- Boost - lists.boost.org

lightweight replacement for boost::function
by John Torjo 19 Aug '04

19 Aug '04

Dear boosters, Did a lightweight replacement for boost::function (compile-wise). Focus was on making it very fast to compile. Rationale: boost::function takes too long to compile. In a library I'm developing (win32gui), using boost::function would almost double the compilation time of some headers. This lightweight replacement takes virtually no extra compilation time. How I tested it: I took a project that took 1 minute to compile - added this to all of the headers, and it compiles in the same amount of time. #including <boost/function.hpp> almost doubled the time. Tested with vc71,gcc3.2. Feel free to use it any way you wish. Any feedback/suggestions deeply appreciated. Best, John -- John Torjo Freelancer -- john(a)torjo.com Contributing editor, C/C++ Users Journal -- "Win32 GUI Generics" -- generics & GUI do mix, after all -- http://www.torjo.com/win32gui/ -- v1.3beta released - check out splitter/simple_viewer, a File Explorer/Viewer all in about 200 lines of code! Professional Logging Solution for FREE -- http://www.torjo.com/code/logging.zip (logging - C++) -- http://www.torjo.com/logview/ (viewing/filtering - Win32) -- http://www.torjo.com/logbreak/ (debugging - Win32) #ifndef JOHN_TORJO_WIN32GUI_function #define JOHN_TORJO_WIN32GUI_function #if defined(_MSC_VER) && (_MSC_VER >= 1020) # pragma once #endif // function.hpp // Win32 GUI Generics library // // Copyright (C) 2004 John Torjo (john(a)torjo.com) // // Permission to copy, use, sell and distribute this software is granted // provided this copyright notice appears in all copies. // Permission to modify the code and to distribute modified code is granted // provided this copyright notice appears in all copies, and a notice // that the code was modified is included with the copyright notice. // // This software is provided "as is" without express or implied warranty, // and with no claim as to its suitability for any purpose. // // To see the version you have now, read win32gui/version.txt // // You can find the latest version of this library at http://www.torjo.com/win32gui/ #include <assert.h> #include <stdexcept> // #include <win32gui/.hpp> namespace win32 { namespace gui { /* Note: removed dependency on boost, and anyway, boost::function does take a lot to compile */ // note: removed dependency on boost, that's why we need this namespace detail { struct no_param {}; template<class func, class base> struct func_exec : base { typedef typename base::result res; typedef typename base::arg1 arg1; typedef typename base::arg2 arg2; typedef typename base::arg3 arg3; typedef typename base::arg4 arg4; typedef typename base::arg5 arg5; func_exec(func f) : f(f) {} // note: one of these is defined virtual in class base res operator()() { return f(); } res operator()(const arg1 & a1) { return f(a1); } res operator()(const arg1 & a1, const arg2 & a2) { return f(a1,a2); } res operator()(const arg1 & a1, const arg2 & a2, const arg3 & a3) { return f(a1,a2,a3); } res operator()(const arg1 & a1, const arg2 & a2, const arg3 & a3, const arg4 & a4) { return f(a1,a2,a3,a4); } res operator()(const arg1 & a1, const arg2 & a2, const arg3 & a3, const arg4 & a4, const arg5 & a5) { return f(a1,a2,a3,a4,a5); } // overrides the base' class base * clone() const { return new func_exec<func,base>(*this); } private: func f; }; template<class res> struct func_keeper0 { typedef func_keeper0<res> self; typedef res result; typedef no_param arg1; typedef no_param arg2; typedef no_param arg3; typedef no_param arg4; typedef no_param arg5; virtual res operator()() = 0; virtual self * clone() const = 0; }; template<class