[J3] [EXTERNAL] Elemental type-bound functions not useful

Mon Mar 29 21:14:42 UTC 2021

Van,

I agree that the result is not as simple/elegant as one might hope.   But on the other hand, you have a situation that runs contrary to the spirit of type extension.   Wanting it both ways is natural, but expecting it is …

Most packages that I’m aware of (in _other_languages) that provide quad precision do it in the manner you have just described.       You either USE the entire package and bring in all the operators or you pick and choose with the equivalent of ONLY.

Cheers,

  *   Tom

From: J3 <j3-bounces at mailman.j3-fortran.org> on behalf of j3 <j3 at mailman.j3-fortran.org>
Reply-To: j3 <j3 at mailman.j3-fortran.org>
Date: Monday, March 29, 2021 at 5:05 PM
To: j3 <j3 at j3-fortran.org>
Cc: Van Snyder <van.snyder at sbcglobal.net>
Subject: Re: [J3] [EXTERNAL] Elemental type-bound functions not useful

On Mon, 2021-03-29 at 20:07 +0000, Clune, Thomas L. (GSFC-6101) wrote:
OK – I have a bit more time now.  Slowing down ….

So, you have a type and want an elemental procedure that acts on that type and returns something of the same type.  And you don’t want to allow that procedure to work on type extensions.     You therefore do _not_ want a type-bound procedure, as that is contrary to the intent of such constructs.    Rather, you just want plain old F90 elemental function.   It will pollute name space a bit more since it will need to be made public, but it otherwise seems to meet your requirements.    You can even overload it with similar functions for type extension now, since the equivalent of the passed-object dummy can now be declared with TYPE instead of CLASS.

Without a type-bound procedure, you can't have a generic binding. So instead of

use MyModule, only: MyType

you need

use MyModule, only: MyType, operator(+), operator(-), operator(*), operator(/), operator(**), operator(==), &
operator(/=), operator(>), operator(<), operator(>=), operator(<=), cos, sin, exp, log, abs, dot_product, matmul, ....

or just

use MyModule

The advantage of the former, which many organizations' style guides require, is that by looking only at the module where it's used, you can see where everything comes from, rather than using "grep" to find it.

From: "Clune, Thomas L. (GSFC-6101)" <thomas.l.clune at nasa.gov>
Date: Monday, March 29, 2021 at 3:12 PM
To: j3 <j3 at mailman.j3-fortran.org>, j3 <j3 at j3-fortran.org>
Cc: Van Snyder <van.snyder at sbcglobal.net>
Subject: Re: [J3] [EXTERNAL] Elemental type-bound functions not useful

Sorry was typing faster than thinking.  NON_OVERRIDABLE is not quite what you want.

From: "Clune, Thomas L. (GSFC-6101)" <thomas.l.clune at nasa.gov>
Date: Monday, March 29, 2021 at 3:07 PM
To: j3 <j3 at mailman.j3-fortran.org>, j3 <j3 at j3-fortran.org>
Cc: Van Snyder <van.snyder at sbcglobal.net>
Subject: Re: [J3] [EXTERNAL] Elemental type-bound functions not useful

If you don’t want that (and I think there are  times when you would), that is precisely what the NON_OVERRIDABLE attribute is for.    As I said before, you can’t overload the same name for these types of type-bound procedures anyway (because they are only distinguished by the return type), so just invent a unique name for each subclass that needs this elemental method and make each one NON_OVERRIDABLE.

From: J3 <j3-bounces at mailman.j3-fortran.org> on behalf of j3 <j3 at mailman.j3-fortran.org>
Reply-To: j3 <j3 at mailman.j3-fortran.org>
Date: Monday, March 29, 2021 at 2:48 PM
To: j3 <j3 at j3-fortran.org>
Cc: Van Snyder <van.snyder at sbcglobal.net>
Subject: Re: [J3] [EXTERNAL] Elemental type-bound functions not useful

On Mon, 2021-03-29 at 18:25 +0000, Clune, Thomas L. (GSFC-6101) wrote:
Why?    The restriction on elemental is that the return type not be polymorphic.  There is no restriction that the passed-object dummy be non-polymorphic.

If the dummy argument is polymorphic, the function can be invoked using an actual argument for which the type is an extension of the result type.

From: J3 <j3-bounces at mailman.j3-fortran.org> on behalf of j3 <j3 at mailman.j3-fortran.org>
Reply-To: j3 <j3 at mailman.j3-fortran.org>
Date: Monday, March 29, 2021 at 2:18 PM
To: j3 <j3 at mailman.j3-fortran.org>
Cc: Van Snyder <van.snyder at sbcglobal.net>
Subject: Re: [J3] [EXTERNAL] Elemental type-bound functions not useful

On Mon, 2021-03-29 at 12:59 +0000, Clune, Thomas L. (GSFC-6101) via J3 wrote:
I meant to add that the use case you are describing is essentially the poster child for templates vs OO.   When the return type varies with the input type, templates are a very natural fit, and trying to force OO is tricky at best.

This doesn't solve the problem unless type-bound functions are allowed not to have polymorphic passed-object dummy arguments.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.j3-fortran.org/pipermail/j3/attachments/20210329/e597f223/attachment-0003.htm>