(j3.2006) (SC22WG5.5162) [ukfortran] [ Draft corrigendum 3]

[N.M. Maclaren]

On Dec 22 2013, Malcolm Cohen wrote:
>
>I (and others) disagree with your assertion that
>   REAL x(*)
>   PARAMETER(x = [ 1,2,3,4,5 ]) is unreasonable and should never be 
> allowed. Disallowing this would break one of our basic design rules for 
> the BNF of declarations, which is that one can specify attributes 
> independently. Yes it is a smaller edit to the standard, and would not 
> remove significant functionality ... but then deleting the PARAMETER 
> statement entirely would not remove significant functionality!

As someone who was not involved, I regrettably agree - though I think
that Bill is right that such code IS unreasonable.  Introducing
gratuitous restrictions should be considered only when they prevent
actual, fairly common, mistakes.  I can't see one that this blocks.

As someone who teaches both Fortran and C++, I really appreciate the
consistency and relatively tiny number of 'gotchas' of Fortran.
Fortran syntax may be verbose and horrible, but it is clean and
consistent (remember that I prefer Algol 68!)

I don't teach setting attributes separately but, if someone asks, I
can simply say that I don't advise it but it can be done.  That's it.
A couple of minutes, and it's covered.


Regards,
Nick Maclaren.