NAs in pluralize vector cause values to not display correctly

[rundel]

See the following reprex:

library(cli)

pluralize('{c(NA_character_)} file{?s}')
#> NA file
pluralize('{c("file1", NA_character_)} file{?s}')
#> NA files
pluralize('{c(NA_character_, "file1")} file{?s}')
#> NA files
pluralize('{c("file1", "file2", NA_character_)} file{?s}')
#> NA files

^{Created on 2024-08-15 with reprex v2.1.1}

I believe the expected behavior should be to have the vector displayed as is with the included NAs.

[rundel]

Seems like this is expected behavior, the underlying cause appears to be in

cli/R/glue.R

Lines 152 to 163 in 2ddcc1a

	collapse_head <- function(x, sep, sep2, last, trunc, width, ellipsis) {
	trunc <- max(trunc, 1L)
	x <- as.character(x)
	lnx <- length(x)
	# special cases that do not need trimming
	if (lnx == 0L) {
	return("")
	} else if (any(is.na(x))) {
	return(NA_character_)
	}

Seems undesirable to me, but maybe this behavior is necessary elsewhere outside of pluralize().

[gaborcsardi]

If the "pluralizing quantity" is a character vector, then the length of the vector is used. E.g. to be able to write:

fs <- c("foo", "bar")
cli_text('Reading file{?s}: {fs}')
#> Reading files: foo and bar

It does not matter if the character vector contains NA elements or not:

v <- c(NA_character_, "bar")
cli_text("Got element{?s}: {v}.")
#> Got elements: NA and bar.

[rundel]

The difference seems to be between pluralize() and cli_text() the latter is doing the correct processing while the former is collapsing to just NA.