`vec_as_names()` repairs `NA_character_` to `"...1"`

[JesseAlderliesten]

vec_as_names() repairs NA_character_ to "...1": vctrs::vec_as_names(NA_character_, repair = "universal_quiet").
[1] "...1"

I would have expected ".NA_character_" as output, just like vctrs::vec_as_names(TRUE, repair = "universal_quiet") that currently returns ".TRUE".

That would also be consistent with the current behaviour for (quoted) "NA_character_": vctrs::vec_as_names("NA_character_", repair = "universal_quiet") returns the expected ".NA_character_", just like vctrs::vec_as_names("TRUE", repair = "universal_quiet") returns ".TRUE".

I use vctrs version 0.7.3.

[DavisVaughan]

The current behavior is expected and is documented in the universal names section of ?vec_as_names

It isn't just ..1, it uses the position of the NA

vctrs::vec_as_names(c(NA_character_, NA_character_), repair = "universal_quiet")
#> [1] "...1" "...2"

It's a recognition that there is no usable information in that slot, so instead we encode via position.

A quoted "NA_character_" (a normal string) is something entirely different from NA_character_ (a character missing value), so I don't think it is fair to compare those