(j3.2006) Interesting F2003-ism
Thu Feb 22 19:38:15 EST 2007
Dan Nagle wrote:
> You're simply making the point that a better abstraction for arrays
> would be helpful in some situations.
I agree...with Van etc...but I was not talking about changing anything
in Fortran 2008 (which can keep its 20-year old array abstractions if it
thinks that's sufficient...at least it has co-arrays which is a major
step forward in itself).
Using the (proposed) Interop TR it is *trivial* to write a single C
routine that handles arrays of different TKRs (this will usually be a
wrapper itself, calling lower level routines or Fortran-ignorant C
libraries). Is this a bad thing and something that we should forbid by
rules in clause 16 that have a completely different intention???? Or
should we finally give this ability to all the computational scientists
that are trying to develop Fortran interfaces to their libraries, hating
Fortran more and more the more they try to interoperate with it?!?
I am not asking that the same (write a single routine that handles
arrays of different TKRs) be true in Fortran. I don't believe Interop is
a symmetric operation. It is the attempt to make sure that everything
written in interoperable C can be written in Fortran and vice versa that
is the pointless excercise in needless restrictions.
More information about the J3