[R] min(NA,"bla") != min("bla", NA)

[Rolf Turner]

On 09/27/13 11:07, Duncan Murdoch wrote:
> On 13-09-26 5:32 PM, Rolf Turner wrote:
>>
>> Just to add to the confusion, on my system I get NA --- which I
>> understand to be
>> the correct value --- from all of min(NA,"bla"), min("bla",NA),
>> min(c(NA,"bla")), and
>> min(c("bla",NA)).  When I append the argument na.rm=TRUE to each of the
>> calls,
>> I get "bla" from each.
>>
>> So, no bug in my system.
>
> I've just built a 3.0.1 version, and I definitely see the bug there. 
> What do you get from these expressions?
>
> str(min(NA, "bla"))
> str(min("bla", NA))
> str(min(NA_character_, "bla"))
>
> I get
>
> > str(min(NA, "bla"))
>  int NA
> > str(min("bla", NA))
>  chr "bla"
> > str(min(NA_character_, "bla"))
>  chr "bla"
>
> on both Windows and OSX R 3.0.1.  After today's patch, I get
>
>  chr NA
>
> for all three.

I get:

 > str(min(NA, "bla"))
  int NA
  > str(min("bla", NA))
  chr NA
 > str(min(NA_character_, "bla"))
  chr NA

Which looks to me to be as it should be.  How come my system's so good
compared to others? :-)

     cheers,

     Rolf