C++Talk.NET C++ language newsgroups   Archives   FAQ   Search   Memberlist   Usergroups   Register   Profile   Log in to check your private messages   Log in  Imaginary friends          C++Talk.NET Forum Index -> C++ Language (Moderated) View previous topic :: View next topic   Author Message White Wolf Guest Posted: Sun Jul 31, 2005 1:15 am    Post subject: Imaginary friends I have a dilemma. It all boils down to this: // Header struct X { friend void f(X&); private: somestuff }; // Implementation static void friend f(X&x) { whatever } Why static? Because I do not like implementation details to leak out. Yes, the outside world will see a rogue f() in X's namespace, but calling it will lead to linker errors. And I only call f() inside the x.cpp, but: Is this legal? What does the standard say about it? g++ just says that f is not friend, while Comeau: "ComeauTest.c", line 6: error: external/internal linkage conflict with previous declaration static void f(X&x) {x.i=42;} -- WW aka Attila ::: Never wrestle with a pig: You both get all dirty, and the pig likes it. [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Tokyo Tomy Guest Posted: Sun Jul 31, 2005 11:47 am    Post subject: Re: Imaginary friends White Wolf wrote: Quote: I have a dilemma. It all boils down to this: // Header struct X { friend void f(X&); private: somestuff }; // Implementation static void friend f(X&x) { whatever } Why static? Because I do not like implementation details to leak out. Yes, the outside world will see a rogue f() in X's namespace, but calling it will lead to linker errors. And I only call f() inside the x.cpp, but: Is this legal? What does the standard say about it? g++ just says that f is not friend, while Comeau: "ComeauTest.c", line 6: error: external/internal linkage conflict with previous declaration static void f(X&x) {x.i=42;} -- IMHO, the "friend" is a key word of access control. The "friend void f(X&)" in class X declares that f is a friend to class X and is permitted to access to the private part of class X. The f in class X is not a member function of class X. The "friend" in "static void friend f(X&x)" has no meaning because it is not clear who is a friend to whom, and may be illegal. Therefore the code below should work, ... yes with VC 7.1, but no with g++. The g++ reported errors: storage class specifiers mismatch: the line (1) was declared 'extern', and the line (2) was declared 'static' I thought this was reasonable, and I added 'static' to the line (1), but g++ did not allow me to do so. I could not find out any solutions except to delete "static" at the line (2). [code] // Header struct X { friend static void f(X&); // (1) private: //somestuff int n; }; // Implementation static void /*friend*/ f(X&x) { // (2) //whatever x.n = 3; //access to private part of x } int main( ) { X x, y; f(x); //x.f(y); // error return 0; } [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Carl Barron Guest Posted: Sun Jul 31, 2005 11:48 am    Post subject: Re: Imaginary friends White Wolf <wolof (AT) freemail (DOT) hu> wrote: Quote: I have a dilemma. It all boils down to this: // Header struct X { friend void f(X&); private: somestuff }; // Implementation static void friend f(X&x) { whatever } Why static? Because I do not like implementation details to leak out. Yes, the outside world will see a rogue f() in X's namespace, but calling it will lead to linker errors. And I only call f() inside the x.cpp, but: Is this legal? What does the standard say about it? g++ just says that f is not friend, while Comeau: "ComeauTest.c", line 6: error: external/internal linkage conflict with previous declaration static void f(X&x) {x.i=42;} class X { void friendly_f_impl(); // impliment f here!! public: friend void f(X &x) {x.friendly_f_impl();} }; any rogue f's will create linker errors as f is an inline function whose definition is given in the class definition, any other is an error. violation of one defintion rule and even if the compiler does not inline f, it should be caught by the linker. [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Mirek Fidler Guest Posted: Sun Jul 31, 2005 11:53 am    Post subject: Re: Imaginary friends White Wolf wrote: Quote: I have a dilemma. It all boils down to this: // Header struct X { friend void f(X&); private: somestuff }; // Implementation static void friend f(X&x) { whatever } Why static? Because I do not like implementation details to leak out. Yes, Why not struct X { private: static void f(X&); }; ? Seems like natural solution to me. Of course, perhaps it does not match you scenario.... Mirek [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Richard Corden