test_that("Likelihoods can be calculated", {
  chains <- c(1, 1, 4, 7)
  set.seed(12)
  expect_lt(
    likelihood(
      chains = chains,
      statistic = "size",
      offspring_dist = rpois,
      lambda = 0.5
    ),
    0
  )
  expect_lt(
    likelihood(
      chains = chains,
      statistic = "size",
      offspring_dist = rpois,
      lambda = 0.5,
      exclude = 1
    ),
    0
  )
  expect_lt(
    likelihood(
      chains = chains,
      statistic = "size",
      offspring_dist = rpois,
      lambda = 0.5,
      stat_threshold = 5
    ),
    0
  )
  expect_lt(
    likelihood(
      chains = chains,
      statistic = "size",
      offspring_dist = rpois,
      lambda = 0.5,
      obs_prob = 0.5,
      nsim_obs = 1
    ),
    0
  )
  expect_lt(
    likelihood(
      chains = chains,
      statistic = "size",
      offspring_dist = rpois,
      lambda = 0.5,
      stat_threshold = 5,
      obs_prob = 0.5,
      nsim_obs = 1
    ),
    0
  )
  expect_lt(
    likelihood(
      chains = chains,
      statistic = "length",
      offspring_dist = rbinom,
      size = 1,
      prob = 0.5
    ),
    0
  )
  expect_gte(
    likelihood(
      chains = chains,
      statistic = "length",
      offspring_dist = rbinom,
      size = 1,
      prob = 0.5,
      log = FALSE
    ),
    0
  )
  expect_gte(
    likelihood(
      chains = chains,
      statistic = "length",
      offspring_dist = rbinom,
      size = 1,
      prob = 0.5,
      individual = FALSE,
      log = FALSE
    ),
    0
  )
})

test_that("likelihood() works with epichains and epichains_summary objects", {
  # Simulate an <epichains> object
  set.seed(32)
  chains_tree_eg <- simulate_chains(
    n_chains = 10,
    pop = 100,
    percent_immune = 0,
    statistic = "size",
    offspring_dist = rpois,
    stat_threshold = 10,
    generation_time = function(n) rep(3, n),
    lambda = 0.9
  )
  # Simulate an <epichains_summary> object
  set.seed(32)
  chains_summary_eg <- simulate_chain_stats(
    n_chains = 10,
    pop = 100,
    percent_immune = 0,
    statistic = "size",
    offspring_dist = rpois,
    stat_threshold = 10,
    lambda = 0.9
  )
  # Use the simulated <epichains> object to calculate likelihood
  expect_equal(
    likelihood(
      chains = chains_tree_eg,
      statistic = "size",
      offspring_dist = rpois,
      lambda = 0.9
    ),
    -23.538996774
  )
  # Use the simulated <epichains_summary> object to calculate likelihood
  expect_equal(
    likelihood(
      chains = chains_summary_eg,
      statistic = "size",
      offspring_dist = rpois,
      lambda = 0.9
    ),
    -23.538997
  )
})

test_that("Likelihoods are numerically correct", {
  chains <- c(1, 1, 4, 7)
  set.seed(12)
  expect_identical(
    round(
      likelihood(
        chains = chains,
        statistic = "size",
        offspring_dist = rpois,
        lambda = 0.5
      ), 5
    ),
    -8.6072
  )
  expect_identical(
    round(
      likelihood(
        chains = chains,
        statistic = "size",
        offspring_dist = rnbinom,
        mu = 0.5,
        size = 0.2
      ), 5
    ),
    -9.13437
  )
  expect_identical(
    round(
      likelihood(
        chains = chains,
        statistic = "size",
        offspring_dist = rgborel,
        prob = 0.5,
        size = 0.2
      ), 5
    ),
    -11.21929
  )
  expect_identical(
    round(
      likelihood(
        chains = chains,
        statistic = "length",
        offspring_dist = rpois,
        lambda = 0.5
      ), 5
    ),
    -9.39945
  )
  expect_identical(
    round(
      likelihood(
        chains = chains,
        statistic = "length",
        offspring_dist = rgeom,
        prob = 0.5
      ), 5
    ),
    -12.48639
  )
  expect_identical(
    round(
      unlist(
        likelihood(
          chains = chains,
          statistic = "size",
          offspring_dist = rpois,
          lambda = 0.5,
          log = FALSE,
          individual = TRUE
        )
      ),
      2
    ),
    c(0.61, 0.61, 0.05, 0.01)
  )
})

test_that("Errors are thrown", {
  chains <- c(1, 1, 4, 7)
  set.seed(12)
  expect_error(
    likelihood(
      chains = chains,
      offspring_dist = list(),
      statistic = "size",
      lambda = 0.5
    ),
    "Must be a function"
  )
  expect_error(
    likelihood(
      chains = chains,
      offspring_dist = rpois,
      statistic = "size",
      lambda = 0.5,
      obs_prob = 3 # obs_prob > 1
    ),
    "is not <= 1"
  )
  expect_error(
    likelihood(
      chains = chains,
      offspring_dist = rpois,
      statistic = "size",
      lambda = 0.5,
      obs_prob = 0.5
    ),
    "must be specified"
  )
  expect_error(
    likelihood(
      chains = chains,
      statistic = "size",
      offspring_dist = rpois,
      nsim_obs = 100,
      lambda = 0.5,
      log = "s" # log must be a logical
    ),
    "Must be of type"
  )
  expect_error(
    likelihood(
      chains = as.factor(chains), # chains cannot be a factor
      statistic = "size",
      offspring_dist = rpois,
      lambda = 0.5
    ),
    "Must be of type"
  )
  expect_error(
    likelihood(
      chains = chains,
      statistic = "size",
      offspring_dist = rpois,
      lambda = 0.5,
      stat_threshold = NULL # stat_threshold must be an integer or Inf
    ),
    "Assertion failed"
  )
  expect_error(
    likelihood(
      chains = c(1:2, Inf), # Inf in data is not allowed if stat_threshold = Inf
      statistic = "size",
      offspring_dist = rpois,
      lambda = 0.5,
      stat_threshold = Inf
    ),
    "must be censored with a finite value"
  )
})

test_that("Warnings are thrown", {
  # Simulate an <epichains> object
  set.seed(32)
  chains_tree_eg <- simulate_chains(
    n_chains = 10,
    pop = 100,
    percent_immune = 0,
    statistic = "size",
    offspring_dist = rpois,
    stat_threshold = 10,
    generation_time = function(n) rep(3, n),
    lambda = 0.9
  )
  # We expect a warning when using the <epichains> object that contains Infs
  # and stat_threshold is Inf
  expect_warning(
    likelihood(
      chains = chains_tree_eg,
      statistic = "size",
      offspring_dist = rpois,
      lambda = 0.9,
      stat_threshold = Inf
    ),
    "`stat_threshold` specified but will be ignored."
  )
})