test_that("matching claim passes core audit", {
  ref <- stat_reference(10.4, "unemployment rate", "Spain", "2025", "percent", "Eurostat")
  clm <- stat_claim(10.4, "unemployment rate", "Spain", "2025", "%", "Eurostat")
  aud <- audit_stat_ai(ref, clm)
  expect_s3_class(aud, "stat_fidelity_audit")
  expect_equal(aud$components$numerical$score, 1)
  expect_equal(aud$components$unit$score, 1)
  expect_false(aud$critical_fail)
})

test_that("wrong indicator is non-compensatory", {
  ref <- stat_reference(10.4, "unemployment rate", "Spain", "2025", "percent", "Eurostat")
  clm <- stat_claim(10.4, "employment rate", "Spain", "2025", "%", "Eurostat")
  aud <- audit_stat_ai(ref, clm)
  expect_true(aud$critical_fail)
  expect_equal(aud$certification, "FAIL")
})

test_that("wrong source fails", {
  ref <- stat_reference(10.4, "unemployment rate", "Spain", "2025", "percent", "Eurostat")
  clm <- stat_claim(10.4, "unemployment rate", "Spain", "2025", "%", "Other source")
  aud <- audit_stat_ai(ref, clm)
  expect_equal(aud$components$provenance$score, 0)
  expect_true(aud$components$provenance$critical)
})


test_that("batch audit returns component score columns", {
  refs <- list(
    stat_reference(10.4, "unemployment rate", "Spain", "2025", "percent", "Eurostat"),
    stat_reference(7.0, "unemployment rate", "France", "2025", "percent", "Eurostat")
  )
  claims <- list(
    stat_claim(10.4, "unemployment rate", "Spain", "2025", "%", "Eurostat"),
    stat_claim(7.2, "unemployment rate", "France", "2025", "%", "Eurostat")
  )
  out <- audit_batch(refs, claims, ids = c("ES", "FR"))
  expect_s3_class(out, "stat_fidelity_batch")
  expect_equal(nrow(out), 2)
  expect_true(all(c("semantic", "numerical", "context") %in% names(out)))
  expect_equal(out$certification[2], "FAIL")
})