[log in to unmask] wrote:
>> [log in to unmask] wrote:
>>> James Giles wrote:
>>>
>>>> 4. provide a way to write code that doesn't need KIND
>>>> specifiers on literals in the first place.
>>>
>>> Unfortunately, Fortran still allows procedures with implicit interface.
>>> So how does the literal get the right kind to go with the dummy
>>> argument?
>>
>> The same way it gets it now: the user is required to insure the
>> match. You response is hardly even worth of notice.
>
> Maybe something more than hand waving would be in order here. If one
> doesn't put the kind type parameter on the literal -- which is after
> all what James is advocating -- how does the user "insure the match?"
Use the explicit KIND in that one (rare) case. Although it's more
likely that a rewrite of the external is what the programmer really
intends.
Eliminating all the explicit KIND specifiers is desired as an alternative
to going through the program robotically putting KIND specifiers on all
the literals: a process prone to inattention and error, and in any case it
makes the code less legible to the next person that reads it. If you go
through that, isn't the call to the external that doesn't have an explicit
interface just as likely to be wrong as if you had specified somewhere
that all literals were to be interpreted as such-and-cuch a KIND? I think
the answer is rather obviously yes. Hence, the issue of how to handle
implicitly interfaced procedures is irrelevant to the thread. (It's apparently
now a gratuitous insult to even point that out. Hmm. I suppose it isn't
a gratuitous insult to characterize someone else's remarks as a "whine".
Hmm. Not where I grew up.)
--
J. Giles
|