C++Talk.NET C++ language newsgroups   Archives   FAQ   Search   Memberlist   Usergroups   Register   Profile   Log in to check your private messages   Log in  Reusable test for presence of a nested typedef?          C++Talk.NET Forum Index -> C++ Language (Moderated) View previous topic :: View next topic   Author Message Glen Low Guest Posted: Mon Oct 18, 2004 4:56 pm    Post subject: Reusable test for presence of a nested typedef? Warning, SFINAE territory ahead... One typical formulation of a test for a nested typedef result_type would be, based on Dewhurstian mechanics: enum yes_no { yes = 1, no = 2 }; template <typename T> char (*may_have_result_type (typename T::result_type*)) [yes]; template <typename T> char (*may_have_result_type (...)) [no]; template <typename T> stuct has_result_type { enum { value = sizeof (*may_have_result_type <T> (NULL)) = yes }; }; So to test if a type U has a result_type, we simply do has_result_type <U>::value. But being a guy who loves reuse, especially when he sees his own code littered with almost similar instances of the above, I wonder whether there's a general way of doing this (without bashing clients with a macro) i.e. has_typedef <U, result_type>::value OR has_typedef <typename U::result_type>::value ? Cheers, Glen Low, Pixelglow Software www.pixelglow.com [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Terje Slettebø Guest Posted: Mon Oct 18, 2004 8:31 pm    Post subject: Re: Reusable test for presence of a nested typedef? "Glen Low" <glenlow (AT) pixelglow (DOT) com> wrote Quote: Warning, SFINAE territory ahead... And if you fail to understand it, don't treat it as an error. ;) Quote: One typical formulation of a test for a nested typedef result_type would be, based on Dewhurstian mechanics: enum yes_no { yes = 1, no = 2 }; template <typename T> char (*may_have_result_type (typename T::result_type*)) [yes]; template <typename T> char (*may_have_result_type (...)) [no]; template <typename T> stuct has_result_type { enum { value = sizeof (*may_have_result_type <T> (NULL)) = yes }; }; So to test if a type U has a result_type, we simply do has_result_type U>::value. But being a guy who loves reuse, especially when he sees his own code littered with almost similar instances of the above, I wonder whether there's a general way of doing this (without bashing clients with a macro) i.e. has_typedef <U, result_type>::value OR has_typedef <typename U::result_type>::value ? I don't think it can be done without a macro. Many good people have tried. In the Concept Traits library ([url]http://www.neoscientists.org/boost_library_proposals.html)[/url], where Tobias Schwinger now is co-author, I used macros for detection where you need a name, as there are headers (in the type_traits_ext directory of the zip-file there. Sorry, the docs are only in the archive at the moment) for detection of: - Non-static members (variables and functions) - Static members (variables and functions) - Nested classes or typedefs - Nested class templates Regards, Terje [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top John M. Dlugosz Guest Posted: Tue Oct 19, 2004 4:22 pm    Post subject: Re: Reusable test for presence of a nested typedef? [email]glenlow (AT) pixelglow (DOT) com[/email] (Glen Low) wrote in message news:<9215d7ac.0410171513.42ac2620 (AT) posting (DOT) google.com>... Quote: But being a guy who loves reuse, especially when he sees his own code littered with almost similar instances of the above, I wonder whether there's a general way of doing this (without bashing clients with a macro) i.e. What happens when you put those templates as template members of a class which it itself a template on the type name in question? You imply that it didn't work? --John [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Glen Low Guest Posted: Tue Oct 19, 2004 5:59 pm    Post subject: Re: Reusable test for presence of a nested typedef? Quote: But being a guy who loves reuse, especially when he sees his own code littered with almost similar instances of the above, I wonder whether there's a general way of doing this (without bashing clients with a macro) i.e. has_typedef <U, result_type>::value OR has_typedef <typename U::result_type>::value ? I don't think it can be done without a macro. Many good people have tried. In the Concept Traits library ([url]http://www.neoscientists.org/boost_library_proposals.html)[/url], where Tobias Schwinger now is co-author, I used macros for detection where you need a name, as there are headers (in the type_traits_ext directory of the zip-file there. I thought so. In the end it wasn't too bad. I was using the presence or absence of a typedef within an enable_if on an partial specialization, it turns out all I needed