fix: error early when .x is non-numeric and default .fun is used (#387)

[LeonidasZhak]

Problem

fct_reorder() produces incorrect results or confusing errors when .x is a character vector or factor and the default .fun = median is used:

f = c("a", "b", "b")
x = c("z", "x", "y")

# Character .x: silently returns NA with warning, wrong ordering
forcats::fct_reorder(f, x)
#> Warning in mean.default(sort(x, partial = half + 0L:1L)[half + 0L:1L]):
#> argument is not numeric or logical: returning NA
#> [1] a b b
#> Levels: a b  # Wrong! Should be b a

# Factor .x: confusing error about numeric data
x = factor(c("z", "x", "y"))
forcats::fct_reorder(f, x)
#> Error in median.default(x, ...): need numeric data

Solution

Added input validation that errors early with a clear message when .x is not numeric and the default .fun = median is used:

fct_reorder(f, x)
#> Error in `fct_reorder()`:
#> ! `.x` must be a numeric vector when using the default `.fun`.
#> i Either supply a numeric `.x` or provide a custom `.fun`.

Users can still use fct_reorder() with character/factor .x by providing a custom .fun:

fct_reorder(f, x, .fun = function(x) x[1])
#> [1] a b b
#> Levels: b a

Changes

• R/reorder.R: Added validation for .x type when using default .fun
• tests/testthat/test-reorder.R: Added 3 tests for character, factor, and custom .fun cases
• tests/testthat/_snaps/reorder.md: Updated snapshots with new error messages

Validation

• All 223 tests pass
• R CMD check passes with Status: OK
• No duplicate PRs found for issue Incorrect results of fct_reorder() with a character vector and duplicates #387

[LeonidasZhak]

Withdrawing this because CI is currently failing and I should not ask maintainers to review an unready automated PR. I will only come back with a clean, focused change if useful. Sorry for the noise, and thank you for maintaining the project.