December 2013 - Boost-users

[serialization] How to deserialize only a part of an XML archive?
by luca 05 Dec '13

05 Dec '13

This is my scenario: I've got a shared library in which these classes are defined: class base { ... serialize() ... } class parent: public base { ... serialize(...) {...serialize base. and itselft too!} ... } the shared lib is used by a computational application that re-implements class parent in an extended version - this because the app is a Qt application that does visualization etc...: class grandParent: public parent { //it also defines some specific serialization capabilities: serialize (...) {serialize parent and also itself} } Now if I serialize into an XML archive an instance of grandParent I am clarly able to deserialize it in a grandParent object, but what about to deserialize it as a parent object? if I serialize from grandParent I would get something like: <grandParent> <parent> <base> </base> </parent> </grandParent> I would need to deserialize only the parent part: <parent> <base> </base> </parent> In practice I would like to be able to deserialize an XML archive to a base class instead that to the derived. But in this case the derived class is not accessible by boost (that thus complains about it) In this way I would get the better inter-operability between my application and the underlaying lib, that must be able to run without the app itself... Many thanks! Luca

1 0

Re: [Boost-users] [graph] Compilation problem using depth_first_visit with a listS as a VertexListS
by alex 05 Dec '13

05 Dec '13

> From that I isolated the following code which compiles without any problem: > > typedef adjacency_list < vecS, vecS, bidirectionalS,GraphvizVertexProperty,GraphvizEdgeProperty, GraphvizGraphProperty > Graph; > ... > depth_first_visit(_graph2, loop_head, default_dfs_visitor(), make_iterator_property_map(reachable_from_head.begin(),get(vertex_index, _graph2), c)); > So this works fine! > Then, I would like to change the type of graph, so more precisely, to use a std::list (or listS) instead of a std..vector (vecS) for the VertexListS > So the first line declaration becomes: > typedef adjacency_list < vecS, listS, bidirectionalS,GraphvizVertexProperty,GraphvizEdgeProperty, GraphvizGraphProperty > Graph; > //only the 2nd argument changes > But then when I compile the code, I get 1 error during the compilation as described hereafter. "If the VertexList of the graph is vecS, then the graph has a builtin vertex indices accessed via the property map for the vertex_index_t property." (http://www.boost.org/doc/libs/1_55_0/libs/graph/doc/adjacency_list.html) So when you use listS instead you have to supply your own vertex index map.

1 0

[log] ThreadID attribute is empty
by Igor R 05 Dec '13

05 Dec '13

Hello, I'm formatting a sink like this: sink->set_formatter ( expr::stream << expr::attr<boost::posix_time::ptime>("TimeStamp") << " T:" << expr::attr<boost::thread::id>("ThreadID") << ":" << expr::smessage ); logging::add_common_attributes(); For some reason ThreadID is empty. I tried to register the attribute explicitly: logging::core::get()->add_global_attribute("ThreadID", logging::attributes::current_thread_id()); but it doesn't help. The program is mutithreaded and linked with mt static libs. Thanks.

1 0

Having trouble compiling interval_map
by Daniel Rubi 04 Dec '13

04 Dec '13

Hi, I'm trying to implement an interval map where the keys are close_interval<long> an the values are set<string>. I modified the party example according to my needs, and came up with this short code: #include <boost/icl/interval.hpp> #include <boost/icl/interval_map.hpp> #include <boost/compressed_pair.hpp> using namespace std; using namespace boost::icl; typedef std::set<string> ids; int main(int argc, char* argv[]) { ids ids1; ids1.insert("T1"); ids ids2; ids2.insert("T2"); interval_map<closed_interval<long>, ids> mymap; closed_interval<long> i1 = closed_interval<long>(2, 7); closed_interval<long> i2 = closed_interval<long>(3, 8); mymap.insert(std::pair<closed_interval<long>, ids>(i1,ids1)); mymap.insert(std::pair<closed_interval<long>, ids>(i2,ids2)); return 0; } However, I get this compilation error: call of overloaded ‘insert(std::pair<boost::icl::closed_interval<long int>, std::set<std::basic_string<char> > >)’ is ambiguous mymap.insert(std::pair<closed_interval<long>, ids>(i2,ids2)); This is after getting compiation errors trying ti use: mymap += make_pair(i1,ids1); Any idea what's wrong?

1 0

[Any] inconsistent return types of any_cast overloads
by Arne Vogel 04 Dec '13

04 Dec '13

Hello Boost Users, Code like the following used to work with Boost 1.53 on x86-64 Linux using GCC 4.7 and -std=c++11: boost::any makeVec() { return std::vector<int>{ 1, 2, 3 }; } void foo() { const auto& vec = boost::any_cast<std::vector<int>>(makeVec()); for ( auto value : vec ) { // ... } } The same code is likely to cause a segmentation violation using Boost 1.55. According to the online documentation *1, any_cast has two overloads accepting reference arguments: template<typename ValueType> ValueType any_cast(any&); template<typename ValueType> ValueType any_cast(const any&); If these were indeed the only overloads accepting reference arguments, there is no reason why the example code shouldn't work. ValueType any_cast(const any&) should be the preferred overload and it returns a vector<int> temporary whose lifetime is then extended to the end of the block by being bound to a local const reference, as required by the C++ standard. However, in any.hpp, I discovered the following overload, which is only declared if the macro BOOST_NO_CXX11_RVALUE_REFERENCES is not set. I suppose this overload was added since 1.53, or that, previously, it was by default omitted. template<typename ValueType> ValueType&& any_cast(any&&); This has higher precedence binding to the temporary returned from makeVec, and returns an rvalue reference to the vector<int> contained in said temporary, which is then implicitly converted to a const lvalue reference as by the usual language rules. Since the lifetime of the temporary returned by makeVec is NOT extended - this only works for temporaries directly bound to a local const reference, whereas the result of makeVec is "hidden" inside the call to any_cast - the temporary is destroyed before the start of the for loop, which renders vec invalid and the code a denizen of undefined behavior land. So much for the analysis, now for my actual question. :-) Why does the new overload return by reference when the old ones return by value? The rationale for the old overloads was, I believe, "you should have to ask for a reference if you want one", which makes sense given precisely that using references tends to be more error prone if temporaries are involved. So this would be the very reason why Any does not declare: template<typename ValueType> const ValueType& any_cast(const any&); even though it could be implemented as easily as the actual overload. If I really wanted to get an const reference out of any_cast, I could write something like boost::any_cast<const ValueType&>(var); I can't think of a reason to be inconsistent in this regard and return a reference if a temporary is passed to any_cast, instead of requiring a user who really wants any_cast to return an rvalue reference to write any_cast<ValueType&&>(...). That the overload breaks existing code and (due to being undocumented and not checked at compile time *2), silently so, whereas using ValueType any_cast(any&&) would at least fail to do any harm, is, in my opinion, another reason to be sceptical of the current solution. Furthermore, there is a usability reason for preferring ValueType(any &&): If ValueType &&(any &&) is declared and I do NOT want a reference, I need to write more verbose and inelegant code than in the opposite case: // Name the temporary const auto &vecAny = makeVec(); // Now, ValueType&& any_cast(any&&) is eliminated from candidate set (no conversion from an lvalue to an rvalue reference) const auto& vec = boost::any_cast<std::vector<int>>(vecAny); Note that the const is accidental here, using a mutable lvalue would work as well. Looking forward to your opinions, Arne Vogel *1) http://www.boost.org/doc/libs/1_55_0/doc/html/boost/any_cast_idp25247624.ht… *2) A compiler may issue a warning. However, I found that GCC 4.8.1 does not warn, even with -Wall and the definition of any_cast being known at the point of use.

