fix: fct_na_value_to_level() no longer adds unnecessary level when no NAs present (#347)

[LeonidasZhak]

Summary

Fixes #347

When called on a factor with no NA values, fct_na_value_to_level() previously added an unnecessary level (either NA or the custom level). Now it returns the factor unchanged, matching the behavior of the deprecated fct_explicit_na().

Changes

• R/na.R: Added early return when f contains no NA values
• tests/testthat/test-na.R: Added test verifying the fix

Reprex

Before:

library(forcats)
letters[1:3] |> factor() |> fct_na_value_to_level("x") |> levels()
# [1] "a" "b" "c" "x"  # unexpected: "x" level added

After:

library(forcats)
letters[1:3] |> factor() |> fct_na_value_to_level("x") |> levels()
# [1] "a" "b" "c"  # correct: no change when no NAs

Tests

All 225 forcats tests pass, including the new test.