(j3.2006) Interp F03/0091
Thu May 17 19:13:59 EDT 2007
The issue in the interp hinges on the term "named array" as used in
04-007:[78:16]. Lots of things are "arrays", including components
declared with a <component-array-spec> or a DIMENSION attribute
04-007:[51:16], an array constructor, or an array section [107:26]. At
least some of these are clearly not "named" arrays. By our usual
meaning of "name" as a lexical token, something like struct%array would
not qualify as a "name". One might reasonably conclude that
struct%array was not a "named array" even if it is an array.
The main point of the discussion on Friday was that the question in the
interp - whether len type parameters could be used in the
<explicit-shape-spec-list> of an array component declaration, was
already clearly answered by C442 [50:28-30]. (Well, ok, "clearly" might
not be a good description of that constraint's wording, but Jim's point
is still valid). If you decide that an array component is not a "named
array" then that's the end of the story, and there no issue. If an
array component is a "named array" then C542 and C442 are in conflict
and that's the real issue that needs an interp.
Aleksandar Donev wrote:
>On Thursday 17 May 2007 14:58, Van Snyder wrote:
>>If its dimensions aren't colons, what kind of array is it?
>>The only one that fits is an explicit shape array.
>I had thought that "named array" did not include components? It is not clear
>to me just by looking at the text. It would make most sense for every array
>to be one of explicit-shape, assumed-size or assumed-shape, including
>components. In which case we need to undo what we did Friday morning :-)
>J3 mailing list
>J3 at j3-fortran.org
Bill Long longb at cray.com
Fortran Technical Support & voice: 651-605-9024
Bioinformatics Software Development fax: 651-605-9142
Cray Inc., 1340 Mendota Heights Rd., Mendota Heights, MN, 55120
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the J3