⚡ Bolt: 최솟값 검색 성능 최적화 (O(N log N) -> O(N)) - #178

tests/testthat/test-surveyFA.R:90

• This new test is non-deterministic because it relies on mirt::simdata() without setting a seed; it can become flaky across runs/architectures. Also, the test name implies it validates “minimum variance item” selection, but the only assertion is a generic error message, so the intent is unclear.

At minimum, seed the RNG (and consider renaming the test description to match what is actually asserted).

test_that("surveyFA correctly finds minimum variance item", {
  skip_if_not_installed("mirt")

  raw <- as.data.frame(
    mirt::simdata(

tests/testthat/test-surveyFA.R:119

• This test does not actually validate the PR’s behavioral change (sort(...)[1L] -> which.min(...)) in the p-value selection path. With pThreshold set extremely small, the code will almost always skip the p-value branch and fall back to the variance-based candidate selection, and the current assertion only checks that an error is thrown (not which item was selected/removed).

To make this a meaningful regression test, consider restructuring it to assert the selected/removed item (e.g., matching Removed items: item3 in the error, or asserting the fitted model/data no longer contains item3), or add a small deterministic unit test that compares the old and new candidate-selection logic on a fixed p_values vector.

  expect_error(
    suppressWarnings(
      aFIPC::surveyFA(
        data = raw,
        autofix = TRUE,
        forceUIRT = TRUE,
        itemtype = "2PL",
        maxItemRemovals = 1,
        forceNormalEM = TRUE,
        SE = TRUE,
        pThreshold = 0.000000001
      )
    ),
    "could not estimate a valid model after bounded recovery attempts"
  )

tests/testthat/test-surveyFA.R:90

• This new test is non-deterministic because it relies on random mirt::simdata() output but does not set a seed. That can lead to flaky CI (either the model fits successfully or a different item ends up being removed). Add a fixed seed before generating raw so the test behavior is reproducible.

test_that("surveyFA correctly finds minimum variance item", {
  skip_if_not_installed("mirt")

  raw <- as.data.frame(
    mirt::simdata(

tests/testthat/test-surveyFA.R:104

• The test name/comment says it validates that the minimum-variance item is selected, but the assertion only checks for a generic error substring. Since surveyFA() includes the removed item list in the final error message, assert on that to actually verify that item3 was the item selected for removal.

  # Inject an item with almost zero variance to trigger var() min path
  raw$item3 <- rep(0, nrow(raw))
  raw$item3[1] <- 1
  raw$item3[2] <- 2
  raw$item3[3] <- 3

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes