(j3.2006) Select Rank

[Malcolm Cohen]

Bill Long writes:
>An associate name is less disruptive, since it is not declared.

Indeed, one would want to do this identically to associate names.  We don't need 
to put the word ASSOCIATE in there, after all SELECT TYPE doesn't have it.

>  Whether we want to propose this for F2020 is another question.  Does it 
> really solve an existing problem?

Well, extremely few existing codes even *have* assumed-rank.  One might 
reasonably think that once support for assumed-rank becomes more widespread, 
people will start using it.  One might reasonably think that once people start 
using it, they will want to do stuff in Fortran without having to write a 
(grungy) C wrapper to do the job.

There are certainly technical wrinkles, principally to do with how to handle 
assumed-size arrays without blowing both feet off, but the normal case looks 
fairly easy to get right.  Once we have a technically-solid proposal I think it 
would be worth considering.

Cheers,
-- 
................................Malcolm Cohen, Nihon NAG, Tokyo.