September 2010 - Boost-users

overload resolution with boost::function
by Jens Auer 20 Sep '10

20 Sep '10

Hi, I am trying to use objects of boost::function as parameters to overloaded function. However, the compiler fails to automatically resolve the function type when I use a function pointer: #include <boost/function.hpp> void f(int); void accept( boost::function< void () > g); void accept( boost::function< void (int) > g); int main(int argc, char* argv[]) { accept(f); return 0; } boost_function.cpp: In function ┬╗int main(int, char**)┬½: boost_function.cpp:11: Fehler: Aufruf des ├╝berladenen ┬╗accept(void (&)(int))┬½ ist nicht eindeutig boost_function.cpp:6: Anmerkung: Kandidaten sind: void accept(boost::function<vo id()>) boost_function.cpp:7: Anmerkung: void accept(boost::function<vo id(int)>) Sorry for the german error messages, the compiler complains about ambigous functions and shows two candidates. Is there any way to have the accept function call resolved automatically without creating explicit boost::function object?

2 2

[Concept] boost::FloatingPoint<T>
by Ruediger Berlich 20 Sep '10

20 Sep '10

Hi there, I'd like to make sure that some of my classes are only used with particular value types. This is easy enough to implement for integers, using BOOST_CONCEPT_ASSERT((boost::Integer<T>)) . However, there is no boost::FloatingPoint<T>, which is a bit odd. The following, easy code seems to do what I want, but reqires inclusion of boost/concept/detail/concept_def.hpp /*****************************************************************/ namespace boost { BOOST_concept(FloatingPoint, (T)) { BOOST_CONCEPT_USAGE(FloatingPoint) { x.error_type_must_be_a_floating_point_type(); } private: T x; }; template <> struct FloatingPoint<long double> {}; template <> struct FloatingPoint<double> {}; template <> struct FloatingPoint<float> {}; } /*****************************************************************/ >From what I understand, this means that I am relying on implementation details with the above code. Is there any way to do the above in a portable way ? Thanks and Best Regards, Ruediger P.S.: If the above seems acceptable, please feel free to include it in Boost.Concept.

2 2

Any "2D" iterator?
by lfrfly＠icqmail.com 19 Sep '10

19 Sep '10

I have on several occasions encountered situations in which it would be advantageous to view a container of containers as a flat sequence. It seems like there ought to be a natural way to do this using an iterator class. For instance, if you have vector<list<T>> v, you ought to be able to design an iterator which begins at v.begin()->begin() and runs until v.end(), over every element of every list. Does such a thing exist in Boost, or is it planned in the near future?

6 5

Re: [Boost-users] Serialization of functors and MPL use
by Anonymous user 19 Sep '10

