Prev: computing Bernoulli numbers
Next: Which direction is Fortran going?
From: James Giles on 1 Feb 2008 16:41
James Giles wrote:
....
> The following has been discussed before and I think was determined
> to be non-standard for the same reason:
>
> Real :: X(5), Y(size(x))

By Steve's response it's clear that I mis-remembered the case.
The one that I recalled must have been something like:

Real :: X(4) = size(X)

I still stand by the observation:

> Even if such things were standard, I would avoid then for legibility
> reasons.

--
J. Giles

"I conclude that there are two ways of constructing a software
design: One way is to make it so simple that there are obviously
no deficiencies and the other way is to make it so complicated
that there are no obvious deficiencies." -- C. A. R. Hoare


From: Tobias Burnus on 1 Feb 2008 04:55
On Jan 31, 10:23 pm, Steve Lionel <steve.lio...(a)intel.com> wrote:
> On Jan 31, 7:38 am, Tobias Burnus <bur...(a)net-b.de> wrote:
> > Is the following code valid or not?
[...]
> We've been here before. Here's the text of Interp 00090 against F95
> discussing this issue. The text was clarified in F2003.

Thanks for the pointer. Then I initially read the standard correctly
but as all compilers allowed more - even those regarded as picky, I
thought that I misread the standard and tried to make sense of of it
- with little success.

> I'll comment for the record that Intel Fortran doesn't get all the cases
> right either in that we accept and do not warn about certain non-standard
> usages. We intend to fix that in a future release.

Ditto for gfortran.


On Jan 31, 10:41 pm, "James Giles" <jamesgi...(a)worldnet.att.net>
wrote:
> James Giles wrote:
> I still stand by the observation:
> > Even if such things were standard, I would avoid then for legibility
> > reasons.

Well, my question was motivated by deciding whether gfortran needs to
be fixed (for -std=f95/f2003 at least). I did and do not intent to use
such constructs.


Although I have to admit that I find things like the following rather
convenient:

real :: a(1,1,1,2) = reshape( [1.0, 2.0], shape(a) )

Thanks again Steve and James!

Tobias
From: James Giles on 1 Feb 2008 13:43
Tobias Burnus wrote:
....
> Although I have to admit that I find things like the following rather
> convenient:
>
> real :: a(1,1,1,2) = reshape( [1.0, 2.0], shape(a) )

Yes it is irritating that it doesn't work. It's a peculiar
asymmetry that [ARRAY] is valid regardless of the rank
of ARRAY, but ARRAY = [anything] is not valid if
ARRAY is not rank 1. I think that array constructors
should interoperate with arrays of any rank provided
the size is correct. Elements line up in array element
order, of course.

--
J. Giles

"I conclude that there are two ways of constructing a software
design: One way is to make it so simple that there are obviously
no deficiencies and the other way is to make it so complicated
that there are no obvious deficiencies." -- C. A. R. Hoare


 | 
Pages: 1
Prev: computing Bernoulli numbers
Next: Which direction is Fortran going?