R devel - Sep 2021 - formatC(character()) returns length 1 result, but is documented otherwise

Hi all,

I believe I have either found a small bug, or a possible inconsistency in
documentation. formatC() returns a length 1 result if given a length 0
character() as input.

formatC(character())
#> [1] ""

But the return value documentation states that it returns: "A character
object of same size and attributes as x".

I'd love for this to return a size 0 result here, consistent with the docs
and my mental model of size stability for this function.

Here is where this happens (it is explicitly hard coded):
https://github.com/wch/r-source/blob/79298c499218846d14500255efd622b5021c10ec/src/library/base/R/format.R#L149

Thanks,
Davis Vaughan

	[[alternative HTML version deleted]]

>>>>> Davis Vaughan 
>>>>>     on Mon, 13 Sep 2021 16:35:47 -0400 writes:
    > Hi all,

    > I believe I have either found a small bug, or a possible inconsistency
in
    > documentation. formatC() returns a length 1 result if given a length 0
    > character() as input.

    > formatC(character())
    > #> [1] ""

    > But the return value documentation states that it returns: "A
character
    > object of same size and attributes as x".

    > I'd love for this to return a size 0 result here, consistent with
the docs
    > and my mental model of size stability for this function.

    > Here is where this happens (it is explicitly hard coded):
    >
https://github.com/wch/r-source/blob/79298c499218846d14500255efd622b5021c10ec/src/library/base/R/format.R#L149

    > Thanks,
    > Davis Vaughan

    > [[alternative HTML version deleted]]

This may well be a "historical artefact" - definitely
older than R 1.0.0. ... and yes I've known much less about S and
its new dialect R, back in 1998.  ;-)

You are right in all you say, and my mental model corresponds to
yours,  so we will almost surely change this (for R-devel;
possibly even consider "back" porting to R 4.1.1 patched).

Thank you for the report and suggestion,
Martin