From: shouman@comcast.net   
      
   "Steven G. Kargl"  writes:   
      
   > On Thu, 05 Feb 2026 21:07:23 +0200, Michael S wrote:   
   >   
   >> On Thu, 5 Feb 2026 18:38:48 -0000 (UTC)   
   >> Thomas Koenig  wrote:   
   >>   
   >>> ["Followup-To:" nach comp.lang.fortran gesetzt.]   
   >>> Michael S  schrieb:   
   >>> > On Thu, 5 Feb 2026 07:23:41 -0000 (UTC)   
   >>> > Thomas Koenig  wrote:   
   >>> >   
   >>> >   
   >>> >> Parsers are a field where Fortran does not shine, although   
   >>> >> allocatable characters have made this much easier.   
   >>> >   
   >>> > Another field where Fortran "does not shine" is handling of huge   
   >>> > uni-dimensional arrays.   
   >>>   
   >>> Why?   
   >>>   
   >>> If you are referring to pre-Fortran-90 days, you are also 35 years in   
   >>> the past :-)   
   >>>   
   >>> A bit more verbiage than required (usually, I would expect people to   
   >>> put their kind numbers into modules):   
   >>>   
   >>> module my_kinds   
   >>>   use iso_fortran_env, only: real64, int64 implicit none private public   
   >>>   :: ik, rk integer, parameter :: ik = int64, rk = real64   
   >>> end module my_kinds   
   >>>   
   >>> program main   
   >>>   use my_kinds implicit none integer(ik) :: sz real(rk), allocatable,   
   >>>   dimension(:) :: array read (*,*) sz allocate(array(sz))   
   >>> end program main   
   >>>   
   >>> works just fine.   
   >>>   
   >>>   
   >> Not that I understood all incantations, but I guess that above code   
   >> demonstrates that in New Fortran array indices are no longer limited to   
   >> 2**31.   
   >> Good to know.   
   >   
   > Technically, no Fortran standard limited the range of indicies to 2**31.   
   > Going back to Fortran 66, an array index has a type of integer, and in   
   > F66 there is only one integer kind.  The range of the integer is not   
   > specified in the standard.  The integer type could be a signed 16-bit,   
   > 32-bit, 48-bit, etc entity.  The Fortran standard does not specify   
   > implementation detail.   
      
   The size of the integer type was specified to be the same as the size of   
   the real type.  The real type need not be 32-bit floating point, but   
   that is certainly the way to bet.   
      
   --   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)