Split large tests up into smaller units with more meaningful test/error messages

[ttimbers]

Some of your tests are quite large:

test_that("Output errors", {
  # generate an output
  data_split <- train_test_split(X, y)
  # check the output type
  expect_is(data_split, "list")
  # check the output length
  expect_equal(length(data_split), 4)

  # assign outputs
  X_train <- data_split[[1]]
  X_test <- data_split[[2]]
  y_train <- data_split[[3]]
  y_test <- data_split[[4]]

  # check the output types of elements in the list
  expect_true(is.data.frame(X_train) | is.atomic(X_train))
  expect_true(is.data.frame(X_test) | is.atomic(X_test))
  expect_true(is.data.frame(y_train) | is.atomic(y_train))
  expect_true(is.data.frame(y_test) | is.atomic(y_test))


  # check if the dimension equals between split data and origianl data
  # check X and X_train, X_test
  expect_equal(get_nrows(X), get_nrows(X_train)+get_nrows(X_test))
  # check y and y_train, y_test
  expect_equal(get_nrows(y), get_nrows(y_train)+get_nrows(y_test))

})

You should split this up into 4 different test_that statements, using what you currently have as comments as the test/error messages. This strategy makes your tests more useful and efficient, as it's much easier to pinpoint what exactly is broken with more, smaller tests.