Guest Posted: Sun Jul 31, 2005 1:06 pm    Post subject: Re: Imaginary friends White Wolf wrote: [...] Quote: // Header struct X { friend void f(X&); private: somestuff }; // Implementation static void friend f(X&x) { whatever } [...] Quote: Is this legal? What does the standard say about it? As you've written it, to make it work you would have to add a storage specifier to the friend declaration. struct X { friend /* static */ void f(X&); }; // Implementation static void f(X&x) { } But 11.4/7 says you cannot have storage class specifiers on the friend declaration. You can probably change the code to: // h struct X; static void f(X&); struct X { friend void f(X&); }; // implementation static void f(X&) { } 11.4/3 says that the friend declaration will have the same linkage as the previous declaration so I believe this is formulation is OK. If I can be so bold as to make a comment though, I think that without copious commenting there is a strong danger of a future programmer thinking that you've incorrectly made the declarations static and so removing them. There's also the danger that compilers actually stop supporting namespace static declarations - however the relationship between unnamed namespace declarations and export might un-deprecate this. ;) Regards, Richard -- Richard Corden [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top White Wolf Guest Posted: Sun Jul 31, 2005 4:39 pm    Post subject: Re: Imaginary friends Carl Barron wrote: [SNIP] Quote: class X { void friendly_f_impl(); // impliment f here!! public: friend void f(X &x) {x.friendly_f_impl();} I cannot do that (my static f is a friend of another class as well, which musn't show up in X-es interface). And my question wasn't how else I can solve it, but it was "Is this legal? What does the standard say about it?" -- WW aka Attila ::: Always do right. This will gratify some people and astonish the rest. - Mark Twain [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top White Wolf Guest Posted: Sun Jul 31, 2005 4:40 pm    Post subject: Re: Imaginary friends Mirek Fidler wrote: [SNIP] Quote: Why not struct X { private: static void f(X&); }; ? Seems like natural solution to me. Of course, perhaps it does not match you scenario.... [SNIP] For exaclty that reason. And may I stress it again? My question was: Is this legal? What does the standard say about it? -- WW aka Attila ::: Sometimes a majority only means that all the fools are on the same side. [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top White Wolf Guest Posted: Sun Jul 31, 2005 4:41 pm    Post subject: Re: Imaginary friends White Wolf wrote: [SNIP] Quote: static void friend f(X&x) { [SNIP] Damn copy-paste. That friend should not be there. It wasn't in the code I have tested, I have no idea how did it get to my post. Sorry for that. (( -- WW aka Attila ::: Be consistent (but not all the time) [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top White Wolf Guest Posted: Sun Jul 31, 2005 4:41 pm    Post subject: Re: Imaginary friends Tokyo Tomy wrote: Quote: White Wolf wrote: I have a dilemma. It all boils down to this: // Header struct X { friend void f(X&); private: somestuff }; // Implementation static void friend f(X&x) { whatever } [SNIP] IMHO, the "friend" is a key word of access control. The "friend void f(X&)" in class X declares that f is a friend to class X and is permitted to access to the private part of class X. The f in class X is not a member function of class X. The "friend" in "static void friend f(X&x)" has no meaning because it is not clear who is a friend to whom, and may be illegal. There was no static friend in my code. Quote: I thought this was reasonable, and I added 'static' to the line (1), but g++ did not allow me to do so. I could not find out any solutions except to delete "static" at the line (2). [SNIP] Which is no solution to what I have asked, since I have asked if it is legal (not by some compiler, but by the standard) to have it static. -- WW aka Attila ::: Quit smoking on the anniversary of your birth, not the day of your death. - Corky Peterson [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top White Wolf Guest Posted: Sun Jul 31, 2005 4:45 pm    Post subject: Re: Imaginary friends Richard Corden wrote: [SNIP] Quote: You can probably change the code to: // h struct X; static void f(X&); struct X { friend void f(X&); }; // implementation static void f(X&) { } [SNIP] Yep, that seems to work. Except that one gets warnings (with gcc) in every translation unit including the header, that f"(X&) is declared but not used". That would require ifdefs ... I guess I am just better off leaving them non-static. -- WW