context("posterior evaluations")

## DO NOT USE fake_sampling()
combo2 <- run_example("combo2")
combo3 <- run_example("combo3")

test_that("Outputs of posterior_* functions have expected shapes.", {
  codata_combo2_alt <- codata_combo2
  combo2$codata_combo2_alt <- codata_combo2

  iter <- getOption("OncoBayes2.MC.iter")
  warmup <- getOption("OncoBayes2.MC.warmup")
  chains <- getOption("OncoBayes2.MC.chains")
  num_sim <- chains * (iter - warmup)

  pp1 <- with(combo2, posterior_predict(blrmfit, newdata = codata_combo2_alt))
  expect_equal(ncol(pp1), nrow(codata_combo2_alt))
  expect_equal(nrow(pp1), num_sim)
  expect_equal(nsamples(combo2$blrmfit), num_sim)
  pp2 <- with(combo2, posterior_linpred(blrmfit, newdata = codata_combo2_alt))
  expect_equal(ncol(pp2), nrow(codata_combo2_alt))
  expect_equal(nrow(pp2), num_sim)
})

test_that("Unkown groups are rejected in posterior_* functions.", {
  codata_combo2_alt <- codata_combo2
  lev_old <- levels(codata_combo2_alt$group_id)
  levels(codata_combo2_alt$group_id) <- c(paste0("new_", lev_old[1]), lev_old[-1])
  combo2$codata_combo2_alt <- codata_combo2_alt

  expect_error(with(combo2, posterior_predict(blrmfit, newdata = codata_combo2_alt)),
    regexp = "Found unkown factor levels in grouping: new_trial_A"
  )
  expect_error(with(combo2, posterior_linpred(blrmfit, newdata = codata_combo2_alt)),
    regexp = "Found unkown factor levels in grouping: new_trial_A"
  )

  ## same error if the group_id is a character instead
  combo2$codata_combo2_alt$group_id <- as.character(combo2$codata_combo2_alt$group_id)
  expect_error(with(combo2, posterior_predict(blrmfit, newdata = codata_combo2_alt)),
    regexp = "Found unkown factor levels in grouping: new_trial_A"
  )
  expect_error(with(combo2, posterior_linpred(blrmfit, newdata = codata_combo2_alt)),
    regexp = "Found unkown factor levels in grouping: new_trial_A"
  )

  ## flip the level definitions
  combo2$codata_combo2_alt$group_id <- codata_combo2$group_id
  levels(combo2$codata_combo2_alt$group_id)[1:2] <- levels(codata_combo2$group_id)[2:1]
  expect_error(with(combo2, posterior_predict(blrmfit, newdata = codata_combo2_alt)),
    regexp = "Mismatch in factor level defintion of grouping"
  )
  expect_error(with(combo2, posterior_linpred(blrmfit, newdata = codata_combo2_alt)),
    regexp = "Mismatch in factor level defintion of grouping"
  )
})

test_that("Unkown strata are rejected in posterior_* functions.", {
  hist_combo3_alt <- hist_combo3
  old_levs <- levels(hist_combo3_alt$stratum_id)
  levels(hist_combo3_alt$stratum_id)[1] <- "BIDflex"
  combo3$hist_combo3_alt <- hist_combo3_alt

  expect_error(with(combo3, posterior_predict(blrmfit, newdata = hist_combo3_alt)),
    regexp = "Found unkown factor levels in stratum: BIDflex"
  )
  expect_error(with(combo3, posterior_linpred(blrmfit, newdata = hist_combo3_alt)),
    regexp = "Found unkown factor levels in stratum: BIDflex"
  )

  ## same error if the stratum_id is a character instead
  combo3$hist_combo3_alt$stratum_id <- as.character(combo3$hist_combo3_alt$stratum_id)
  expect_error(with(combo3, posterior_predict(blrmfit, newdata = hist_combo3_alt)),
    regexp = "Found unkown factor levels in stratum: BIDflex"
  )
  expect_error(with(combo3, posterior_linpred(blrmfit, newdata = hist_combo3_alt)),
    regexp = "Found unkown factor levels in stratum: BIDflex"
  )

