skip_if_not_installed("future")

test_that("parallel resample", {
  skip_if_not_installed("future.callr")
  skip_if_not_installed("progressr")

  with_future(future.callr::callr, {
    task = tsk("iris")
    learner = lrn("classif.rpart")

    progressr::with_progress({
      rr = resample(task, learner, rsmp("cv", folds = 3))
    })
    expect_resample_result(rr)
    expect_data_table(rr$errors, nrows = 0L)
  })
})

test_that("seeds work identical during sequential and parallel execution", {
  task = tsk("sonar")
  learner = lrn("classif.debug", predict_type = "prob")
  resampling = rsmp("cv", folds = 3L)
  measure = msr("classif.auc")

  rr1 = with_seed(123, with_future(future::sequential, resample(task, learner, resampling)))
  rr2 = with_seed(123, with_future(future::multisession, resample(task, learner, resampling)))

  expect_equal(
    as.data.table(rr1$prediction())$prob.M,
    as.data.table(rr2$prediction())$prob.M
  )
})

test_that("parallel benchmark", {
  skip_if_not_installed("future.callr")

  task = tsk("iris")
  learner = lrn("classif.rpart")

  with_future(future.callr::callr, {
    progressr::with_progress({
      bmr = benchmark(benchmark_grid(task, learner, rsmp("cv", folds = 3)))
    })
  })
  expect_benchmark_result(bmr)
  expect_equal(bmr$aggregate(conditions = TRUE)$warnings, 0L)
  expect_equal(bmr$aggregate(conditions = TRUE)$errors, 0L)

  grid = benchmark_grid(list(tsk("wine"), tsk("sonar")), replicate(2, lrn("classif.debug")), rsmp("cv", folds = 2))
  njobs = 3L
  bmr = with_future(future::multisession, {
    benchmark(grid, store_models = TRUE)
  }, workers = njobs)

  expect_benchmark_result(bmr)
  pids = map_int(as.data.table(bmr)$learner, function(x) x$model$pid)
  expect_equal(length(unique(pids)), njobs)
})

test_that("real parallel resample", {
  skip_if_not_installed("progressr")
  skip_on_os("windows") # currently buggy

  with_future(future::multisession, {
    task = tsk("iris")
    learner = lrn("classif.rpart")

    progressr::with_progress({
      rr = resample(task, learner, rsmp("cv", folds = 3))
    })
    expect_resample_result(rr)
    expect_data_table(rr$errors, nrows = 0L)
  })
})

test_that("parallel seed", {
  skip_if_not_installed("future.callr")

  task = tsk("wine")
  learner = lrn("classif.debug", predict_type = "prob")

  rr1 = with_seed(123, resample(task, learner, rsmp("cv", folds = 3)))
  with_future(future.callr::callr, {
    rr2 = with_seed(123, resample(task, learner, rsmp("cv", folds = 3)))
  })
  expect_equal(rr1$prediction()$prob, rr2$prediction()$prob)
})

test_that("data table threads are not changed in main session", {
  skip_on_os("mac") # number of threads cannot be changed on mac
  skip_on_cran()

  old_dt_threads = getDTthreads()
  on.exit({
    setDTthreads(old_dt_threads)
  }, add = TRUE)
  setDTthreads(2L)

  task = tsk("sonar")
  learner = lrn("classif.debug", predict_type = "prob")
  resampling = rsmp("cv", folds = 3L)
  measure = msr("classif.auc")

  rr1 = with_seed(123, with_future(future::sequential, resample(task, learner, resampling)))
  expect_equal(getDTthreads(), 2L)

  rr2 = with_seed(123, with_future(future::multisession, resample(task, learner, resampling)))
  expect_equal(getDTthreads(), 2L)
})