19 Sep '10

Hello, Thanks for the quick response ! It is a good solution, but hhat I try to do is the following: template<class Object,class Type> struct my_class { �boost::function<Type ()> *m_f; template<typename Functor> my_class(Object &obj,Functor &f,Type & newValue) { �m_f= boost::bind(f,&obj); } �... �void serialize(Archive &ar, const unsigned int version) { ��ar & m_f; //does not compile �} �}; Instantiation:� �MyObject myObject("old value");� my_class�<�MyObject,std::string>�m(myObject,&MyObject::setName,e,"new value");� Thus the idea is to avoid changing the design (no inheritance to create functor) and just use boost::function to create dynamically functors. My problem is to rebuild dynamically the boost::function in the serialization step. I really needs to avoid using inheritance on my classes just to creates functors, thus I really need to keep the current design�and just build dynamically functors through boost::bind. One possible idea was to register a pair of types -> object which contains the function and the functor. Example: template<class Object,class Type> struct my_class { typedef boost::mpl::pair<Object,Type> key; typedef boost::function<Type ()> value; typedef�boost::mpl::map< key,value > > Functions; template<typename Functor> my_class(Object &obj,Functor &f,Type & newValue) { �m_f= boost::bind(f,&obj); typedef�boost::mpl::insert<�Functions,Functor > > t; } ... template <class Archive> void serialize( Archive & ar, constunsignedint version ) { ��typedef boost::mpl::at<key>()::type t; ��ar & t; //not possible -> does not instantiate the real functor address ! } }; The problem is the above solution is that we don't store the functor address but only it's signature. If you have an idea to make this example working or another idea to store functors which have been created dynamicaly, don't hesitate. � Thanks again for your help ! Regards. > ----- Original message ----- > From: "Robert Ramey‎" <ramey(a)rrsd.com> > To: boost-users(a)lists.boost.org > Subject: Re: [Boost-users] Serialization of functors and MPL use > Date: Sun, 19 Sep 2010 11:06:43 -0800 > > >Anonymous user wrote: > > Hello, > > > > Thanks for the response. > > > > Actually, I need to create dynamically functosr, because I call a lot > > of different methods coming from several classes. I don't want to > > implement the operator () inside each object used, because it is not > > appropriate in my code (too much classes). > > > > Can you please , give me an example of registration of functors ? > > I could if I had nothing else to do right now. I've got on my list to > make a "case study" in the documentation on how to do this. But > for now I'll just give a short explanation off the top of my head. > > A functor is just a structure. You can serialize just as you can > any other structure. > > The problem is that you want serialize a data variable which > refers to any one of a number of differently defined structures. > This is a specific instance of something that we want to do > in C++ all the time. The most common way of doing this > is to use a pointer to a virtual base class. e.g. > > struct functor_interface { > virtual int /* or other return type */ operator()(int /* other other > return types */ x, ...); > // virtual makes it polymorphic > // operator() ... makes a functor > }; > > So now we can define any number of functors which use the same interface: > > struct f1 : public functor_interface { > virtual int operator()(int x){ > ... whatever > template<class Archive> > void serialize(Archive &ar, const unsigned int version){ > //.. this functor has no parameters so nothing to do. > // .. serialization still needs this defined though > } > }; > > struct f2 : public functor_interface { > int m_a; // member variable which the functor uses at runtime > virtual int operator()(int x){ > ... whatever - something different than f1 > void serialize(Archive &ar, const unsigned int version){ > ar & m_a; > } > }; > > ... as many as you want. > > Register all the functors > BOOST_EXPORT_CLASS(f1) > ... > > The above makes the functors exportable - that is associates > the functor f1 with the external string name "f1". > > Now you can use a variable > > struct my_class { > functor_interface *m_f; > ... > void serialize(Archive &ar, const unsigned int version){ > ar & m_f; > } > }; > > Which uses the m_f variable as a "variable function". You > call it with > m_f->operator(112) > or maybe with something slicker like > (*m_f)(112) // or something similar > > Now you're done - just serialize m_class like anyother. > when you de-serialize it, it has the same state it did before > (whatever that was). > > Home this is helpful. > > Robert Ramey > > > > _______________________________________________ > Boost-users mailing list > Boost-users(a)lists.boost.org > http://lists.boost.org/mailman/listinfo.cgi/boost-users > -------------------------------------------------------------- Ovi Mail: Making email access easy http://mail.ovi.com

2 1

Re: [Boost-users] Serialization of functors and MPL use
by Anonymous user 19 Sep '10