res, class a1> struct func_keeper1 { typedef func_keeper1<res,a1> self; typedef res result; typedef a1 arg1; typedef no_param arg2; typedef no_param arg3; typedef no_param arg4; typedef no_param arg5; virtual res operator()(const arg1&) = 0; virtual self * clone() const = 0; }; template<class res, class a1, class a2> struct func_keeper2 { typedef func_keeper2<res,a1,a2> self; typedef res result; typedef a1 arg1; typedef a2 arg2; typedef no_param arg3; typedef no_param arg4; typedef no_param arg5; virtual res operator()(const arg1&,const arg2&) = 0; virtual self * clone() const = 0; }; template<class res, class a1, class a2, class a3> struct func_keeper3 { typedef func_keeper3<res,a1,a2,a3> self; typedef res result; typedef a1 arg1; typedef a2 arg2; typedef a3 arg3; typedef no_param arg4; typedef no_param arg5; virtual res operator()(const arg1&,const arg2&, const arg3&) = 0; virtual self * clone() const = 0; }; template<class res, class a1, class a2, class a3, class a4> struct func_keeper4 { typedef func_keeper4<res,a1,a2,a3,a4> self; typedef res result; typedef a1 arg1; typedef a2 arg2; typedef a3 arg3; typedef a4 arg4; typedef no_param arg5; virtual res operator()(const arg1&,const arg2&, const arg3&, const arg4&) = 0; virtual self * clone() const = 0; }; } template<class res> struct func_wrapper0 { typedef res result; typedef detail::no_param arg1; typedef detail::no_param arg2; typedef detail::no_param arg3; typedef detail::no_param arg4; typedef detail::no_param arg5; typedef detail::func_keeper0<res> func_keeper; typedef func_wrapper0<res> self_type; func_wrapper0() : m_f(0) {} template<class func> func_wrapper0(func f) : m_f(new detail::func_exec<func,func_keeper>(f)) {} typedef res (*raw_func)(); func_wrapper0(raw_func f) : m_f(new detail::func_exec<raw_func,func_keeper>(f)){} func_wrapper0(const self_type & other) : m_f(other.m_f ? other.m_f->clone() : 0) {} res operator()() { if ( m_f) return (*m_f)(); else WIN32GUI_THROW std::runtime_error("null function"); } self_type& operator=(const self_type & other) { func_keeper * old = m_f; m_f = other.m_f ? other.m_f->clone() : 0; delete old; return *this; } private: func_keeper * m_f; }; template<class res,class arg1_type> struct func_wrapper1 { typedef res result; typedef arg1_type arg1; typedef detail::no_param arg2; typedef detail::no_param arg3; typedef detail::no_param arg4; typedef detail::no_param arg5; typedef detail::func_keeper1<res,arg1> func_keeper; typedef func_wrapper1<res,arg1> self_type; typedef res (*raw_func)(const arg1&); func_wrapper1() : m_f(0) {} template<class func> func_wrapper1(func f) : m_f(new detail::func_exec<func,func_keeper>(f)) {} func_wrapper1(raw_func f) : m_f(new detail::func_exec<raw_func,func_keeper>(f)){} func_wrapper1(const self_type & other) : m_f(other.m_f ? other.m_f->clone() : 0) {} res operator()(const arg1 &a1) { if ( m_f) return (*m_f)(a1); else WIN32GUI_THROW std::runtime_error("null function"); } self_type& operator=(const self_type & other) { func_keeper * old = m_f; m_f = other.m_f ? other.m_f->clone() : 0; delete old; return *this; } private: func_keeper * m_f; }; template<class res,class arg1_type,class arg2_type> struct func_wrapper2 { typedef res result; typedef arg1_type arg1; typedef arg2_type arg2; typedef detail::no_param arg3; typedef detail::no_param arg4; typedef detail::no_param arg5; typedef detail::func_keeper2<res,arg1,arg2> func_keeper; typedef func_wrapper2<res,arg1,arg2> self_type; typedef res (*raw_func)(const arg1&,const arg2&); func_wrapper2() : m_f(0) {} template<class func> func_wrapper2(func f) : m_f(new detail::func_exec<func,func_keeper>(f)) {} func_wrapper2(raw_func f) : m_f(new