tests/testthat/test_filter_classif.R

In mlr3filters: Filter Based Feature Selection for 'mlr3'

test_that("all classif filters return correct filter values", {
  task = mlr_tasks$get("sonar")
  task$select(head(task$feature_names, 3))
  filters = mlr_filters$mget(mlr_filters$keys())
  filters$permutation$param_set$values = list(nmc = 2)

  for (f in filters) {
    if ("classif" %in% f$task_types && all(require_namespaces(f$packages, quietly = TRUE))) {
      f$calculate(task)
      expect_filter(f, task = task)
    }
  }
})


test_that("filters throw errors on missing values", {
  task = tsk("sonar")
  data = task$data(cols = c(task$target_names, head(task$feature_names, 3)))
  data$V1[1] = NA
  task = as_task_classif(data, target = task$target_names)

  filters = mlr_filters$mget(mlr_filters$keys())

  for (f in filters) {
    if ("classif" %nin% f$task_types) {
      next
    }

    if (!all(require_namespaces(f$packages, quietly = TRUE))) {
      next
    }

    if ("missings" %in% f$properties) {
      f$calculate(task)
    } else {
      expect_error(f$calculate(task), "missing values")
    }
  }
})