19 Sep '10

Hello, Thanks for the response. Actually, I need to create dynamically functosr, because I call a lot of different methods coming from several classes.�I don't want to implement the operator () inside each object used, because it is not appropriate in my code (too much classes).� Can you please , give me an example of registration of functors ? Thanks in advance. Regards. > ----- Original message ----- > From: "OvermindDL1‎" <overminddl1(a)gmail.com> > To: boost-users(a)lists.boost.org > Subject: Re: [Boost-users] Serialization of functors and MPL use > Date: Fri, 17 Sep 2010 00:22:11 -0600 > > >On Thu, Sep 16, 2010 at 2:43 PM, Anonymous user <nokiac5eel(a)ovi.com> wrote: > > Hello everybody, > > > > > > > > I'm currently trying to serialize functors. > > > > Apparently, it is not possible to serialize boost::function objects, > > > > is it possible to find a way to serialize pointer to function ? (in order to > > make working the example below*) > > > > If it is definitively not possible, is there a way to 'register' a relation > > betwwen objects and functor ? (using MPL library). > > > > The idea is to rebuild the boost function according to the 'Object and > > 'Type' class of the template'. > > > > > > > > Example: > > > > > > > > mpl::pair<Object,Function> container; > > > > serialize() > > > > { > > > > boost::function<Type ()> m_getter=mpl::at< container,Object>; > > > > } > > > > > > > > > > > > *Here is the problem that� I'm trying to resolve: > > > > > > > > MyObject myObject("before"); > > > > > > > > SetCommand<std::string,MyObject> > > SetCommand(myObject,&MyObject::setName,&MyObject::getName,"after"); > > > > > > > > SetCommand.do(); � // call functor 'MyObject::setName("after")' > > > > SetCommand.undo(); // call functor 'MyObject::setName("before")' > > > > > > > > > > > > template<typename Type,typename Object> > > > > class SetCommand : public Command > > > > { > > > > > > > > > > > > � public: > > > > > > > > � � � � template <typename Setter, typename Getter> > > > > � � � � SetCommand(Object& o, const Setter& setter, const Getter& getter, > > const Type& value) : > > > > � � � � � � Command(), > > > > � � � � � � m_getter(boost::bind(getter, &o)), > > > > � � m_setter(boost::bind(setter, &o, _1)), > > > > � � � � � � initialValue(m_getter()), > > > > � � � � � � finalValue(value) > > > > � � � � { > > > > � � � � } > > > > > > > > virtual void undo() > > > > � � � � { > > > > � � � � � � m_setter(initialValue) ; > > > > � � � � } > > > > > > > > � � � � virtual void do() > > > > � � � � { > > > > � � � � � � m_setter(finalValue) ; > > > > � � � � } > > > > > > > > �� private: > > > > > > > > friend class boost::serialization::access; > > > > > > > > SetCommand(){} > > > > > > > > template <class Archive> > > > > void serialize( Archive & ar, const unsigned int version ) > > > > { > > > > > > > > � ar & m_getter; ERROR -> does not compile !!! > > > > � ar & m_setter; ERROR -> does not compile !!! > > > > > > > > � ar & initialValue; > > > > � ar & finalValue; > > > > } > > > > > > > > boost::function<Type ()> m_getter ; > > > > � � � � boost::function<void (Type)> m_setter ; > > > > > > > > � � � � Type initialValue ; > > > > � � � � Type finalValue ; > > > > }; > > > > > > > > Thanks in advance. > > You need to somehow register the function in there, if you know what a > limited set of types it might be, an integer that can rebuild it would > work, but the most common way is to just make a base class with an > operator() and have a (shared) pointer to that be your 'function', > then base classes (that can be properly serialized) to act as the > functions. Not pretty, not the easiest, but it works well and is > expandable. > _______________________________________________ > Boost-users mailing list > Boost-users(a)lists.boost.org > http://lists.boost.org/mailman/listinfo.cgi/boost-users > -------------------------------------------------------------- Ovi Mail: Making email access easy http://mail.ovi.com

2 1

Re: [Boost-users] Serialization of functors and MPL
by Anonymous user 19 Sep '10

19 Sep '10

Hello, Thanks for the response. Actually, I need to create dynamically functor, because I call a lot of different methods coming from several classes. see below: MyObject myObject("before"); SetCommand<std::string,MyObject> cmd(myObject,&MyObject::setName,&MyObject::getName,"after"); Thus, I'm wondering if it is not possible to use the MPL library to register the link between a class and it functor. Example: template<class Object,class Type> class SetCommand { typedef mpl::pair<MyObject,Function>::type �container; template<class Functor> SetCommand() { typedef mpl::insert<container,mpl::pair<Object,Functor>::type myType; } }; serialize() { boost::function<Type ()> getter=mpl::at< container,MyObject>::type; } MyObject myObject("before"); SetCommand<std::string,MyObject> cmd(myObject,&MyObject::setName,&MyObject::getName,"after"); The mechanism should be like the registration of serializable class with boost. The problem here is the registration of the functor, we don't register the address of the functor, but only its type(function signature). Thus it is difficult to allocate this functor ! If someone can give me a solution to this problem, it would be great :) Remark: I don't want to implement the operator () inside each object used, because it is not appropriate in my code (too much classes).� Thanks in advance. Regards. > ----- Original message ----- > From: "Robert Ramey‎" <ramey(a)rrsd.com> > To: boost-users(a)lists.boost.org > Subject: Re: [Boost-users] Serialization of functors and MPL > Date: Thu, 16 Sep 2010 14:20:06 -0800 > > >Anonymous user wrote: > > Hello everybody, > > > > I'm currently trying to serialize functors. > ... > > Thanks in advance. > > > Without going through your specific situation, I can say that this is an > easy problem. > Here are a couple of ways to do it. > > a) use boost::variant > > Make your functors: f, g, h > make a variant boost::variant<f, g, h> v; > > v = f or g or h ... > > #include boost/serialization/variant.hpp > serialize v > > b) make your functors f, g, h all derived from a base class - functors > make the base class polymorphic by making at least one member virtual > EXPORT f, g, and h > > functor *fx; > fx = &f, or &g, or &h > > serialize fx > > c) There might be a way to do it with boost::any also. > > Robert Ramey > > > > _______________________________________________ > Boost-users mailing list > Boost-users(a)lists.boost.org > http://lists.boost.org/mailman/listinfo.cgi/boost-users > -------------------------------------------------------------- Ovi Mail: Making email access easy http://mail.ovi.com

