- Boost - lists.boost.org

Re: [boost] Answers re BENUM library for automated enum streaming
by Tony D 27 Sep '08

27 Sep '08

Hi Marcus, Thank you for your guidance, and my turn to apologise for slow turn-around and mistaking the name of EnumIO. Noted the heads-up re boost code formatting standards - will have to do something I guess :-(. As you say, my library relies on "exotic features", and while I was happy to discover Sun Studio has supported __VA_ARGS__ for a long time, VC++ support only came is 2005, and I'm not so happy to be unable to find explicit mention of it in the discussions for C++0x... so avoidance is definitely a strength for EnumIO. I did receive one suggestion elsewhere - think it's worth sharing as you might consider it for EnumIO. Idea is: "inheriting" from an existing enum and adding new values. An "extendable" enum would be packaged in a class that actually stores the value in an int and has a protected set_from_int(int) method. A derived class defines an enum with additional symbols/values, constructors from the base's enum and their own enum - both of which call the base set_from_int(). (A protected constructor from int() could be used - and might be more efficient - but in the case of BENUM would complicate end-user control over whether construction from int is implicit or explicit.) Thought that was a pretty neat idea. Lots to do on the docs, examples etc.. Regards, Tony ----- Original Message ---- From: Marcus Lindblom <macke(a)yar.nu> To: boost(a)lists.boost.org Sent: Friday, 19 September, 2008 17:53:24 Subject: Re: [boost] Answers re BENUM library for automated enum streaming (Sorry for not replying sooner) Tony D wrote: > Hi all, > > Should I expect any further feedback re BENUM? David, Marcus: we've discussed specific issues (portability of variadic macros, smart_enum features) - do you have further questions, comments re my replies, or recommended actions before submission for formal review? Should I be doing more to pursue suggestions or solicit feedback? Is anyone else currently reviewing the code in the vault, experiment with usage in a real system, or planning to do so? If so, can we keep in touch re progress? If I don't hear anything to the contrary this week, I'll start doing the preparations for requesting a formal review, specifically my TODO list currently looks like: Small detail: My little embryo is called EnumIO, not smart_enum. Smart_enum is probably something different altogether. I'm quite happy with the answers, and I think it suits as a good replacement for my lib. I took a brief look at the code, and besides it not being formatted according to boost standards, it looked pretty ok. I still think my little lib is useful, mainly because it doesn't rely on exotic features and it can also be used in a non-intrusive fashion. But I see no bigger conflict as your will probably win the future. (Besides, I'll have to figure out a few things before I can publish mine for review, as noted in the link I posted a few mails back. If I ever get there there will probably be some discussion about integration, etc.) Cheers, /Marcus _______________________________________________ Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost

1 0

[thread] TSS cleanup
by Felipe Magno de Almeida 26 Sep '08

26 Sep '08

Hello, I would like to know if the storage created for boost::thread_specific_ptr in the tss specific code for win32 is free'd after the destruction of said boost::thread_specific_ptr, or if it stays until the end of the process. For example: for(int i = very_big_number;i--;i > 0) boost::thread_specific_ptr<int> p(new int); To my knowledge, the pointer returned by new in this case is freed. But does the int* that is created to store the pointer to new int is also freed, or this program will get only bigger and bigger? I'm using boost 1.36. The boost in 1.34 did get bigger and bigger, and I modified to use a map, which made the complexity bigger (log N), but made it manageable to use spirit 1.8, which creates boost::thread_specific_ptr and destroys a lot of them. Thanks in advance, -- Felipe Magno de Almeida

