April 2007 - Boost-users - lists.boost.org

Maps: one-to-one one-to-many many-to-many
by Travis Parks 24 Apr '07

24 Apr '07

Hello: I am constantly trying to jump into boost. However, I can only read so many pages of documentation before my brain explodes. I have always been interested in a data structure for storing pairs of of keys. In math it is a one-to-one mapping. I find it useful from time to time to have a container that allows me to treat it like an enum, but then to do the operation the other way. I usually do it lazily and just do a map and then do a linear search for the value when I need to. However, I have been thinking about it, and why should I have to suffer linear half the time? So today I wrote a very ugly class that does it log n both ways for me. I did it in less than 30 minutes so it has three methods, namely, insert, get_by_first and get_by_second. It is exceptionally ugly because you have to guaruntee that you will never change the data structure by returning a reference to one of the keys. Worse, if you want to save memory and make sure comparisons work the way you expect them, you have to use two maps of pointers with custom comparison functions that call the user's comparison function with the key-pointers dereferenced! I am wondering if there is something like this already out there already. Or is there a better way of dealing with my issue? I want structures that let me go on to things like one-to-many and many-to-many as well. Of course, what would "many" look like? Would it be an array of the mapped values? Perhaps I should stick with one-to-one! Finally, what would be a good name for this structure? I have been calling it bi-directional map. But this name doesn't carry well for one-to-many and many-to-many. And what should the interface look like? This structure is kind of hard to map to a STL container's interface due to the two keys (no operator[] for instance and no mapped_type or key_type). Thanks for any input! Ask if you want my horrible source code. Ever see two const_casts in one line? Want to?

4 3

SPACE problem in replace_all_regex
by llwaeva＠21cn.com 24 Apr '07

24 Apr '07

Hi there, I am using replace_all_regex to do a batch replace, for instance replace_all_regex( source, boost::regex("(xxx)|(yyy)"), string("(?1 FIRST)(?2 SECOND)"), format_all ); the format string is seperated with (?N XXX) pattern, where N is a decimal number. Using the format string shown above, all xxx will be replaced with " FIRST" (no quotation mark) and yyy will be replaced with " SECOND" (no quotation mark) instead of "FIRST" and "SECOND" (no extra space). Someone said the problem can be solved by rewriting the format string as string("(?1FIRST)(?2SECOND)") However, if the replacing string also begins with a decimal string, this approach doesn't work. For example, if I want to replace xxx with "55FIRST" (without quotation mark), the following function will TRANSLATE: replace the N0. 155 pattern with FIRST, but not replace N0.1 pattern with 55FIRST replace_all_regex( source, boost::regex("(xxx)"), string("(?155FIRST)"), format_all ); How can I solve such probelm? Thanks

1 0

Help installing (building) boost - Windows2K w/ MSVC++
by Beauregard Dommit 23 Apr '07

23 Apr '07

I'm trying to install boost on Windows 2000 running Visual C++, and having lots of problems. It's W2K w/ SP4, and SP6 for Visual Studio. First, I take it there is no precompiled "install" package, even for Windows? First problem - I unpack boost and put bjam in the boost directory. I run it from a command prompt w/ something like (I've tried a few variations): bjam "-sTOOLS=msvc" "--prefix=C:\program files\boost" "--PYTHON-ROOT=C:\Program Files\Python25" "--with-python-version=2.5" "--builddir=G:\temp\buildboost" install > install_log.txt (Sorry for whatever the mailer might be doing to the text - the above is one line.) I also tried the command line with lower case for the Python stuff - the above was my last desperate attempt. In my output file I see: skipping Boost.Python library build due to missing or incorrect configuration couldn't find Python.h in "c:/Python25/include" You can configure the location of your python installation by setting: PYTHON_VERSION - The 2-part python Major.Minor version number (e.g. "2.2", NOT "2.2.1") - currently "2.5" etc................ I guess I'm doing something stupid here, but what? Jam is apparently finding the prefix and builddir OK - it does build boost, or at least part of it, and installs a bunch of libraries and includes in the right place. I do have Python installed and working in C:\Program Files\Python25, and Python.h is there. Second problem - after the above, I get about 2000 lines of errors, or mostly errors. A lot of them are duplicates, but still that doesn't seem good. The first ones are about the Program Options package. That was the part of Boost I was going to try using first, so I'd like to get that fixed. Here are the error msgs: vc-C++ G:\temp\buildboost\bin\boost\libs\program_options\build\libboost_program_options.lib\msvc\debug\runtime-link-static\cmdline.obj cmdline.cpp G:\Temp\newboost\boost_1_33_1\boost/program_options/options_description.hpp(156) : error C2258: illegal pure syntax, must be '= 0' G:\Temp\newboost\boost_1_33_1\boost/program_options/options_description.hpp(156) : error C2252: 'm_default_line_length' : pure specifier can only bespecified for functions G:\Temp\newboost\boost_1_33_1\boost/program_options/options_description.hpp(159) : error C2065: 'm_default_line_length' : undeclared identifier Then a bunch of errors about the Bind package, etc, etc, etc. If someone can help me fix the above problems I'll be very grateful. Gard

