(j3.2006) Intrinsic modules

[Craig Dedo]

Van and Everyone Else:
	No, there is no way to tell a compiler to prefer intrinsic modules
to nonintrinsic modules of the same names.  In fact, such behavior would
directly contradict the normative text of the Fortran 2003 standard and the
current draft.  The relevant text is in section 11.2.1 in Fortran 2003
(03-007) [254:8-10] and section 11.2.2p3 in the current draft (09-007r3)
[273:27-28].

[Begin quote]
A use-stmt without a module-nature provides access either to an intrinsic or
to a nonintrinsic module. If the module-name is the name of both an
intrinsic and a nonintrinsic module, the nonintrinsic module is accessed.
[End of quote]

	Obviously, any change in this preference rule would create a
backwards incompatibility with Fortran 2003.

	I find this preference rule to be somewhat surprising.  I remember
that when we discussed the IEEE intrinsic modules, I raised a question about
a particular example in one of John Reid's drafts.  He had left out the ",
INTRINSIC" module-nature in the example.  The answer I got was that it did
not matter; as proposed, the compiler would pick the intrinsic module if the
module-nature was left off the USE statement.

	Does anyone remember why J3 chose to prefer nonintrinsic modules
over intrinsic modules of the same name?

	FWIW, if J3 decides to develop a way to state the opposite
preference, then it should be done with explicit syntax at the procedure or
module levels, immediately before the relevant USE statements.  This would
be similar to IMPLCIT or IMPLICIT NONE statements overriding default
implicit typing.

Sincerely,
Craig T. Dedo
17130 W. Burleigh Place
P. O. Box 423                  Mobile Phone:  (414) 412-5869
Brookfield, WI   53008-0423    E-mail:  <craig at ctdedo.com>
USA
Linked-In:  http://www.linkedin.com/in/craigdedo


-----Original Message-----
From: j3-bounces at j3-fortran.org [mailto:j3-bounces at j3-fortran.org] On Behalf
Of Van Snyder
Sent: Friday, November 20, 2009 20:08
To: j3
Subject: (j3.2006) Intrinsic modules

For the compiler vendors:

Is there a way to tell your compiler to prefer intrinsic modules
(hopefully a list of specific ones) to a nonintrinsic one of the same
name where <module-nature> is not specified in a USE statement
(contradicting 09-007r3:11.2.2p3), even if there's a nonintrinsic one of
the same name in the module search paths?  Or is the only way to do it
simply not to have such a one in any directory in the module search
paths?

I have a 1/3-million line program in which we (perhaps unwisely) have
"use IEEE_Arithmetic" in most modules.  For platforms that don't offer
the intrinsic module yet, we've written our own.  It would be quite
painful to go through the entire program and put either that or "use,
intrinsic :: IEEE_Arithmentc" in place of what we have, controlled by a
preprocessor variable.  Perhaps we should have put "use
My_IEEE_Arithmetic" in each place, so we could put "use, intrinsic ::
IEEE_Arithmetic" in that module.

If we were to add to the list in 16.3.1p2:

  o  is the name of the module or program in which the local identifier
     appears

that would also solve the problem (and several others).  This wouldn't
cause any trouble (for the standard) because module names only appear in
MODULE, END MODULE, USE, and SUBMODULE statements.

-- 
Van Snyder                    |  What fraction of Americans believe 
Van.Snyder at jpl.nasa.gov       |  Wrestling is real and NASA is fake?
Any alleged opinions are my own and have not been approved or
disapproved by JPL, CalTech, NASA, the President, or anybody else.

_______________________________________________
J3 mailing list
J3 at j3-fortran.org
http://j3-fortran.org/mailman/listinfo/j3