4 7

[threadpool] sub_tasks and active waiting
by vicente.botet 26 Sep '08

26 Sep '08

Hi, How can I define pool so the following code do not deadlocks? ???? pool; int fib_seq(int n) { if (n <= 1) return n; else return fib_seq(n−1) + fib_seq(n−2); } void fib_par(int n) { if (n <= 5) // granularity ctl return fib_seq(n); else { tp::task<int> t1(pool.submit(bind(fib_par, n-1)); tp::task<int> t2(pool.submit(bind(fib_par, n-2)); return t1.get_future().get() + t2.get_future().get(); } } int main() { return fib_par(10); } Well, we can see that the number worker threads depends on the fib_par parameter. As soon as the fib_par function waits on its sub_tasks t1 and t2, the worker thread cannot handle other tasks, so we need a thread available for fib_par(9) and fib_par(8). When this worker thread call fib_par(9) after launching fib_par(8) and fib_par(7) will be blocked to manage other task, ... Evidently this do not scale very well, so we can not use the threadpool library to address this kind of problems, in which the tasks is split on sub-task launched to be run in parallel, and the the parennt task waits on its child sub-task to do somthing. IMO, what we need is that instead of creating a task we create a subtask. When the function will wait until the results of its subtasks are ready, we don't need to block the worker thread, instead we can implement replace the pasive wait by an active one, which will check if the waited task is ready and if this is not the case it will do a scheduling step on the associated thread pool and check again. void fib_par(int n) { if (n <= 5) // granularity ctl return fib_seq(n); else { tp::sub_task<int> t1(pool.submit_subtask(bind(fib_par, n-1)); tp::sub_task<int> t2(pool.submit_subtask(bind(fib_par, n-2)); return t1.get() + t2.get(); } } The submison of a sub_task can be done differently of a task, because there is a parent task that will be waiting for the end of the sub_task, so maybe we can add it on LIFO order. Do you think that this is an interesting feature for your library? could it be integrated in your library without disruption? Best, ______________________ Vicente Juan Botet Escribá

3 3

[Phoenix] Monomorphic functors don't expose result_type
by Mathias Gaunard 26 Sep '08

26 Sep '08

I thought bind was supposed to generate monomorphic functors when used with function pointers, pointers to members, or monomorphic functors. Yet the functors it generates don't expose result_type, but only the result template. Moreover, I would expect functors to stay monomorphic when generating lambdas. If foo and bar are monomorphic (and thus expose result_type), foo < bar should be monomorphic too.

2 2

[serialization] "stream error" when port from boost_1.33 to boost_1.36
by ipangth 26 Sep '08

26 Sep '08

I just port a project using boost_1.33_1 to use boost_1.36. The project has a Server class using a std::set<Server> serialization, the serialization save and load functions code see behind. I use gcc 4.1 and run on FC6. The porting problem is, when the old boost_1.33 generates(save) a xml file, then use boost_1.36 load the xml throws an exception "stream error"! the exception error code line : boost::archive::xml_iarchive ia(ifs); How can i sovle it? class Server { public: int id; std::string ip; int port; private: static std::set<Server> lst_; static int dirty_; }; #endif #ifdef USE_SERVER_SERIALIZATION #ifndef _SERVER_SERIALIZATION_ #define _SERVER_SERIALIZATION_ BOOST_CLASS_VERSION(Server, 1) namespace boost { namespace serialization { template <class archive> void serialize(archive & ar, Server & t, const unsigned int ver) { ar & BOOST_SERIALIZATION_NVP(t.id); ar & BOOST_SERIALIZATION_NVP(t.ip); ar & BOOST_SERIALIZATION_NVP(t.port); } } } #endif #endif int Server::save( const std::string & file) { try { std::ofstream ofs(file.c_str(), std::ios::binary); { boost::archive::xml_oarchive oa(ofs); oa << BOOST_SERIALIZATION_NVP(lst_); } } catch(...) { return ERROR_ERROR; } return ERROR_SUCCEED; } int Server::load(const std::string & file) { std::set<Server> Server::lst_; try { std::ifstream ifs(file.c_str(), std::ios::binary); { boost::archive::xml_iarchive ia(ifs); ia >> BOOST_SERIALIZATION_NVP(lst_); } } catch(std::exception & e) { std::cout << e.what() << std::endl; } return 0; } -- Best regards, PangYongQiang

