Renaming module procedures
in [Fortran]
|
Prev: ENTRY statement
Next: Namelist/module-question
From: helvio on 30 Sep 2009 22:59 On Oct 1, 1:59 am, nos...(a)see.signature (Richard Maine) wrote: > I will continue to insist that > this does not "rename" the function. This gives the function a generic > name in addition to its specific one. That is not the same thing as > renaming. It can seem similar on the surface in limited circumstances. I > do see that you quoted "rename". I'm just emphasizing that the quotes > are very much needed. I don't like referring to it that way because as > soon as you do so, someone will ignore the quotes and think it really is > a renaming... and then they will wonder why it doesn't work like they > would expect. Oh, that's true... I already knew that generic interfaces only give you aliases for the original procedures, which can still be called by their original names, but somehow I forgot that. So then it's very different from the true renaming that happens with '=>' in USE statements. So, if I understood everything correctly: I'm perfectly allowed to construct a generic INTERFACE to a given MODULE PROCEDURE (PROC1) inside another of the same scope (PROC2) -- modulo compiler bugs --, but I cannot re-use its original name (PROC1) to any other purpose than to call the very same procedure? In fact, that sounds reasonable. But is there a fundamental reason why truly renaming a procedure inside another of the same scope should not be allowed by a compiler? Something like a generalized USE? I know that it can't be done with actual compilers conforming the actual standards, but is there any fundamental logical obstruction for it to be a future standard feature of Fortran? This is just an academic question, I just want to understand how the compiling works, for this case. -- helvio
From: Richard Maine on 30 Sep 2009 23:36 helvio <helvio.vairinhos(a)googlemail.com> wrote: > In fact, that sounds reasonable. But is there a fundamental reason why > truly renaming a procedure inside another of the same scope should not > be allowed by a compiler? Something like a generalized USE? I know > that it can't be done with actual compilers conforming the actual > standards, but is there any fundamental logical obstruction for it to > be a future standard feature of Fortran? This is just an academic > question, I just want to understand how the compiling works, for this > case. I can't think of any fundamental reason why there couldn't be such a feature. I'm not a compiler writer, so I'm certainly not the one to say definitively, but it doesn't even seem to me like it would necessarily be particularly difficult. There is already a separate scope for names in a module procedure. There are already ways that a name can be "blocked" from host association into that scope. I don't see anything here that seems like it would greatly upset existing ways of doing things or introduce particularly new concepts. That doesn't mean I think such a feature is particularly likely to come about, though. I suspect it would be on the wrong end of a lot of the tradeoffs for new features. There are always zillions of new features that people are asking for, but there are finite resources to do them. Plus, even if resources weren't an issue, adding every feature that anyone asks for would make the language a huge mess that nobody could fully understand. Some criticise it as being too much that way already. So one tends to need to show fairly strong justification for new features. That it might prove handy in a few special cases usually isn't enough justification. -- Richard Maine | Good judgment comes from experience; email: last name at domain . net | experience comes from bad judgment. domain: summertriangle | -- Mark Twain
From: Richard Maine on 30 Sep 2009 23:36
helvio <helvio.vairinhos(a)googlemail.com> wrote: > On Oct 1, 1:59 am, nos...(a)see.signature (Richard Maine) wrote: > > > I will continue to insist that > > this does not "rename" the function. > Oh, that's true... I already knew that generic interfaces only give > you aliases for the original procedures, which can still be called by > their original names, but somehow I forgot that. It is more than that. A generic name is not just an alias. It is a different kind of thing. There are places where you can use a specific name, but you can't use the generic name. I already mentioned one: you can't pass a generic as an actual argument. There are others. -- Richard Maine | Good judgment comes from experience; email: last name at domain . net | experience comes from bad judgment. domain: summertriangle | -- Mark Twain |