tests/testthat/test_convertToLlamaCVFolds.R

In aslib: Interface to the Algorithm Selection Benchmark Library

context("convertToLlamaCVFolds")

test_that("convertToLlamaCVFolds", {
  skip_on_cran()
  folds = createCVSplits(testscenario1, folds = 2L, reps = 1L)
  llama.scenario = convertToLlamaCVFolds(testscenario1, cv.splits = folds)
  lrn = makeLearner("classif.rpart")
  res = classify(classifier = lrn, data = llama.scenario)

  # FIXME: re-add when llama can handle more than 1 rep
  # folds = createCVSplits(testscenario1, folds = 2L, reps = 2L)
  # llama.scenario = convertToLlamaCVFolds(testscenario1, cv.splits = folds)
  # res = classify(classifier = makeLearner("classif.rpart"), data = llama.scenario)
})

test_that("convertToLlamaCVFolds check matching", {
  folds = createCVSplits(testscenario1, folds = 2L, reps = 1L)
  llama.scenario = convertToLlamaCVFolds(testscenario1, cv.splits = folds)
  for (i in unique(folds$fold)) {
    iid1 = as.character(llama.scenario$data[llama.scenario$test[[i]],]$instance_id)
    iid2 = as.character(subset(folds, folds$fold == i)$instance_id)
    expect_true(setequal(iid1, iid2))
  }

  folds = createCVSplits(testscenario6, folds = 2L, reps = 1L)
  llama.scenario = convertToLlamaCVFolds(testscenario6, cv.splits = folds)
  for (i in unique(folds$fold)) {
    iid1 = as.character(llama.scenario$data[llama.scenario$test[[i]],]$instance_id)
    iid2 = as.character(subset(folds, folds$fold == i)$instance_id)
    expect_true(setequal(iid1, iid2))
  }
})