Re: [OctDev] Can we implement isinteger this way?

[Quentin S. <qsp...@ie...>]

William Poetra Yoga H wrote:

>--- Andy Adler <ad...@nc...> wrote:
>  
>
>>Are you happy with the way this function will handle large
>>exponents where the interger part is lost? For example
>>
>>octave:6> isinteger(1e10/3)
>>ans = 0
>>octave:7> isinteger(1e20/3)
>>ans = 1
>>
>>Andy
>>
>>    
>>
>
>Well... I'm not. But I can't think of a better way...
>Btw, actually 1e20/3 should be treated as an integer anyway (or a floating
>point with zero behind the decimal point) because Octave can't handle it. I
>meant to use this function to check for integer count arguments passed to
>functions, and I think most people would do the same thing, so...
>
>In the meantime, I've modified it a bit (thanks Paul):
>
>function retval = isinteger(A)
>## -*- texinfo -*-
>## @deftypefn {Function File} {} isinteger (@var{A})
>## Return 1 if @var{A} is an integer array.  Otherwise, return 0.
>## @end deftypefn
>##
>## @seealso{isa, isnumeric, isfloat}
>
>if (nargin ~= 1)
>  usage("isinteger(A)");
>end
>
>retval = ismatrix(A) && all((round(A) == A)(:));
>
>And now I also have an isfloat function:
>
>function retval = isfloat(A)
>## -*- texinfo -*-
>## @deftypefn {Function File} {} isfloat (@var{A})
>## Return 1 if @var{A} is a floating point array.  Otherwise, return 0.
>## @end deftypefn
>##
>## @seealso{isa, isinteger, isnumeric}
>
>if (nargin ~= 1)
>  usage("isfloat(A)");
>end
>
>retval = ismatrix(A) && ~all((round(A) == A)(:));
>
>
>  
>



If we're aiming for compatibility with the Matlab functions isinteger 
and isfloat, we should look at what they do:
 >> isinteger(1)
ans =
     0

 >> isinteger(int8(1))
ans =
     1

So, the outputs of these functions are based not on whether the contents 
of the array happen to be integers, which can happen with a double 
array, but on whether they are _stored_ as integers. In addition to the 
compatibility issue, checking all elements of an array to see if they 
are an integer could be much slower with a large array than just 
checking the array type. Given this, it would seem that the correct 
implementation of isinteger would involve parsing the output of the 
typeinfo function.

-Quentin