  ## flip the level definitions
  combo3$hist_combo3_alt$stratum_id <- hist_combo3$stratum_id
  levels(combo3$hist_combo3_alt$stratum_id)[1:2] <- levels(hist_combo3$stratum_id)[2:1]
  expect_error(with(combo3, posterior_predict(blrmfit, newdata = hist_combo3_alt)),
    regexp = "Mismatch in factor level defintion of stratum"
  )
  expect_error(with(combo3, posterior_linpred(blrmfit, newdata = hist_combo3_alt)),
    regexp = "Mismatch in factor level defintion of stratum"
  )
})


## check blrm_trial objects (basic)
check_trial_posterior_functions <- function(example) {
  with(example, {
    suppressWarnings(suppressMessages(trial <- blrm_trial(histdata, dose_info, drug_info)))

    expect_error(posterior_linpred(trial), ".*configure the prior.$")
    expect_error(posterior_interval(trial), ".*configure the prior.$")
    expect_error(predictive_interval(trial), ".*configure the prior.$")
    expect_error(nsamples(trial), ".*configure the prior.$")
  })
}

check_trial_with_prior_posterior_functions <- function(example) {
  with(example, {
    suppressWarnings(suppressMessages(trial <- blrm_trial(histdata, dose_info, drug_info, simplified_prior = TRUE)))

    items <- nrow(histdata)
    draws <- nsamples(trial)
    dims <- c(draws, items)
    expect_equal(dims, dim(posterior_linpred(trial)))
    expect_equal(dims, dim(posterior_predict(trial)))
    expect_equal(posterior_interval(trial$blrmfit), posterior_interval(trial))
    expect_equal(c(items, 2), dim(predictive_interval(trial)))
  })
}

check_trial_posterior_functions(examples$single_agent)
check_trial_with_prior_posterior_functions(examples$single_agent)

check_trial_posterior_functions(examples$combo2)
check_trial_with_prior_posterior_functions(examples$combo2)


# test S3 methods in alphabetical order
test_that("as_draws and friends have resonable outputs", {
  draws <- as_draws(combo2$blrmfit, variable = "mu_log_beta[1,1]")
  expect_s3_class(draws, "draws_list")
  expect_equal(variables(draws), "mu_log_beta[1,1]")
  expect_equal(ndraws(draws), nsamples(combo2$blrmfit))

  draws <- suppressMessages(as_draws_matrix(combo2$blrmfit, variable = "mu_log_beta[1,1]"))
  expect_s3_class(draws, "draws_matrix")
  expect_equal(variables(draws), "mu_log_beta[1,1]")
  expect_equal(ndraws(draws), nsamples(combo2$blrmfit))

  draws <- as_draws_array(combo2$blrmfit, variable = "mu_log_beta[1,1]")
  expect_s3_class(draws, "draws_array")
  expect_equal(variables(draws), "mu_log_beta[1,1]")
  expect_equal(ndraws(draws), nsamples(combo2$blrmfit))

  draws <- as_draws_df(combo2$blrmfit, variable = "mu_log_beta[1,1]")
  expect_s3_class(draws, "draws_df")
  expect_equal(variables(draws), "mu_log_beta[1,1]")
  expect_equal(ndraws(draws), nsamples(combo2$blrmfit))
  
  draws <- as_draws_list(combo2$blrmfit, variable = "mu_log_beta[1,1]")
  expect_s3_class(draws, "draws_list")
  expect_equal(variables(draws), "mu_log_beta[1,1]")
  expect_equal(ndraws(draws), nsamples(combo2$blrmfit))
  
  draws <- as_draws_rvars(combo2$blrmfit)
  expect_s3_class(draws, "draws_rvars")
  expect_true(nvariables(draws) > 0)
  expect_equal(ndraws(draws), nsamples(combo2$blrmfit))
})