was: template <typename T> struct exists { enum { value = true }; }; Simple, no? Why does it work... well in the partial specialization I do: template <typename T, typename Enable = void> class thing; template <typename T> class thing <T, typename enable_if ::type> { ... }; So in classic SFINAE style, this will be silently dropped from the overload set when typename T::result_type doesn't exist. The exists + enable_if combination only serves to form the type "void" when result_type does exist. Of course the exists template is of no use in a non-substitutionary context, and it might yet blow up in my face on a non-gcc compiler, but I'm fairly sure it will work on a Standard compliant one. Cheers, Glen Low, Pixelglow Software www.pixelglow.com [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Larry Evans Guest Posted: Wed Oct 20, 2004 4:26 pm    Post subject: Re: Reusable test for presence of a nested typedef? On 10/18/2004 03:31 PM, Terje Slettebø wrote: [snip] Quote: I don't think it can be done without a macro. Many good people have tried. In the Concept Traits library ([url]http://www.neoscientists.org/boost_library_proposals.html)[/url], where Tobias Schwinger now is co-author, I used macros for detection where you need a name, as there are headers (in the type_traits_ext directory of the zip-file there. Sorry, the docs are only in the archive at the moment) for detection of: - Non-static members (variables and functions) - Static members (variables and functions) - Nested classes or typedefs - Nested class templates One example of using boost/mpl/aux_/has_xxx.hpp is: <------------------ //Purpose: // Test use of SFINAE to detect T::nesting_type for any type, T. #include namespace boosts { BOOST_MPL_HAS_XXX_TRAIT_DEF(nesting_type) } struct bool_false{ char c[1];}; struct bool_true { char c[2];}; template<typename T> struct has_nesting_type { template<class C> static bool_true has_nest_type(C::nesting_type*); template<typename C> static bool_false has_nest_type(...); enum{ result = sizeof(has_nest_type<T>(0)) == sizeof(bool_true)}; }; struct not_nester{}; struct yes_nester { typedef not_nester nesting_type; }; #include <iostream> using namespace boost::mpl::aux; int main(void) { std::cout<<"yes_nester="< std::cout<<"not_nester="< std::cout<<"===>boostn"; std::cout<<"yes_nester="< std::cout<<"not_nester="< return 0; } Quote: ------------------ The output of this was: <------------------ /opt/intel_cc_80/bin/icc -c -MMD -g -I/home/evansl/prog_dev/boost-root.ln/boost_dev -I/home/evansl/prog_dev/boost-root.ln -o has_nesting_type.o has_nesting_type.cpp /opt/intel_cc_80/bin/icc -g -o has_nesting_type.exe has_nesting_type.o running has_nesting_type ../has_nesting_type.exe yes_nester=1 not_nester=0 ===>boost yes_nester=1 not_nester=0 Compilation finished at Sat Mar 6 16:45:52 Quote: ------------------ [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Terje Slettebø Guest Posted: Wed Oct 20, 2004 6:34 pm    Post subject: Re: Reusable test for presence of a nested typedef? "Glen Low" <glenlow (AT) pixelglow (DOT) com> wrote Quote: But being a guy who loves reuse, especially when he sees his own code littered with almost similar instances of the above, I wonder whether there's a general way of doing this (without bashing clients with a macro) i.e. has_typedef <U, result_type>::value OR has_typedef <typename U::result_type>::value ? I don't think it can be done without a macro. Many good people have tried. In the Concept Traits library ([url]http://www.neoscientists.org/boost_library_proposals.html)[/url], where Tobias Schwinger now is co-author, I used macros for detection where you need a name, as there are headers (in the type_traits_ext directory of the zip-file there. I thought so. In the end it wasn't too bad. I was using the presence or absence of a typedef within an enable_if on an partial specialization, it turns out all I needed was: template <typename T> struct exists { enum { value = true }; }; Simple, no? Why does it work... well in the partial specialization I do: template <typename T, typename Enable = void> class thing; template <typename T> class thing <T, typename enable_if ::type { ... }; So in classic SFINAE style, this will be silently dropped from the overload set when typename T::result_type doesn't exist. The exists + enable_if combination only serves to form the type "void" when result_type does exist. Clever... This does actually work. Come to think of it, I've played with this technique, as well - I even used the same name: "exists" (!) and found it marvellous simple. As you say, it doesn't work as a general trait (outside templates) - which was the original request - but it does work in this context, of eliminating overloads/specialisations during