Re: write recursive call

[Richard E Maine <Richard dot Maine at nasa dot gov>]

On Nov 29, 2005, at 12:04 AM, Erik Edelmann wrote:

> On Tue, Nov 29, 2005 at 08:36:49AM +0100, Pierre-Matthieu anglade 
> wrote:
> > Then it seems that recursive IO is not something desirable. Right?
>
> In my opinion, recursive IO as allowed by the F2003 standard
> (i.e. where the "lower-level" IO is to/from internal files), is highly
> desirable.

I agree. And so did enough of the standard's committee in order to have 
it added to the f2003 standard. I don't think there has ever been 
significant doubt that it was desirable - that is to say useful. The 
only questions have been as to exactly how desirable it was, and thus 
how much investment it was worth, versus what the cost of 
implementation was estimated to be.

In times gone by, there was concern that it would cause serious 
implementation problems. In f77, recursion in any form was disallowed 
by the standard and was difficult to implement well on some 
architectures. Even in early f90 days, there was concern that some 
run-time I/O support libraries might not be able to handle it. Although 
the feature would have been desirable, it wasn't deemed so important as 
to justify requiring complete rewrite of the run-time support libraries 
of some extant compilers.

Today, pretty much all run-time support libraries can handle it anyway, 
making the cost negligable. Furthermore, the f2003 feature of 
user-defined derived-type I/O was going to force the issue; that 
feature is almost useless without recursive I/O.

--
Richard Maine                |  Good judgment comes from experience;
Richard.Maine@nasa.gov       |  experience comes from bad judgment.
                            |        -- Mark Twain