1 0

[serialization] When do I need to overload operators
by Sean Farrow 19 Sep '10

19 Sep '10

Hi: In trhe demo.cpp file in the Serialization examples, the << and >> operators are made friends of the gps_position class and are overridden. In what circumstances do I need to override these operators and when can I get away with just writing a Serialize template function. Regards Sean.

2 1

Performance of boost::serialization
by Daniel Herb 18 Sep '10

18 Sep '10

Hi! I'm having performance issues with boost::serilaization. First a short description what my data looks like: I serialize a container of boost::shared_ptr<Base> where i put polymorph objects in (no very deep hierarchy ~2 layers). The objects are ~40 byte each. The container has 100 elements in my test case. In the test application i can only serialize that container ~40 times per second. That's to slow for me because the data is used for the network communication. What i've found is this thread: http://old.nabble.com/-Serialization--Speedding-up-client-server-communicat… But it seems that i have to recreate the archive in my case because the data can get send, changed and send again. The documentation says that this behaviour can be disabled through disabling tracking. But that doesn't work for me because I use pointers. Robert Ramey also mentioned that implementing a lightweight version of stringstream could improve the performance. I liked to look at that but it seems that i'm missing something completely. class mystreambuf : public std::streambuf { public: virtual std::streamsize xsputn(const char*_Ptr, std::streamsize _Count) { } }; So this stream buffer does nothing right? But somehow serialization throughput is even worse with that implementation (compared to stringstream). I've never implemented a stream buffer so help would be nice. What are the requirements of boost::serialization for the stream buffer? Also generally it would be nice if you could share your oppinion if you think if the performance of boost::serialization can be tuned to serialize data for network communication. Is it only possible without polymorphism/shared_ptr? Any help is welcome! Thanks -- View this message in context: http://old.nabble.com/Performance-of-boost%3A%3Aserialization-tp29739639p29… Sent from the Boost - Users mailing list archive at Nabble.com.

