R devel - Jun 2009 - Documentation/software inconsistency in `:` and seq

In 2.8.1/Windows:

According to ? :

     Details:
     For numeric arguments 'from:to' is equivalent to 'seq(from,
to)' ...

     Value:
     For numeric arguments, a numeric vector.  This will be of type
     'integer' if 'from' and 'to' are both integers and
representable
     in the integer type, otherwise of type 'numeric'....

The first claim is inconsistent with the second; it appears that the first
one is true and the second is false.

Actual behavior:

   storage.mode(1:2.5) => integer
   storage.mode(`:`(from=1L,to=2.5)) => integer   # just to be completely
explicit

But of course 2.5 is not an integer or equal to one.

This is the same behavior as seq:

   storage.mode(seq(1,2.5)) => integer
   storage.mode(seq(from=1L,to=2.5)) => integer

---------------------------------

On the other hand, according to ? seq

     Currently, the default method returns a result of type
'"integer"'
     if 'from' is (numerically equal to an) integer

This is not correct:
> seq(1e16,1e16+10)
 [1] 1e+16 1e+16 1e+16 1e+16 1e+16 1e+16 1e+16 1e+16 1e+16 1e+16
1e+16
> seq(1e16,1e16+10)-1e16
 [1]  0  0  2  4  4  4  6  8  8  8 10   # not enough precision to represent
all the distinct integers, OK
> storage.mode(seq(1e16,1e16+10))
[1] "double"

In this case, the ? : documentation is correct:

     This will be of type 'integer' if ... representable in the integer
type, otherwise of type 'numeric'.

	[[alternative HTML version deleted]]

Hi Stavros,
>>>>> "SM" == Stavros Macrakis <macrakis at
alum.mit.edu>
>>>>>     on Tue, 23 Jun 2009 15:53:30 -0400 writes:
    SM> In 2.8.1/Windows:
    SM> According to ? :

    SM> Details:
    SM> For numeric arguments 'from:to' is equivalent to
'seq(from, to)' ...

that *is* definitely true as    seq(from,to)  leads to  return(from:to)

    SM> Value:
    SM> For numeric arguments, a numeric vector.  This will be of type
    SM> 'integer' if 'from' and 'to' are both
integers and representable
    SM> in the integer type, otherwise of type 'numeric'....

    SM> The first claim is inconsistent with the second; it appears that the
first
    SM> one is true and the second is false.

{1st, 2nd ... where do you start counting ?}

    SM> Actual behavior:

    SM> storage.mode(1:2.5) => integer
    SM> storage.mode(`:`(from=1L,to=2.5)) => integer   # just to be
completely
    SM> explicit

    SM> But of course 2.5 is not an integer or equal to one.

Interestingly, because of an earlier posting on the topic,
two days ago (*), I have fixed the help page (and very slightly also
the code) of ":"  

[for R-devel (to be 2.10.0 in October) only, because R 2.9.1
 has already been in pre-release feature deep freeze ].


    SM> This is the same behavior as seq:

    SM> storage.mode(seq(1,2.5)) => integer
    SM> storage.mode(seq(from=1L,to=2.5)) => integer

{of course, as I mentioned above}

    SM> ---------------------------------

    SM> On the other hand, according to ? seq

    SM> Currently, the default method returns a result of type
'"integer"'
    SM> if 'from' is (numerically equal to an) integer

    SM> This is not correct:

Yes, seq(from, to)  {without further arguments} and "default"
'from', indeed returns from:to
and hence should probably either defer to the ":" help page
or copy-paste from it.


Martin Maechler, ETH Zurich

--------------------------------------------
(*)
------------------------------------------------------------------------
r48823 | maechler | 2009-06-22 17:57:05 +0200 (Mon, 22 Jun 2009)
Improve doc of ":" {and code very slightly, at the boundary of
MAX_INT}
------------------------------------------------------------------------



    >> seq(1e16,1e16+10)
    SM> [1] 1e+16 1e+16 1e+16 1e+16 1e+16 1e+16 1e+16 1e+16 1e+16 1e+16 1e+16
    >> seq(1e16,1e16+10)-1e16
    SM> [1]  0  0  2  4  4  4  6  8  8  8 10   # not enough precision to
represent
    SM> all the distinct integers, OK
    >> storage.mode(seq(1e16,1e16+10))
    SM> [1] "double"

    SM> In this case, the ? : documentation is correct:

    SM> This will be of type 'integer' if ... representable in the
integer
    SM> type, otherwise of type 'numeric'.

    SM> [[alternative HTML version deleted]]
    	^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

  {I'm amazed that you still did not find out how to avoid the
   above, e.g., by following the posting guide ...}

    SM> ______________________________________________
    SM> R-devel at r-project.org mailing list
    SM> https://stat.ethz.ch/mailman/listinfo/r-devel