empty models list", {
  skip_if_not_installed("parsnip")
  skip_if_not_installed("workflowsets")

  data <- palmerpenguins::penguins
  data <- na.omit(data[, c("bill_length_mm", "bill_depth_mm", "species")])

  expect_error(
    tidymodels_bridge(data, "species", character(0)),
    "cannot be empty"
  )
})

test_that("tidymodels_bridge rejects duplicates", {
  skip_if_not_installed("parsnip")
  skip_if_not_installed("workflowsets")

  data <- palmerpenguins::penguins
  data <- na.omit(data[, c("bill_length_mm", "bill_depth_mm", "species")])

  expect_error(
    tidymodels_bridge(data, "species", c("rpart", "rpart")),
    "Duplicate models detected"
  )
})

test_that("tidymodels_bridge rejects unsupported models", {
  skip_if_not_installed("parsnip")
  skip_if_not_installed("workflowsets")

  data <- palmerpenguins::penguins
  data <- na.omit(data[, -c(2, 7, 8)])

  expect_error(
    tidymodels_bridge(data, "species", c("rpart", "fake_engine")),
    "Unsupported models requested: fake_engine"
  )
})

test_that("tidymodels_bridge auto-computes range robustly", {
  skip_if_not_installed("parsnip")
  skip_if_not_installed("workflowsets")
  skip_if_not_installed("rpart")

  # 1. Successful auto-computation
  data <- palmerpenguins::penguins
  data <- na.omit(data[, c("bill_length_mm", "bill_depth_mm", "species")])

  res <- tidymodels_bridge(data, "species", c("rpart"), resolution = 5)
  expect_true(is.list(res$boundary_data))

  # 2. Rejects > 2 predictors without explicit range
  data_3d <- na.omit(palmerpenguins::penguins[, c("bill_length_mm", "bill_depth_mm", "flipper_length_mm", "species")])
  expect_error(
    tidymodels_bridge(data_3d, "species", c("rpart")),
    "exactly 2 numeric features"
  )

  # 3. Rejects categorical predictors in auto-computation
  data_cat <- data.frame(
    x1 = as.factor(c("A", "B", "A", "B")),
    x2 = c(1, 2, 3, 4),
    y = as.factor(c("Yes", "No", "Yes", "No"))
  )
  expect_error(
    tidymodels_bridge(data_cat, "y", c("rpart")),
    "exactly 2 numeric features"
  )

  # 4. Rejects NA/Inf only data
  data_na <- data.frame(
    x1 = as.numeric(c(NA, NA, NA)),
    x2 = c(1, 2, 3),
    y = as.factor(c("Yes", "No", "Yes"))
  )
  expect_error(
    tidymodels_bridge(data_na, "y", c("rpart")),
    "Could not find numeric range"
  )
})

test_that("find_workspace_models correctly identifies relevant objects", {
  # Create a temporary environment to simulate the global environment
  test_env <- new.env()

  # Populate with relevant and irrelevant objects
  test_env$my_wf <- structure(list(), class = c("workflow", "list"))
  test_env$my_spec <- structure(list(), class = c("model_spec", "list"))
  test_env$my_fit <- structure(list(), class = c("model_fit", "list"))

  test_env$not_a_model <- data.frame(x = 1:5)
  test_env$just_a_string <- "hello"

  found <- find_workspace_models(test_env)

  expect_type(found, "character")
  expect_length(found, 3)
  expect_setequal(found, c("my_wf", "my_spec", "my_fit"))

  # Test empty environment
  empty_env <- new.env()
  found_empty <- find_workspace_models(empty_env)
  expect_length(found_empty, 0)
  expect_type(found_empty, "character")
})