3 4

libboost*-mt.so (thread safety)
by Stanisław Findeisen 17 Sep '10

17 Sep '10

Hi! It looks that in some older Boost versions (up to 1.40 or so) several shared objects were coming in 2 versions: with or without -mt suffix. For instance: libboost_filesystem-gcc42-1_34_1.so.1.34.1 libboost_filesystem-gcc42-mt-1_34_1.so.1.34.1 libboost_program_options-gcc42-1_34_1.so.1.34.1 libboost_program_options-gcc42-mt-1_34_1.so.1.34.1 libboost_regex-gcc42-1_34_1.so.1.34.1 libboost_regex-gcc42-mt-1_34_1.so.1.34.1 (These are the names from my Debian GNU/Linux installation, perhaps they are a bit different than when building from source.) This seems to no longer be the case with Boost 1.44. I've just built it from source passing threading=multi parameter to bjam, and can't see any -mt shared objects. This is what I have: drwxr-xr-x 2 ... 4096 ... 11:55 . drwxr-xr-x 4 ... 4096 ... 11:43 .. -rw-r--r-- 1 ... 118790 ... 11:49 libboost_date_time.a lrwxrwxrwx 1 ... 28 ... 11:48 libboost_date_time.so -> libboost_date_time.so.1.44.0 -rwxr-xr-x 1 ... 92296 ... 11:48 libboost_date_time.so.1.44.0 -rw-r--r-- 1 ... 294068 ... 11:50 libboost_filesystem.a lrwxrwxrwx 1 ... 29 ... 11:48 libboost_filesystem.so -> libboost_filesystem.so.1.44.0 -rwxr-xr-x 1 ... 171758 ... 11:48 libboost_filesystem.so.1.44.0 -rw-r--r-- 1 ... 534858 ... 11:51 libboost_graph.a lrwxrwxrwx 1 ... 24 ... 11:49 libboost_graph.so -> libboost_graph.so.1.44.0 -rwxr-xr-x 1 ... 379135 ... 11:49 libboost_graph.so.1.44.0 -rw-r--r-- 1 ... 286678 ... 11:52 libboost_math_c99.a -rw-r--r-- 1 ... 282978 ... 11:52 libboost_math_c99f.a lrwxrwxrwx 1 ... 28 ... 11:44 libboost_math_c99f.so -> libboost_math_c99f.so.1.44.0 -rwxr-xr-x 1 ... 133041 ... 11:44 libboost_math_c99f.so.1.44.0 -rw-r--r-- 1 ... 288144 ... 11:52 libboost_math_c99l.a lrwxrwxrwx 1 ... 28 ... 11:45 libboost_math_c99l.so -> libboost_math_c99l.so.1.44.0 -rwxr-xr-x 1 ... 132259 ... 11:44 libboost_math_c99l.so.1.44.0 lrwxrwxrwx 1 ... 27 ... 11:44 libboost_math_c99.so -> libboost_math_c99.so.1.44.0 -rwxr-xr-x 1 ... 137877 ... 11:44 libboost_math_c99.so.1.44.0 -rw-r--r-- 1 ... 1714884 ... 11:51 libboost_math_tr1.a -rw-r--r-- 1 ... 1723476 ... 11:52 libboost_math_tr1f.a lrwxrwxrwx 1 ... 28 ... 11:44 libboost_math_tr1f.so -> libboost_math_tr1f.so.1.44.0 -rwxr-xr-x 1 ... 295638 ... 11:44 libboost_math_tr1f.so.1.44.0 -rw-r--r-- 1 ... 1716390 ... 11:52 libboost_math_tr1l.a lrwxrwxrwx 1 ... 28 ... 11:44 libboost_math_tr1l.so -> libboost_math_tr1l.so.1.44.0 -rwxr-xr-x 1 ... 287712 ... 11:44 libboost_math_tr1l.so.1.44.0 lrwxrwxrwx 1 ... 27 ... 11:43 libboost_math_tr1.so -> libboost_math_tr1.so.1.44.0 -rwxr-xr-x 1 ... 287831 ... 11:43 libboost_math_tr1.so.1.44.0 -rw-r--r-- 1 ... 90942 ... 11:54 libboost_prg_exec_monitor.a lrwxrwxrwx 1 ... 35 ... 11:46 libboost_prg_exec_monitor.so -> libboost_prg_exec_monitor.so.1.44.0 -rwxr-xr-x 1 ... 77833 ... 11:46 libboost_prg_exec_monitor.so.1.44.0 -rw-r--r-- 1 ... 742886 ... 11:53 libboost_program_options.a lrwxrwxrwx 1 ... 34 ... 11:45 libboost_program_options.so -> libboost_program_options.so.1.44.0 -rwxr-xr-x 1 ... 440715 ... 11:45 libboost_program_options.so.1.44.0 -rw-r--r-- 1 ... 8868 ... 11:53 libboost_random.a lrwxrwxrwx 1 ... 25 ... 11:46 libboost_random.so -> libboost_random.so.1.44.0 -rwxr-xr-x 1 ... 13058 ... 11:45 libboost_random.so.1.44.0 -rw-r--r-- 1 ... 2102620 ... 11:50 libboost_regex.a lrwxrwxrwx 1 ... 24 ... 11:49 libboost_regex.so -> libboost_regex.so.1.44.0 -rwxr-xr-x 1 ... 1256561 ... 11:49 libboost_regex.so.1.44.0 -rw-r--r-- 1 ... 984416 ... 11:54 libboost_serialization.a lrwxrwxrwx 1 ... 32 ... 11:46 libboost_serialization.so -> libboost_serialization.so.1.44.0 -rwxr-xr-x 1 ... 624366 ... 11:46 libboost_serialization.so.1.44.0 -rw-r--r-- 1 ... 139202 ... 11:54 libboost_signals.a lrwxrwxrwx 1 ... 26 ... 11:46 libboost_signals.so -> libboost_signals.so.1.44.0 -rwxr-xr-x 1 ... 99503 ... 11:46 libboost_signals.so.1.44.0 -rw-r--r-- 1 ... 15498 ... 11:49 libboost_system.a lrwxrwxrwx 1 ... 25 ... 11:48 libboost_system.so -> libboost_system.so.1.44.0 -rwxr-xr-x 1 ... 20109 ... 11:48 libboost_system.so.1.44.0 -rw-r--r-- 1 ... 1476598 ... 11:49 libboost_test_exec_monitor.a -rw-r--r-- 1 ... 149246 ... 11:54 libboost_thread.a lrwxrwxrwx 1 ... 25 ... 11:47 libboost_thread.so -> libboost_thread.so.1.44.0 -rwxr-xr-x 1 ... 118667 ... 11:47 libboost_thread.so.1.44.0 -rw-r--r-- 1 ... 1634106 ... 11:54 libboost_unit_test_framework.a lrwxrwxrwx 1 ... 38 ... 11:47 libboost_unit_test_framework.so -> libboost_unit_test_framework.so.1.44.0 -rwxr-xr-x 1 ... 941323 ... 11:47 libboost_unit_test_framework.so.1.44.0 -rw-r--r-- 1 ... 2909510 ... 11:55 libboost_wave.a lrwxrwxrwx 1 ... 23 ... 11:48 libboost_wave.so -> libboost_wave.so.1.44.0 -rwxr-xr-x 1 ... 1686755 ... 11:48 libboost_wave.so.1.44.0 -rw-r--r-- 1 ... 694192 ... 11:54 libboost_wserialization.a lrwxrwxrwx 1 ... 33 ... 11:46 libboost_wserialization.so -> libboost_wserialization.so.1.44.0 -rwxr-xr-x 1 ... 464203 ... 11:46 libboost_wserialization.so.1.44.0 Is it so, that all Boost libraries are now thread safe?? Or maybe I overlooked some build options??... I had some 50 million build errors though (something related to Python) so some targets were skipped by bjam. -- http://eisenbits.homelinux.net/~stf/ OpenPGP: DFD9 0146 3794 9CF6 17EA D63F DBF5 8AA8 3B31 FE8A Like hardship, risk & challenge? --- Follow Jesus!!