4 3

Re: [Boost-users] [Serialization, XML] - how to get info "where am i?" in serialization helper (encoding is not recursive)?
by Frank Bergemann 23 Apr '07

23 Apr '07

Hi Robert, when i use this xml_oarchive wrapper: /*************************** * * xml_oarchive wrapper * ***************************/ template <class Archive> class xml_oarchive_orga_impl : public xml_oarchive_impl<Archive> { #ifdef BOOST_NO_MEMBER_TEMPLATE_FRIENDS public: #else friend class detail::interface_oarchive<Archive>; friend class xml_oarchive_impl<Archive>; friend class save_access; protected: #endif xml_oarchive_orga_impl(std::ostream & os, unsigned int flags) : xml_oarchive_impl<Archive>(os, flags) { }; ~xml_oarchive_orga_impl() { } ; template<class T> void save_override( #ifndef BOOST_NO_FUNCTION_TEMPLATE_ORDERING const #endif boost::serialization::nvp<T> & t, int i ){ MakeNVPPathConcat(t.name()); xml_oarchive_impl<Archive>::save_override(t, i); } }; // do not derive from this class. If you want to extend this functionality // via inhertance, derived from xml_iarchive_orga_impl instead. This will // preserve correct static polymorphism. class xml_oarchive_orga : public xml_oarchive_orga_impl<xml_oarchive_orga> { public: xml_oarchive_orga(std::ostream & os, unsigned int flags = 0) : xml_oarchive_orga_impl<xml_oarchive_orga>(os, flags) {} ~xml_oarchive_orga() { }; }; ... i got a compile error: /nfs/usta/FBergemann/SCMws/OPSC_DIAXHBS_dev_R1.8_provcorba/include/boost/archive/detail/interface_oarchive.hpp: In member function `Archive& boost::archive::detail::interface_oarchive<Archive>::operator<<(T&) [with T = const boost::archive::version_type, Archive = boost::archive::xml_oarchive_orga]': /nfs/usta/FBergemann/SCMws/OPSC_DIAXHBS_dev_R1.8_provcorba/include/boost/archive/detail/common_oarchive.hpp:34: instantiated from `void boost::archive::detail::co mmon_oarchive<Archive>::vsave(boost::archive::version_type) [with Archive = boost::archive::xml_oarchive_orga]' /nfs/usta/FBergemann/SCMws/OPSC_DIAXHBS_dev_R1.8_provcorba/include/boost/archive/detail/iserializer.hpp:557: instantiated from here /nfs/usta/FBergemann/SCMws/OPSC_DIAXHBS_dev_R1.8_provcorba/include/boost/archive/detail/interface_oarchive.hpp:78: error: no matching function for call to `boost::archive::xml_oarchive_orga:: save_override(const boost::archive::version_type&, int)' Do i have to add different overloads for save_override(...) in my class - even if i do NOT want to change the ones from the base class? \Frank > -----Ursprüngliche Nachricht----- > Von: "Frank Bergemann" <FBergemann(a)web.de> > Gesendet: 23.04.07 08:25:14 > An: boost-users(a)lists.boost.org > Betreff: Re: [Boost-users] [Serialization, XML] - how to get info "where am i?" in serialization helper (encoding is not recursive)? > > Hi Robert, > > > -----Ursprüngliche Nachricht----- > > Von: boost-users(a)lists.boost.org > > Gesendet: 23.04.07 01:37:58 > > An: boost-users(a)lists.boost.org > > Betreff: Re: [Boost-users] [Serialization, XML] - how to get info "where am i?" in serialization helper (encoding is not recursive)? > > > > > > Frank Bergemann wrote: > > > i use an external XML format for configuration of CORBA (MICO) data > > > and managed to align that with boost::serialization > > > (To not write my own encoder/decoder and to minimize efforts for my > > > module, in case the customer wants more CORBA objects/methods in > > > future. Actually i think it would be a good options to let generate > > > the boost::serialitzation helper file directly by MICO idl compiler.) > > > > What is the boost::serialitzation helper file? Oh maybe its what i called a > > wrapper > > i meant the serialize(...) functions. > The ones, which are related to the CORBA data types are plain simple now, directly related to the IDL file - e.g. > > template<class