library(parsnip)
library(recipes)
library(hardhat)

data(two_class_dat, package = "modeldata")

xgb <- boost_tree(trees = 3) %>% set_mode("classification")
rec <-
   recipe(Class ~ A + B, two_class_dat) %>%
   step_normalize(A) %>%
   step_normalize(B)

sparse_bp <- default_recipe_blueprint(composition = "dgCMatrix")


test_that('update model', {
   expect_error(
      new_set <- update_workflow_model(two_class_res, "none_cart", spec = xgb),
      regexp = NA
   )
   expect_true(
      inherits(extract_spec_parsnip(new_set, id = "none_cart"),
               "boost_tree")
   )
   expect_equal(new_set$result[[1]], list())

   expect_error(
      new_new_set <-
         update_workflow_model(new_set,
                               "none_glm",
                               spec = xgb,
                               formula = Class ~ log(A) + B),
      regexp = NA
   )
   new_wflow <- extract_workflow(new_new_set, "none_glm")
   expect_equal(
      new_wflow$fit$actions$model$formula,
      Class ~ log(A) + B
   )

})

test_that('update recipe', {
   expect_error(
      new_set <- update_workflow_recipe(two_class_res, "yj_trans_cart", recipe = rec),
      regexp = NA
   )
   new_rec <- extract_recipe(new_set, id = "yj_trans_cart", estimated = FALSE)


   expect_true(all(tidy(new_rec)$type == "normalize"))
   expect_equal(new_set$result[[4]], list())

   expect_error(
      new_new_set <-
         update_workflow_recipe(
            new_set,
            "yj_trans_cart",
            recipe = rec,
            blueprint = sparse_bp
         ),
      regexp = NA
   )
   new_wflow <- extract_workflow(new_new_set, "yj_trans_cart")
   expect_equal(new_wflow$pre$actions$recipe$blueprint, sparse_bp)

})