detail::func_exec<raw_func,func_keeper>(f)){} func_wrapper2(const self_type & other) : m_f(other.m_f ? other.m_f->clone() : 0) {} res operator()(const arg1 &a1, const arg2 &a2) { if ( m_f) return (*m_f)(a1,a2); else WIN32GUI_THROW std::runtime_error("null function"); } self_type& operator=(const self_type & other) { func_keeper * old = m_f; m_f = other.m_f ? other.m_f->clone() : 0; delete old; return *this; } private: func_keeper * m_f; }; template<class res,class arg1_type,class arg2_type,class arg3_type> struct func_wrapper3 { typedef res result; typedef arg1_type arg1; typedef arg2_type arg2; typedef arg3_type arg3; typedef detail::no_param arg4; typedef detail::no_param arg5; typedef detail::func_keeper3<res,arg1,arg2,arg3> func_keeper; typedef func_wrapper3<res,arg1,arg2,arg3> self_type; typedef res (*raw_func)(const arg1&,const arg2&,const arg3&); func_wrapper3() : m_f(0) {} template<class func> func_wrapper3(func f) : m_f(new detail::func_exec<func,func_keeper>(f)) {} func_wrapper3(raw_func f) : m_f(new detail::func_exec<raw_func,func_keeper>(f)){} func_wrapper3(const self_type & other) : m_f(other.m_f ? other.m_f->clone() : 0) {} res operator()(const arg1 &a1, const arg2 &a2, const arg3 &a3) { if ( m_f) return (*m_f)(a1,a2,a3); else WIN32GUI_THROW std::runtime_error("null function"); } self_type& operator=(const self_type & other) { func_keeper * old = m_f; m_f = other.m_f ? other.m_f->clone() : 0; delete old; return *this; } private: func_keeper * m_f; }; template<class res,class arg1_type,class arg2_type,class arg3_type,class arg4_type> struct func_wrapper4 { typedef res result; typedef arg1_type arg1; typedef arg2_type arg2; typedef arg3_type arg3; typedef arg4_type arg4; typedef detail::no_param arg5; typedef detail::func_keeper4<res,arg1,arg2,arg3,arg4> func_keeper; typedef func_wrapper4<res,arg1,arg2,arg3,arg4> self_type; typedef res (*raw_func)(const arg1&,const arg2&,const arg3&,const arg4&); func_wrapper4() : m_f(0) {} template<class func> func_wrapper4(func f) : m_f(new detail::func_exec<func,func_keeper>(f)) {} func_wrapper4(raw_func f) : m_f(new detail::func_exec<raw_func,func_keeper>(f)){} func_wrapper4(const self_type & other) : m_f(other.m_f ? other.m_f->clone() : 0) {} res operator()(const arg1 &a1, const arg2 &a2, const arg3 &a3, const arg4 &a4) { if ( m_f) return (*m_f)(a1,a2,a3,a4); else WIN32GUI_THROW std::runtime_error("null function"); } self_type& operator=(const self_type & other) { func_keeper * old = m_f; m_f = other.m_f ? other.m_f->clone() : 0; delete old; return *this; } private: func_keeper * m_f; }; /* void f() { std::cout << "f" << std::endl; } void f1(int i) { std::cout << "f1(" << i << ")" << std::endl; } bool f2(int i, char c) { std::cout << "f2(" << i << "," << c << ")" << std::endl; return true; } int f3(int i, char c, char d) { std::cout << "f3(" << i << "," << c << d << ")" << std::endl; return 2; } int f4(int i, char c, char d, char e) { std::cout << "f4(" << i << "," << c << d << e << ")" << std::endl; return 3; } struct titi { titi() : idx(0) {} void operator()() { std::cout << "titi" << idx++ << std::endl; } int idx; }; int main() { func_wrapper0<void> test(&f); test(); titi t; func_wrapper0<void> test2( t ); test2(); func_wrapper0<void> test3( test2 ); test2(); test2(); test3(); func_wrapper0<void> test4; test4 = test3; test4(); test4 = test; test4(); func_wrapper1<void,long> func1(f1); func1(0); func1(2); func_wrapper2<int,long,int> func2(f2); int i = func2(10,97); assert(i == 1); func_wrapper3<int,long,int,char> func3(f3); i = func3(11,97,'b'); assert(i == 2); func_wrapper4<int,long,int,char,char> func4(f4); i = func4(12,97,'b','c'); assert(i == 3); std::cin.get(); return 0; } */ }} #endif // JOHN_TORJO_WIN32GUI_function

