tests/testthat/test-orbital.R

In orbital: Predict with 'tidymodels' Workflows in Databases

test_that("orbital works with workflows - recipe", {
	skip_if_not_installed("recipes")
	skip_if_not_installed("tidypredict")
	skip_if_not_installed("workflows")
	rec_spec <- recipes::recipe(mpg ~ ., data = mtcars) %>%
		recipes::step_normalize(recipes::all_numeric_predictors())

	lm_spec <- parsnip::linear_reg()

	wf_spec <- workflows::workflow() %>%
		workflows::add_recipe(rec_spec) %>%
		workflows::add_model(lm_spec)

	wf_fit <- parsnip::fit(wf_spec, mtcars)

	res <- orbital(wf_fit)

	expect_s3_class(res, "orbital_class")
	expect_true(is.character(res))
	expect_named(res)
	expect_length(res, 1 + (ncol(mtcars) - 1))
})

test_that("orbital works with workflows - formula", {
	skip_if_not_installed("tidypredict")
	skip_if_not_installed("workflows")
	lm_spec <- parsnip::linear_reg()

	wf_spec <- workflows::workflow() %>%
		workflows::add_formula(mpg ~ .) %>%
		workflows::add_model(lm_spec)

	wf_fit <- parsnip::fit(wf_spec, mtcars)

	res <- orbital(wf_fit)

	expect_s3_class(res, "orbital_class")
	expect_true(is.character(res))
	expect_named(res, ".pred")
	expect_length(res, 1)
})

test_that("orbital works with workflows - variables", {
	skip_if_not_installed("tidypredict")
	skip_if_not_installed("workflows")
	lm_spec <- parsnip::linear_reg()

	wf_spec <- workflows::workflow() %>%
		workflows::add_variables(outcomes = "mpg", predictors = everything()) %>%
		workflows::add_model(lm_spec)

	wf_fit <- parsnip::fit(wf_spec, mtcars)

	res <- orbital(wf_fit)

	expect_s3_class(res, "orbital_class")
	expect_true(is.character(res))
	expect_named(res, ".pred")
	expect_length(res, 1)
})

test_that("orbital errors on non-trained workflow", {
	skip_if_not_installed("recipes")
	skip_if_not_installed("workflows")
	rec_spec <- recipes::recipe(mpg ~ ., data = mtcars) %>%
		recipes::step_normalize(recipes::all_numeric_predictors())

	lm_spec <- parsnip::linear_reg()

	wf_spec <- workflows::workflow(rec_spec, lm_spec)

	expect_snapshot(
		error = TRUE,
		orbital(wf_spec)
	)
})

test_that("orbital works with recipe", {
	skip_if_not_installed("tidypredict")
	skip_if_not_installed("recipes")
	rec_spec <- recipes::recipe(mpg ~ ., data = mtcars) %>%
		recipes::step_normalize(recipes::all_numeric_predictors())

	rec_prep <- recipes::prep(rec_spec)

	res <- orbital(rec_prep)

	expect_s3_class(res, "orbital_class")
	expect_true(is.character(res))
	expect_named(res)
})

test_that("orbital errors untrained recipe", {
	skip_if_not_installed("recipes")
	rec_spec <- recipes::recipe(mpg ~ ., data = mtcars) %>%
		recipes::step_normalize(recipes::all_numeric_predictors())

	expect_snapshot(
		error = TRUE,
		orbital(rec_spec)
	)
})

test_that("orbital works with parsnip", {
	skip_if_not_installed("tidypredict")
	lm_spec <- parsnip::linear_reg()

	lm_fit <- parsnip::fit(lm_spec, mpg ~ ., data = mtcars)

	res <- orbital(lm_fit)

	expect_s3_class(res, "orbital_class")
	expect_true(is.character(res))
	expect_named(res, ".pred")
})

test_that("orbital errors on non-trained parsnip", {
	skip_if_not_installed("parsnip")
	lm_spec <- parsnip::linear_reg()

	expect_snapshot(
		error = TRUE,
		orbital(lm_spec)
	)
})

test_that("orbital errors nicely on post-processing", {
	skip_if_not_installed("workflows")
	lm_spec <- parsnip::linear_reg()

	wf_spec <- workflows::workflow() %>%
		workflows::add_formula(mpg ~ .) %>%
		workflows::add_model(lm_spec)

	wf_fit <- parsnip::fit(wf_spec, mtcars)

	# fake post-processing happening
	wf_fit$post$actions <- list(thing = 1)

	expect_snapshot(
		error = TRUE,
		orbital(wf_fit)
	)
})

test_that("orbital errors on wrong input", {
	expect_snapshot(
		error = TRUE,
		orbital(lm(mpg ~ ., data = mtcars))
	)
})

test_that("orbital printing works", {
	skip_if_not_installed("recipes")
	skip_if_not_installed("tidypredict")
	skip_if_not_installed("workflows")
	rec_spec <- recipes::recipe(mpg ~ ., data = mtcars) %>%
		recipes::step_normalize(recipes::all_numeric_predictors())

	lm_spec <- parsnip::linear_reg()

	wf_spec <- workflows::workflow(rec_spec, lm_spec)

	wf_fit <- parsnip::fit(wf_spec, mtcars)

	expect_snapshot(
		orbital(wf_fit)
	)

	expect_snapshot(
		print(orbital(wf_fit), digits = 2)
	)

	expect_snapshot(
		print(orbital(wf_fit), truncate = FALSE)
	)
})

test_that("prefix argument works", {
	skip_if_not_installed("recipes")
	skip_if_not_installed("parsnip")
	skip_if_not_installed("workflows")
	skip_if_not_installed("tidypredict")

	rec_spec <- recipes::recipe(mpg ~ ., data = mtcars) %>%
		recipes::step_normalize(recipes::all_numeric_predictors())

	lm_spec <- parsnip::linear_reg()

	wf_spec <- workflows::workflow(rec_spec, lm_spec)

	wf_fit <- parsnip::fit(wf_spec, mtcars)

	orb_obj <- orbital(wf_fit, prefix = "pred")

	expect_true("pred" %in% names(orb_obj))
	expect_false(".pred" %in% names(orb_obj))
})