2 1

boost::asio::io_service destructor never returns
by custgmx.net 17 Sep '10

17 Sep '10

Hello, he timer also starts a thread in which an io_service hosts deadline timer until thread is terminated. so a continuous timer. The boost thread which needs the timer is create from within DLL. The DLL ExitInstance function is called, when the hosting App. unloads the DLL. The ExitInstance function stops the thread and before timer. But the io_service destruction in my timer never returns, so the app hangs. This happens NOT, when i am able to call a Dispose Function before ExitInstance is called. However, some application loading my DLL, give not the chance to call this expose function. Anyone know, how to work around this problem ? Here is the code of my timer core. It is the thread which restarts the service until thread is stopped. The thread get's stopped by setting _stop flag and setting deadlime timer to 1 ms in future. summary: hangs when within Dll::ExitInstance destructed. Hangs not, if destrcuted before Dll::ExitInstance Thank you Code: void tcTimerThread::timerLoop(void) { _running=true; _io_service = new boost::asio::io_service; /// create timer _pTimer = new boost::asio::deadline_timer(*_io_service); while(_stop==false) { _pTimer->expires_from_now(boost::posix_time::milliseconds(_delay)); /// bind timer event function _pTimer->async_wait(boost::bind(&tcTimerThread::timerEvent,this,boost::asio: :placeholders::error)); try { if(_stop==false) { /// reset async _io_service->reset(); /// wait for timer event _io_service->run(); } } catch(...) { } } try { _io_service->stop(); delete _pTimer; /// !! never returns if called within ExitInstance of DLL;returns always is called before ExitInstance of DLL delete _io_service; } catch(...) { } _running=false; }

1 0