4 6

Java style GUI in C++
by Johannes Brunen 19 Aug '04

19 Aug '04

Hi Reece, did you know of the existence of the following list of C++ GUI frameworks? I think they are currently the most important ones. http://www.fltk.org/index.php http://www.fox-toolkit.org/ http://gtkmm.sourceforge.net/ http://inti.sourceforge.net/ http://www.trolltech.com/ http://vcf.sourceforge.net/ http://www.wxwindows.org/ http://www.torjo.com/win32gui/ All of them have some strength and some weaknesses. It may be worth to look at least at the gtkmm, vcf and the win32gui frameworks. Johannes

7 6

boost::noncopyable broken?
by Bronek Kozicki 19 Aug '04

19 Aug '04

Hello when testing some regression errors I updated my second local copy from anonymous CVS @ sf.net (first copy is always from meta-comm and is producing regression tests you can see on www.meta-comm.com) Then I started running some tests, and received many errors similar to following: (vc71) boost\noncopyable.hpp(32) : error C2371: 'boost::noncopyable' : redefinition; different basic types boost\utility_fwd.hpp(26) : see declaration of 'boost::noncopyable' boost\test\unit_test_log.hpp(138) : error C2504: 'boost::noncopyable' : base class undefined (como-win32-4.3.3-vc7.1) boost/noncopyable.hpp", line 32: error #256: invalid redeclaration of type name "boost::noncopyable" (declared at line 26 of "boost/utility_fwd.hpp") typedef noncopyable_::noncopyable noncopyable; ^ I then took archived boost.tar.bz2 from meta-comm (boost-04-08-18-1200) and found the same error in this archive. Fortunatelly I also have older copy of this archive (boost-04-08-18-0600) and this archive is currently used to produce regression tests you will see soon - without error mentioned above. Apparently boost::noncopyable (or test unit?) got broken somewhere between Aug 18th 06:00 and Aug 18th 12:00 (I do not know what timezone) and this will produce many errors in next regression tests run. Unless it's quickly fixed. B.

2 2

link in python docs
by John Maddock 18 Aug '04

18 Aug '04

Dave, In libs/python/doc/tutorial/index.html there is: <link rel="next" href="quickstart.html"> But the file quickstart.html does not exist - and this is fooling bcp, I assume that it should point to doc/quickstart.html ? Whatever it should be can you fix? Thanks, John.

3 2

Boost Software License Progress As Of 20040817
by Stephan T. Lavavej 18 Aug '04

18 Aug '04

http://stl.nuwen.net/bsl3.html (now with serialization, though not python) * 5037 BSLed files. * 39 files converted which have to be manually verified. * 1 file that could be converted, but hasn't yet been. That's boost/variant/detail/move.hpp. It's dual-copyrighted and dual-licensed, but both Eric Friedman and Andrei Alexandrescu are now in blanket-permission.txt. The file should be converted. * 619 files that cannot be converted yet. * 83 files without a recognizable license (including files from format, lambda, pending, any, graph, iterator, mpl, property_map, and utility). * 248 files without a recognizable copyright holder (all mpl, except for one file from serialization, where bcp is getting confused - that's got a copyright and a variant of the BSL). Still need many authors for the BSL. [David Abrahams] > What do you mean? What was broken? python and serialization used to cause bcp to fail. serialization has since been fixed, but something is still wrong with python: **** exception(205): std::runtime_error: Bad file name: libs\python\doc\tutorial\quickstart.html ******** errors detected; see standard output for details ******** [John Maddock] > Done, I've added the Aliases to blanket-permission.txt, and > fixed up the regex scanning to pick up on Robert Ramey's web > address. Cool! > I use "." as the module name, of course then it flags up a > whole load of new problems :-( I tried what you suggest, but it didn't work: C:\Temp\boost\boost>..\bcp --report . bsl4.html **** exception(205): std::exception: boost::filesystem::is_directory: ":sys:stat.h": The parameter is incorrect. ******** errors detected; see standard output for details ******** If you know how to scan everything, perhaps you could post the results? The recent flurry of additions to blanket-permission.txt is /awesome/. Thanks, Stephan T. Lavavej [I don't speak for Microsoft]

