C++Talk.NET C++ language newsgroups   Archives   FAQ   Search   Memberlist   Usergroups   Register   Profile   Log in to check your private messages   Log in  Omitting unspecialized template function definitions.          C++Talk.NET Forum Index -> C++ Language (Moderated) View previous topic :: View next topic   Author Message Al Guest Posted: Sat May 12, 2007 9:10 am    Post subject: Omitting unspecialized template function definitions. Hi, Is it legal to declare a template function with no unspecialized definition, only definitions for the used specializations? template <class T> T foo(); template <> int foo() { ...; } template <> float foo() { ...; } It seems to work on a couple of platforms. This way there's no need to throw "not implemented" at runtime, since using a wrong version would always be caught at link time (presumably). Thanks. -Al- -- [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Rune Allnor Guest Posted: Sat May 12, 2007 7:04 pm    Post subject: Re: Omitting unspecialized template function definitions. On 12 May, 11:11, Al <t...@haik.us> wrote: Quote: Hi, Is it legal to declare a template function with no unspecialized definition, only definitions for the used specializations? template <class T T foo(); template <> int foo() { ...; } template <> float foo() { ...; } It seems to work on a couple of platforms. This way there's no need to throw "not implemented" at runtime, since using a wrong version would always be caught at link time (presumably). Wouldn't what you ask for be taken care of by virtual functions? If this can be done your way, you would have to declare at coding time exactly what types to implement somewhere -- which is a requirement of virtual functions as well. The basic philosophy behind template that "the type of the argument is unknown at coding time" is broken, since you at coding time have this finite and well-specified list of types you allow. I can't really see that templates are the tool for the job? Or am I missing some very subtle detail in either the problem specification or the proposed solution? Rune -- [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Sebastian Redl Guest Posted: Sat May 12, 2007 7:54 pm    Post subject: Re: Omitting unspecialized template function definitions. On Sat, 12 May 2007, Al wrote: Quote: Is it legal to declare a template function with no unspecialized definition, only definitions for the used specializations? Yes, absolutely legal and always caught at link time. (Unless, of course, you have a output type that isn't linked completely, such as static libraries or, I believe, POSIX shared objects.) Sebastian Redl -- [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Al Guest Posted: Sat May 12, 2007 7:54 pm    Post subject: Re: Omitting unspecialized template function definitions. Hi, Rune Allnor wrote: <snip> Quote: Wouldn't what you ask for be taken care of by virtual functions? If this can be done your way, you would have to declare at coding time exactly what types to implement somewhere -- which is a requirement of virtual functions as well. The basic philosophy behind template that "the type of the argument is unknown at coding time" is broken, since you at coding time have this finite and well-specified list of types you allow. I don't think that "basic philosophy" is comprehensive of all template uses. I think of templates as a general-purpose meta-programming tool. Quote: I can't really see that templates are the tool for the job? Or am I missing some very subtle detail in either the problem specification or the proposed solution? I don't think the task at hand is particularly subtle. I am using a boost::variant (a true discriminated union) with a finite set of types (e.g. float, int, string and a few more). I don't think invoking the full OO machinery is necessary for dispatching these types when a simple template (with the help of boost::static_visitor) will work. I just wanted to know if the technique is a legal one. Thanks, -Al- -- [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Daniel Krügler Guest Posted: Sun May 13, 2007 2:39 am    Post subject: Re: Omitting unspecialized template function definitions. On 12 Mai, 21:54, Sebastian Redl <e0226...@stud3.tuwien.ac.at> wrote: Quote: On Sat, 12 May 2007, Al wrote: Is it legal to declare a template function with no unspecialized definition, only definitions for the used specializations? Yes, absolutely legal and always caught at link time. (Unless, of course, you have a output type that isn't linked completely, such as static libraries or, I believe, POSIX shared objects.) Nope, from the standpoint of the language this is illegal and causes UB, because it violates the ODR. The concept of "caught at link time" does not exist in C++, although you are right that this is a typical implementation behaviour. I hope that my corresponding answer will come up soon, where I provide more details. Btw.: Thanks for correcting my error to qualify a class invariant as a postcondition in another posting. Greetings from Bremen, Daniel Krügler -- [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Alberto Ganesh Barbati Guest Posted: Tue May 15, 2007 12:38 am    Post subject: Re: Omitting unspecialized template function definitions. Richard Corden ha scritto: Quote: Al wrote: template <class T T foo(); template <> int foo() { ...; } template <> float foo() { ...; } [As a BTW - this exact example is 'strictly' ill-formed, since 'T' can only be deduced from the function argument types - however AFAIK most, if not all, compilers do the right thing(tm). ] Why do you say it's ill-formed? Could you please elaborate? I think it is perfectly legal. Of course, the only way to invoke such a template is to explicitly specify the template parameters, as in foo<int>(), but this fact doesn't make the *definitions* ill-formed. Ganesh -- [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Abhishek Padmanabh Guest Posted: Wed May 16, 2007 3:41 pm    Post subject: Re: Omitting unspecialized template function definitions. On 14 May, 18:58, Richard Corden <richard_cor...@hotmail.com> wrote: Quote: // a2.cc template <class T> T foo (); template <> int foo<int>() {} void b2 () { foo<int> (); } The above is either 'ill-formed no diagnostic required', or just undefined behaviour. What's the problem with b2 function in the part a2.cc? -- [ 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: Wed May 16, 2007 5:19 pm    Post subject: Re: Omitting unspecialized template function definitions. Abhishek Padmanabh wrote: Quote: On 14 May, 18:58, Richard Corden <richard_cor...@hotmail.com> wrote: // a2.cc template <class T> T foo (); template <> int foo<int>() {} void b2 () { foo<int> (); } The above is either 'ill-formed no diagnostic required', or just undefined behaviour. What's the problem with b2 function in the part a2.cc? I'm not an expert on what actually happens regarding implicit instantiation matching by linkers etc, however, I think the problem is that a1.o and a2.o will both have a different definitions for 'foo<int>'. This then leads to: 1) linker error, or 2) definition from a1.cc being used, or 3) definition from a2.cc being used, or 4) other. The standard just says that the explicit specialization shall be declared before the first use of that specialization would cause an implicit instantiation to take place, no diagnostic required.[14.7.3/6] 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 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.