5 6

boost::regex Character set including braces
by Tobias 04 Dec '13

04 Dec '13

Dear boost-users, in the post http://stackoverflow.com/questions/20351369/regular-expression-cant-find-seq uence-of-numbers I discovered that if one uses an closing bracket as the first character of a character class the character class includes this bracket. This works with the standard setting of boost::regex (i.e., perl-regular expressions) but it is not documented in the manual page http://www.boost.org/doc/libs/1_55_0/libs/regex/doc/html/boost_regex/syntax/ perl_syntax.html#boost_regex.syntax.perl_syntax.character_sets Is this an undocumented feature, a bug or did I misinterpret something in the manual? The version information in version.hpp is: #define BOOST_LIB_VERSION "1_53" Info: I tried this specification of the bracket in the character set because I am used to it from emacs regular expressions. In the remainder there follows a program for reproduction of the phenomenon: #include <string> #include <boost/regex.hpp> #include <iostream> using namespace std; int main() { std::cout << "main()\n"; string s = "98[2]00"; string exp = "^[][0-9]*$"; const boost::regex e(exp); bool isSequence = boost::regex_match(s,e); //isSequence is boolean and should be equal to 1 cout << isSequence << endl; return 0; } /** Local Variables: compile-command: "g++ -g test.cc -o test.exe -lboost_regex-mt; ./test.exe" End: */ Best regards / Mit freundlichen Grüßen Tobias Zawada Engineering / Software Development ______________________________ ITI Events We say Thank you! for an outstanding ITI Symposium 2013! Make sure you save the date for next year: November 3-5, 2014 www.itisim.com ITI GmbH Schweriner Straße 1 01067 Dresden Germany T +49 (351) 26050 147 F +49 (351) 260 50 155 mailto:Zawada@itisim.com Managing Director/Geschäftsführer Jens O. Schindler · Dr. Andreas Uhlig Commercial Register/Handelsregister Amtsgericht Dresden · HRB 2331 USt.-IdNr.: DE140297717 Diese E-Mail ist streng vertraulich und nur für den Empfänger bestimmt. Sie kann Informationen enthalten, die rechtliche, geschäftliche oder andere Vergünstigungen bedeuten. Sollten Sie nicht der ausdrückliche Empfänger sein, dürfen Sie diese Informationen nicht benutzen, veröffentlichen oder vervielfältigen. The information in this e-mail is strictly confidential, is intended solely for the addressee, and may be privileged. If you are not the intended addressee you must not disclose, print or take action based upon this transmission and should delete it from your system.

2 1

[Filesystem] [Locale] Unicode filenames doc example error
by Egor Tensin 04 Dec '13

04 Dec '13

Why doesn't this simple code from the Boost.Locale docs work? #include <boost/filesystem.hpp> #include <boost/filesystem/fstream.hpp> #include <boost/locale.hpp> #include <exception> #include <iostream> int main() { try { std::locale::global(boost::locale::generator().generate("")); boost::filesystem::path::imbue(std::locale()); boost::filesystem::ofstream hello("привет.txt"); } catch (const std::exception& e) { std::cerr << e.what() << std::endl; } return 0; } Compiled with >cl /EHsc /D _UNICODE /D UNICODE b.cpp /I%BOOST_ROOT% /MD /link /LIBPATH:%BOOST_ROOT%\stage\lib it gives me >b.exe boost::filesystem::path codecvt to wstring: error Could someone please explain what's the problem there and how do I handle Unicode file names on Windows using Boost.Filesystem & Boost.Locale? I'm using Boost 1.55.0 & compiling with MS Visual C++ 2010 compiler v.16.00.30319.01.

1 0

[log] connecting a logger to a specific sink
by Igor R 04 Dec '13

04 Dec '13

Hello, I'm integrating Boost.Log to my project, and have a couple of questions. First of all, I'd like each instance of some class to write its log to a separate rotating file. I understand how to setup such a sink, but can't find a way to connect a particular source (logger) to the sink. Is it possible? Besides, when setting up the file, I configure it with: keywords::format = "[%TimeStamp%]: %Message%" but %TimeStamp% doesn't seem to get recognized, so I get empty [ ] in the file. Do I miss some header file? Thanks.