4 5

Fw: Boost
by John Maddock 18 Aug '04

18 Aug '04

Another permission: ----- Original Message ----- From: "Maarten" <mkeijzer(a)xs4all.nl> To: <boost-owner(a)lists.boost.org> Sent: Wednesday, August 18, 2004 8:30 PM Subject: Boost > License Permission > Date: Wed, 18 Aug 2004 21:30:43 +0200 > User-Agent: KMail/1.6.2 > MIME-Version: 1.0 > Content-Disposition: inline > Content-Type: text/plain; > charset="us-ascii" > Content-Transfer-Encoding: 7bit > Message-Id: <200408182130.43251.mkeijzer(a)xs4all.nl> > Status: R > X-Status: NQ > X-KMail-EncryptionState: > X-KMail-SignatureState: > X-KMail-MDN-Sent: > > I Maarten Keijzer, give > permission for all my contributions to Boost to be moved to the > Boost Software License Version 1 > > I think this involves a single file in the detail section of the > interval library. FYI. > > -Maarten- >

1 0

Re: lightweight replacement for boost::function
by Max Khesin 18 Aug '04

18 Aug '04

John, just in case you missed this there is a similar effort happening here (from a competing GUI library!): http://tinytl.sourceforge.net/ This email and any files transmitted with it are confidential and intended solely for the use of the individual or entity to whom they are addressed. If you have received this email in error please notify the system manager. This message contains confidential information and is intended only for the individual named. If you are not the named addressee you should not disseminate, distribute or copy this e-mail.

2 1

python/embedding fails under Intel 8
by Stefan Slapeta 18 Aug '04

18 Aug '04

Maybe somebody who is more experienced with boost.python than me could have a look at this: http://tinyurl.com/6p8zf. Thanks, Stefan

3 7

RE: [boost] Re: ETI bug and iterators
by Yitzhak Sapir 18 Aug '04

18 Aug '04

> -----Original Message----- > From: David Abrahams > > "Yitzhak Sapir" writes: > > > I can't send my own code, but the following code causes the > > previously described error with an unpatched minimum_category.hpp > > and an ICE when compiled with patched and unpatched > > minimum_category.hpp: > > Not for me. Of course I had to remove the include of stdafx.h since > I don't have that, but the first error is: > > Dave Abrahams The problem is with debug STLPort iterators, not the pointer iterators that you get in non-debug iterators mode. Yes, release compiles fine for me with the additional BOOST_TT_BROKEN_COMPILER_SPEC. But I also want to be able to compile in debug, with the debug STLPort iterators. :-) Also, when I speak of ETI workarounds removed, I speak only of the #ifdefed lines in the following block: typedef minimum_category_impl< # if BOOST_WORKAROUND(BOOST_MSVC, == 1200) // ETI workaround is_same<T2,int>::value || # endif ::boost::is_convertible<T1,T2>::value , ::boost::is_convertible<T2,T1>::value # if BOOST_WORKAROUND(BOOST_MSVC, == 1200) // ETI workaround || is_same<T1,int>::value # endif > outer; I didn't notice there was another ETI workaround before.

2 1

algorithm/minmax_element regression with Comeau
by Bronek Kozicki 18 Aug '04

18 Aug '04

Following error http://tinyurl.com/5t7cj is due to too strict requirements of libcomo (at least I believe so). I received fix from Comeau Computing (stl_algo.h, too large to send as attachment with this message) and deployed it, tommorow we shall see this error gone. Again, I deployed it without stopping regression tests, thus next results in como-win32-4.3.3-vc7.1 may be screwed. B.

2 1