overload resolution/specialisation selection. It does actually generalise to detection of member class template, too: template <typename T, typename Enable = void> class thing; template <typename T> class thing< T, typename boost::enable_if< exists Quote: ::type { }; class A { public: template<class T> struct nested {}; }; class B {}; int main() { thing<A> a; // Ok thing<B> b; // Error } Moreover, unlike the pointer-to-member type technique, this one actually works for reference type members, too (you can't form a pointer to a reference). And - I just realised - it also works for private members (!), as the type or class template is never actually used (and hence checked for access), only checked for existence. Thanks for reminding me of this technique. :) Quote: Of course the exists template is of no use in a non-substitutionary context, and it might yet blow up in my face on a non-gcc compiler, but I'm fairly sure it will work on a Standard compliant one. It works on Intel C++ 7.1, too (EDG). Regards, Terje [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Glen Low Guest Posted: Wed Oct 20, 2004 9:42 pm    Post subject: Re: Reusable test for presence of a nested typedef? Quote: What happens when you put those templates as template members of a class which it itself a template on the type name in question? You imply that it didn't work? Don't understand, please elaborate with an example. Cheers, Glen Low, Pixelglow Software www.pixelglow.com [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Glen Low Guest Posted: Thu Oct 21, 2004 7:40 pm    Post subject: Re: Reusable test for presence of a nested typedef? Quote: One example of using boost/mpl/aux_/has_xxx.hpp is: ------------------ //Purpose: // Test use of SFINAE to detect T::nesting_type for any type, T. I wanted the converse, i.e. to detect T: for any nested type x. Cheers, Glen Low, Pixelglow Software www.pixelglow.com [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Glen Low Guest Posted: Thu Oct 21, 2004 7:41 pm    Post subject: Re: Reusable test for presence of a nested typedef? Quote: This does actually work. Come to think of it, I've played with this technique, as well - I even used the same name: "exists" (!) and found it marvellous simple. As you say, it doesn't work as a general trait (outside templates) - which was the original request - but it does work in this context, of eliminating overloads/specialisations during overload resolution/specialisation selection. Thanks for reminding me of this technique. No worries. Unfortunately it can be somewhat deceptive e.g. template <typename T> struct X <T, enable_if <T>::value>::type> { } never actually works for an integral type, since the exists <T::inner>::value causes the substitution failure anyway. Although for the case of &&, it does work correctly and as you'd expect. I find it a little more straightforward and reusable than SFINAE on overload functions; I've changed my macstl code to use it. Quote: It works on Intel C++ 7.1, too (EDG). Works on MS Visual C++ .NET 2003 too. Cheers, Glen Low, Pixelglow Software www.pixelglow.com [ See http://www.gotw.ca/resources/clcm.htm for info about ] [ comp.lang.c++.moderated. First time posters: Do this! ] Back to top Terje Slettebø Guest Posted: Fri Oct 22, 2004 5:16 pm    Post subject: Re: Reusable test for presence of a nested typedef? "Larry Evans" <cppljevans (AT) cox-internet (DOT) com> wrote Quote: On 10/18/2004 03:31 PM, Terje Slettebø wrote: [snip] I don't think it can be done without a macro. Many good people have tried. In the Concept Traits library ([url]http://www.neoscientists.org/boost_library_proposals.html)[/url], where Tobias Schwinger now is co-author, I used macros for detection where you need a name, as there are headers (in the type_traits_ext directory of the zip-file there. Sorry, the docs are only in the archive at the moment) for detection of: - Non-static members (variables and functions) - Static members (variables and functions) - Nested classes or typedefs - Nested class templates One example of using boost/mpl/aux_/has_xxx.hpp is: <snip> Yeah, but, uhm, here you're showing one version using an MPL macro to do it (BOOST_MPL_HAS_XXX_TRAIT_DEF(nesting_type)), and then one way to do it by making the trait class "manually": Quote: template<typename T struct has_nesting_type { template has_nest_type(C::nesting_type*); template<typename C> static bool_false has_nest_type(...); enum{ result = sizeof(has_nest_type<T>(0)) == sizeof(bool_true)}; }; So... what was the point? OP mentioned both ways, but inquired about the possibility of doing it without macros, and without having to define a template like the above. Regards, Terje [ 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.