2 4

Print boost graph with custom vertex label
by Saha, Shambwaditya 03 Dec '13

03 Dec '13

Hi All, I want to print a Boost Graph having custom(string) vertex labels instead of the default vertex numbering label 0,1,2... I have initialized the graph as : typedef adjacency_list <vecS, vecS, directedS, property<vertex_name_t,string>> Graph; Graph g; set<string> names; map<string,Graph::vertex_descriptor> vertex ; for(auto it = names.begin() ; it != names.end(); ++it ) vertex[*it] = add_vertex(*it,g) ; Now how should I print this graph such that I get edges of the form abc -> xyz; instead of 1->2; Thanks and Regards, Shambwaditya

1 0

[phoenix][spirit] catch_all issues in V3
by Igor R 03 Dec '13

03 Dec '13

Hello, The following code doesn't compile (MSVC10, Boost 1.55). If I modify "catch_all" line like this: query = id[phx::try_[qi::_val = phx::bind(&string2test, qi::_1)].catch_all [ qi::_pass = false ], qi::_val = qi::_val ]; ...it compiles, but doesn't seem to invoke the semantic action in runtime. It there a workaround that would both compile and work? Thanks! #define BOOST_SPIRIT_USE_PHOENIX_V3 #include <string> #include <boost/spirit/include/qi.hpp> #include <boost/spirit/include/phoenix_core.hpp> #include <boost/spirit/include/phoenix_operator.hpp> #include <boost/phoenix/bind/bind_function.hpp> #include <boost/phoenix/statement/try_catch.hpp> namespace qi = boost::spirit::qi; namespace phx = boost::phoenix; struct test { }; test string2test(const std::string &str) { if (str == "throw") throw "exception!"; return test(); } template <typename Iterator = std::string::const_iterator> struct test_parser : qi::grammar<Iterator, test(), qi::blank_type> { test_parser() : test_parser::base_type(query) { query = id[phx::try_[qi::_val = phx::bind(&string2test, qi::_1)].catch_all [ qi::_pass = false ] ]; id = qi::lexeme[*qi::char_("a-zA-Z_0-9")]; } qi::rule<Iterator, test(), qi::blank_type> query; qi::rule<Iterator, std::string(), qi::blank_type> id; }; int main() { const std::string input = "qwerty"; test_parser<> p; test output; bool ok = phrase_parse(input.begin(), input.end(), p, qi::blank, output); } // 1> boost\boost\proto\traits.hpp(297): error C2338: 0 == Expr::proto_arity_c 1> boost\boost\mpl\eval_if.hpp(63) : see reference to class template instantiation 'boost::proto::result_of::value<Expr>' being compiled 1> with 1> [ 1> Expr=boost::phoenix::try_catch_actor<boost::proto::exprns_::basic_expr<boost::phoenix::tag::try_catch,boost::proto::argsns_::list2<boost::phoenix::try_catch_actor<boost::proto::exprns_::basic_expr<boost::phoenix::tag::try_catch,boost::proto::argsns_::list1<boost::phoenix::actor<boost::proto::exprns_::expr<boost::proto::tagns_::tag::assign,boost::proto::argsns_::list2<boost::proto::exprns_::expr<boost::proto::tagns_::tag::terminal,boost::proto::argsns_::term<boost::spirit::attribute<0>>,0>,boost::phoenix::actor<boost::proto::exprns_::basic_expr<boost::phoenix::detail::tag::function_eval,boost::proto::argsns_::list2<boost::proto::exprns_::basic_expr<boost::proto::tagns_::tag::terminal,boost::proto::argsns_::term<boost::phoenix::detail::function_ptr<1,test,test (__cdecl *)(const std::string &)>>,0>,boost::phoenix::actor<boost::spirit::argument<0>>>,2>>>,2>>>,1>>,boost::proto::exprns_::basic_expr<boost::phoenix::tag::catch_all,boost::proto::argsns_::list1<boost::phoenix::actor<boost::proto::exprns_::expr<boost::proto::tagns_::tag::assign,boost::proto::argsns_::list2<boost::proto::exprns_::basic_expr<boost::proto::tagns_::tag::terminal,boost::proto::argsns_::term<boost::phoenix::argument<3>>,0>,boost::phoenix::actor<boost::proto::exprns_::expr<boost::proto::tagns_::tag::terminal,boost::proto::argsns_::term<bool>,0>>>,2>>>,1>>,2>> 1> ]

2 3