2 5

[boos][switch] What do you think?
by Kasra Nassiri(Math & ComSci) 26 Sep '08

26 Sep '08

Hi all, I have been writing a BOOST_SWITCH(...) statement, which tries to emulate the C++ switch() for C++ objects. string variable = "some string"; string string0 = "0", string1 = "1", string2 = "2"; BOOST_SWITCH( variable , BOOST_CASE( string0, std::cout << "error" << std::endl; BOOST_BREAK; ); BOOST_CASE( string1, std::cout << "error" << std::endl; BOOST_BREAK; ); BOOST_DEFAULT( BOOST_SWITCH( string2, BOOST_DEFAULT( std::cout << "embedded switch statments work" << std::endl; ); ); std::cout << "end of secound switch statment" << std::endl; std::cout << "reached default" << std::endl; ); ); The above code would work correctly (msvc). However, for "const char *", the compiler could not deduce template. I have written an specialized "const char *" template however, still could not deduce template. Any suggestions? template<class T> switch_; template<> switch_<const char *>; With best regards Kasra Nassiri

1 0

Re: [boost] Phoenix review
by Alp Mestan 26 Sep '08

26 Sep '08

I really want to give my approval, for what it is worth... I vote to accept Phoenix into Boost. However, I think it is important to keep Boost.Bind as many of Boost users don't know (yet) Phoenix but intensively use Bind, that's to say they only use Bind because they just need it, no more. The documentation is excellent - I had to use it for some personal projects - and I got used to Phoenix very quickly. The syntax is quite simple compared to the work done in background by the library. Thank to Phoenix, we can now write FP code in C++ quickly, efficiently. Thanks guys.

2 1

[graph] MAXINT problem
by Janek Kozicki 25 Sep '08

25 Sep '08

Hi, I just noticed that in file http://svn.boost.org/svn/boost/trunk/boost/graph/sloan_ordering.hpp there is this line: #define MAXINT 2147483647 I'm afraid that it can conflict with MAXINT defined elsewhere. Or at least cause some confusion between MAXINT and INT_MAX from <limits.h> Wouldn't it be better to just #include <limits.h> and use INT_MAX defined there? -- Janek Kozicki |

3 3

boost::signal poor performance
by Hansi 25 Sep '08

25 Sep '08

Hello, I have posted this already to the user list, maybe wrong forum... I have made now a small test and I don't know if I make something wrong. I have the following performance comparsion between a boost::function and a boost::signal. Okay, I know that boost::signal has a lot of more functionality and for that it is for sure slower than boost::function. But 100times slower? May be I do something wrong? On my pc I have for boost::function a runtime of 728ms and for boost::signals a runtime of 64128ms There is one slot used for signals. Can anyone give me some hint what I do wrong? my test program was: class TestBind { public: int counter; void OnEventReceived(char b) { counter++; } }; class TestSignal { public: int counter; void OnEventReceived(char b) { counter++; } }; int _tmain(int argc, _TCHAR* argv[]) { TestBind testBind; PerformanceCounter timer; boost::function<void (char)> f = boost::bind(&TestBind::OnEventReceived,&testBind,_1); for(int i = 0; i < 100000000; i++) { f('a'); } TestSignal testSignal; boost::signal<void (char)> sig; sig.connect(boost::bind(&TestSignal::OnEventReceived,&testSignal,_1));; for(int i = 0; i < 100000000; i++) { sig('a'); } } Best regards Hansjörg

6 8

function_traits and functors
by Michael Marcin 24 Sep '08

24 Sep '08

Is it possible to use function_traits to decompose functors? Specifically is this supposed to work? void test_helper( int a ) { } template< class Handler > void test( Handler h ) { typedef typename boost::function_traits<Handler>::arg1_type ArgType; } int main() { test( boost::bind(&test_helper,_1) ); } Thanks, -- Michael Marcin

2 2