aka Attila ::: What's the difference between roast beef and pea soup? Anyone can roast beef. [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Richard Corden Guest Posted: Mon Aug 01, 2005 10:26 am    Post subject: Re: Imaginary friends White Wolf wrote: Quote: Richard Corden wrote: [SNIP] You can probably change the code to: // h struct X; static void f(X&); struct X { friend void f(X&); }; // implementation static void f(X&) { } [SNIP] Yep, that seems to work. Except that one gets warnings (with gcc) in every translation unit including the header, that f"(X&) is declared but not used". That would require ifdefs ... I guess I am just better off leaving them non-static. I thought of a different way that I think achieves the same goal. struct SPECIAL_TYPE_ONLY_DEFINED_IN_X_CPP; struct X { friend void f (X&, SPECIAL_TYPE_ONLY_DEFINED_IN_X_CPP const &); }; // implementation struct SPECIAL_TYPE_ONLY_DEFINED_IN_X_CPP {}; void f (X&, SPECIAL_TYPE_ONLY_DEFINED_IN_X_CPP const &) { } void bar (X & x) { f (x, SPECIAL_TYPE_ONLY_DEFINED_IN_X_CPP ()); } What I like about this approach is that the name you give the type can help document for other developers that this function should not be used outside of this translation unit. Regards, Richard -- Richard Corden [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top kanze@gabi-soft.fr Guest Posted: Mon Aug 01, 2005 12:33 pm    Post subject: Re: Imaginary friends Richard Corden wrote: Quote: White Wolf wrote: [...] // Header struct X { friend void f(X&); private: somestuff }; // Implementation static void friend f(X&x) { [I presume that the presence of friend here is a copy/paste error.] Quote: whatever } [...] Is this legal? What does the standard say about it? As you've written it, to make it work you would have to add a storage specifier to the friend declaration. struct X { friend /* static */ void f(X&); }; // Implementation static void f(X&x) { } But 11.4/7 says you cannot have storage class specifiers on the friend declaration. You can probably change the code to: // h struct X; static void f(X&); struct X { friend void f(X&); }; // implementation static void f(X&) { } 11.4/3 says that the friend declaration will have the same linkage as the previous declaration so I believe this is formulation is OK. I may work, but it certainly isn't legal. Function f is a distinct entity in each translation unit in which it occurs, and the one definition rule says that all names in a class definition must bind to the same entity in all translation units. In practice, of course, it will almost certainly work. In fact, one wonders if there wasn't some intent for it to work, at least at one time. Otherwise, what's the point of §11.4/3: "A function first declared in a friend declaration has external linkage. Otherwise, the funtion retains its previous linkage." If the class definition is to be in a header file, the only way you can avoid a violation of the one definition rule is if all friends have external linkage. -- James Kanze GABI Software Conseils en informatique orientée objet/ Beratung in objektorientierter Datenverarbeitung 9 place Sémard, 78210 St.-Cyr-l'École, France, +33 (0)1 30 23 00 34 [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top White Wolf Guest Posted: Mon Aug 01, 2005 6:02 pm    Post subject: Re: Imaginary friends Richard Corden wrote: [SNIP] Quote: I thought of a different way that I think achieves the same goal. struct SPECIAL_TYPE_ONLY_DEFINED_IN_X_CPP; struct X { friend void f (X&, SPECIAL_TYPE_ONLY_DEFINED_IN_X_CPP const &); }; [SNIP] What I like about this approach is that the name you give the type can help document for other developers that this function should not be used outside of this translation unit. [SNIP] This sounds good. But it still "contaminates" the namespace, and ADL may find that function. I think if we could combine the above with in some way with SFINAE, we may just even eliminate that last issue. -- WW aka Attila ::: I started out with nothing... I still have most of it. [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Display posts from previous: All Posts1 Day7 Days2 Weeks1 Month3 Months6 Months1 Year Oldest FirstNewest First         C++Talk.NET Forum Index -> C++ Language (Moderated) All times are GMT Page 1 of 1   Jump to: Select a forum C/C++----------------C LanguageC Language (Moderated)C++ Language (Moderated)C++ language (comp.lang.c++)C++ language, library and standards Non-English C/C++ Newsgroups----------------C++ (German)C++ (French)  You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum Powered by phpBB © 2001, 2006 phpBB Group SEO toolkit © 2004-2006 webmedic.