test_that("hmix returns a list with expected structure", {
  result <- hmix(dummy_set$AMZN, horizon = 5, centers = 4, n_hidden = 4, n_samples = 200, seed = 123, n_tests = 3, warmup = 0.3)

  expect_type(result, "list")
  expect_named(result, c("model", "error_sets"))
  expect_type(result$model, "list")

  model <- result$model

  expect_type(model, "list")
  expect_named(model, c('hmm_model', 'pred_funs'))
  expect_type(model$hmm_model, "list")
  expect_named(model$hmm_model, c('observations', 'hmm', 'trained_hmm'))
  expect_named(model$pred_funs, c('t1', 't2', 't3', 't4', 't5'))
  expect_true(is.numeric(model$pred_funs$t1$rfun(5)) & length(model$pred_funs$t1$rfun(5)) == 5)
  expect_type(model$pred_funs$t1$qfun(c(0, 1)), "double")
  expect_type(model$pred_funs$t1$pfun(tail(ts)), "double")
  expect_type(model$pred_funs$t1$dfun(tail(ts)), "double")
  expect_true(is.numeric(model$pred_funs$t5$rfun(5)) & length(model$pred_funs$t5$rfun(5)) == 5)
  expect_type(model$pred_funs$t5$qfun(c(0, 1)), "double")
  expect_type(model$pred_funs$t5$pfun(tail(ts)), "double")
  expect_type(model$pred_funs$t5$dfun(tail(ts)), "double")

  expect_type(result$error_sets, "list")
  crps_set <- result$error_sets$crps_set
  expect_type(crps_set, "double")
  expect_true(all(crps_set >= 0))
  expect_true(prod(dim(crps_set)) == 15)
  })

test_that("hmix works with larger hyperparameters", {
  result <- hmix(dummy_set$IBM, horizon = 20, centers = 15, n_hidden = 8, n_samples = 1000, seed = 123, n_tests = 10, warmup = 0.6)

  expect_type(result, "list")
  expect_named(result, c("model", "error_sets"))
  expect_type(result$model, "list")

  model <- result$model

  expect_type(model, "list")
  expect_named(model, c('hmm_model', 'pred_funs'))
  expect_type(model$hmm_model, "list")
  expect_named(model$hmm_model, c('observations', 'hmm', 'trained_hmm'))
  expect_named(model$pred_funs, c('t1', 't2', 't3', 't4', 't5', 't6', 't7', 't8', 't9', 't10', 't11', 't12', 't13', 't14', 't15', 't16', 't17', 't18', 't19', 't20'))
  expect_true(is.numeric(model$pred_funs$t1$rfun(5)) & length(model$pred_funs$t1$rfun(5)) == 5)
  expect_type(model$pred_funs$t1$qfun(c(0, 1)), "double")
  expect_type(model$pred_funs$t1$pfun(tail(ts)), "double")
  expect_type(model$pred_funs$t1$dfun(tail(ts)), "double")
  expect_true(is.numeric(model$pred_funs$t20$rfun(5)) & length(model$pred_funs$t20$rfun(5)) == 5)
  expect_type(model$pred_funs$t20$qfun(c(0, 1)), "double")
  expect_type(model$pred_funs$t20$pfun(tail(ts)), "double")
  expect_type(model$pred_funs$t20$dfun(tail(ts)), "double")

  expect_type(result$error_sets, "list")
  crps_set <- result$error_sets$crps_set
  expect_type(crps_set, "double")
  expect_true(all(crps_set >= 0))
  expect_true(prod(dim(crps_set)) == 200)
})


test_that("hmix produces consistent results with the same seed", {

  result1 <- hmix(dummy_set$NVDA, horizon = 10, centers = 5, n_hidden = 3, n_samples = 500, seed = 123, n_tests = 5, warmup = 0.4)
  result2 <- hmix(dummy_set$NVDA, horizon = 10, centers = 5, n_hidden = 3, n_samples = 500, seed = 123, n_tests = 5, warmup = 0.4)

  expect_equal(result1$model$hmm_model$hmm$transProbs, result2$model$hmm_model$hmm$transProbs)
  expect_equal(result1$model$hmm_model$hmm$emissionProbs, result2$model$hmm_model$hmm$emissionProbs)
  expect_equal(result1$error_sets$crps_set, result2$error_sets$crps_set)
})