gof <- hydroeval_release_get_export("gof")

test_that("gof stays release-stable for default, all, and explicit metric selection", {
  scenario <- hydroeval_release_scenario("positive_reference")
  default_summary <- gof(scenario$observed, scenario$simulated)
  all_summary <- gof(scenario$observed, scenario$simulated, metrics = "all")
  explicit_summary <- gof(
    scenario$observed,
    scenario$simulated,
    metrics = c("pbias", "rho", "kge_2012")
  )

  expect_s3_class(default_summary, "hydroeval_gof")
  expect_identical(default_summary$metric_ids, hydroeval_release_gof_default_metric_ids)
  expect_identical(default_summary$request$metrics, "default")
  expect_identical(default_summary$request$resolved_metric_ids, hydroeval_release_gof_default_metric_ids)
  expect_identical(default_summary$request$selection_mode, "gof_default")

  expect_s3_class(all_summary, "hydroeval_gof")
  expect_identical(all_summary$metric_ids, hydroeval_release_gof_all_metric_ids)
  expect_identical(all_summary$request$metrics, "all")
  expect_identical(all_summary$request$resolved_metric_ids, hydroeval_release_gof_all_metric_ids)
  expect_identical(all_summary$request$selection_mode, "gof_all")

  expect_identical(explicit_summary$metric_ids, c("pbias", "rho", "kge_2012"))
  expect_identical(explicit_summary$display_labels, c("PBIAS (%)", "\u03C1", "KGE (2012)"))
  expect_identical(explicit_summary$request$metrics, c("pbias", "rho", "kge_2012"))
  expect_identical(explicit_summary$request$selection_mode, "gof_explicit")
})

test_that("gof values agree with standalone public wrappers and coercion views", {
  scenario <- hydroeval_release_scenario("positive_reference")
  summary <- gof(scenario$observed, scenario$simulated)
  expected_values <- vapply(
    hydroeval_release_gof_default_metric_ids,
    FUN.VALUE = numeric(1),
    FUN = function(metric_name) {
      hydroeval_release_call_wrapper(metric_name, scenario$observed, scenario$simulated)
    }
  )

  expect_named(summary, c("values", "metric_ids", "display_labels", "request"))
  expect_identical(names(summary$values), summary$metric_ids)
  expect_equal(summary$values, expected_values, tolerance = 1e-12)
  hydroeval_release_expect_gof_views_agree(summary)
})

test_that("gof display-label governance stays registry-facing rather than canonical-id-facing", {
  scenario <- hydroeval_release_scenario("positive_reference")
  summary <- gof(
    scenario$observed,
    scenario$simulated,
    metrics = c("pbias", "nrmse", "kge_2009", "kge_2012", "r_squared", "rho")
  )
  printed <- paste(capture.output(print(summary)), collapse = "\n")

  expect_identical(
    rownames(as.matrix(summary)),
    c("PBIAS (%)", "NRMSE", "KGE (2009)", "KGE (2012)", "R\u00B2", "\u03C1")
  )
  expect_match(printed, "PBIAS \\(%\\)")
  expect_match(printed, "NRMSE")
  expect_match(printed, "KGE \\(2009\\)")
  expect_match(printed, "KGE \\(2012\\)")
  expect_match(printed, "R\u00B2")
  expect_match(printed, "\u03C1")
  expect_no_match(printed, "kge_2009")
  expect_no_match(printed, "r_squared")
})

test_that("gof handles perfect-fit, missing-data, and invalid-selection public contracts", {
  perfect <- hydroeval_release_scenario("perfect_fit")
  missing <- hydroeval_release_scenario("missing_omit")
  single_point <- hydroeval_release_scenario("single_point")
  scenario <- hydroeval_release_scenario("positive_reference")

  perfect_summary <- gof(perfect$observed, perfect$simulated)
  omit_summary <- gof(
    missing$observed,
    missing$simulated,
    metrics = c("mae", "nse", "kge_2009"),
    na_policy = "omit"
  )
  kept_summary <- gof(
    missing$kept_observed,
    missing$kept_simulated,
    metrics = c("mae", "nse", "kge_2009"),
    na_policy = "omit"
  )

  expect_length(perfect_summary$values, length(hydroeval_release_gof_default_metric_ids))
  expect_equal(perfect_summary$values[["me"]], 0, tolerance = 1e-12)
  expect_equal(perfect_summary$values[["nse"]], 1, tolerance = 1e-12)
  expect_equal(perfect_summary$values[["kge_2012"]], 1, tolerance = 1e-12)
  expect_equal(omit_summary$values, kept_summary$values, tolerance = 1e-12)

  expect_error(
    gof(
      missing$observed,
      missing$simulated,
      metrics = c("mae", "nse", "kge_2009"),
      na_policy = "fail"
    ),
    class = "hydroeval_validation_error",
    regexp = "Missing `NA` values are not allowed"
  )
  expect_error(
    gof(single_point$observed, single_point$simulated),
    class = "hydroeval_validation_error",
    regexp = "Need at least 2 complete paired observations"
  )
  expect_error(
    gof(scenario$observed, scenario$simulated, metrics = c("MAE")),
    class = "hydroeval_metric_plan_error",
    regexp = "Metric `MAE`"
  )
})

test_that("gof rejects malformed metric selectors and preserves request metadata", {
  scenario <- hydroeval_release_scenario("positive_reference")
  summary <- gof(
    scenario$observed,
    scenario$simulated,
    metrics = c("mae", "rsr", "kge_2012"),
    na_policy = "omit"
  )

  expect_identical(summary$request$metrics, c("mae", "rsr", "kge_2012"))
  expect_identical(summary$request$resolved_metric_ids, c("mae", "rsr", "kge_2012"))
  expect_identical(summary$request$na_policy, "omit")
  expect_identical(summary$request$default_metric_ids, hydroeval_release_gof_default_metric_ids)

  expect_error(
    gof(scenario$observed, scenario$simulated, metrics = c("all", "mae")),
    class = "hydroeval_metric_plan_error",
    regexp = "mixture"
  )
  expect_error(
    gof(scenario$observed, scenario$simulated, metrics = c("mae", "not_a_metric")),
    class = "hydroeval_metric_plan_error",
    regexp = "not_a_metric"
  )
})