Archive> > void serialize(Archive & ar, > ::EmiHLRINData::HLRINData_V4::DCsiList_s &pInstance, > const unsigned int version) > { > ar & make_nvp_wrapper( "d_csi_opt", pInstance.d_csi_opt ); > ar & make_nvp_wrapper( "neg_ph_vlr", pInstance.neg_ph_vlr ); > ar & make_nvp_wrapper( "cap", pInstance.cap ); > ar & make_nvp_wrapper( "d_csi", pInstance.d_csi ); > ar & make_nvp_wrapper( "lowph", pInstance.lowph ); > ar & make_nvp_wrapper( "lowac", pInstance.lowac ); > } > > I saved it in a file.cc.template and let it be expanded for the Archives i use (e.g. file_xml_oarchive.cc). > So i can pre-compile the whole stuff in a library. And only re-compile, if the IDL files change. > ( Using templates it took ages to compile my module :-) ) > > > > > > OK, for this now i have to check, that the (external, non by > > > boost::serialization generated) XML image is _actually_ compliant with > > > C/C++ data structures - especially for dynamic sized data. That is > > > also covered now, the checks do fine. > > > > How much time/effort did it take to do this? > > > > Was not that much. > I have a XML "catalogue" of CORBA commands, for each cmd referring to CORBA object, method of object & related root data objects (holding event/subscriber variables inside, to be expanded). > Then I use boost::serialization save(xml_oarchive,...) to create a *dummy* reference object. > I traverse this one with some XML lib. And while traversing, i lookup corresponding positions in my XML formatted configuration file (for data to fill-into the boost::serialization image) and update the value in boost::serialization XML image. > Then i use load(xml_oarchive, ...) to create the C/C++ object, populated with my "real" data now. > I save all the objects needed for a command in a boost::any container (for invoking a generic CORBA-object/method dispatcher). > All this alignment stuff will be shifted to process startup phase. > In the end it's better than writing my own XML decoder for each CORBA data object. > It is much more generic using your boost::serialization lib :-) > > > > But what i didn't manage is to get the _location_ of error in the XML > > > file - in case there is an error. > > > > Hmmm - location as ? point to the line/character? current name? or ? > > location in XML data structure (XML-path). > > > > > > opsc_hlr :2204 084723:D:### invoked for <value> > > > opsc_hlr :2204 084723:D:### c'tor > > > opsc_hlr :2204 084723:D:### done for <value> > > > opsc_hlr :2204 084723:D:### d'tor > > > > > So there is no recursion(???). > > > > > Is there another way to get a "where am i" information within a > > > serialize (load/save) helper? > > > > make_...wrapper is not recurssive, save/load are. > > > > I would derive from xml_?archive and override the save(NVP(T)..) and/or > > load(NVP...) to capture/whatever the tags as they are read/writting. Don't > > forget to call the original save/load. > > Could you pls refer to short example for deriving from xml_?archive ? > > > Robert Ramey > > > > \Frank > _______________________________________________________________ SMS schreiben mit WEB.DE FreeMail - einfach, schnell und kostenguenstig. Jetzt gleich testen! http://f.web.de/?mc=021192

2 1

[multi_index] using boost::tuple and hash_unique
by Ovanes Markarian 23 Apr '07

23 Apr '07

Hello, unfortunately I can't figure out how to specify member access for a hash function to a hash indexed mult-index container containing boost::tuple. I would like to hash by the first tuple member only. Can someone help? Here is an example: typedef boost::tuples::tuple<std::string, int, some_other_type> my_tuple; struct hash {}; multi_index_container< my_tuple , indexed_by< hashed_unique<tag<hash>, ???? , other_index_type... > > hashed_container; I would like this tuple to be hashed by the first member. boost::tuple has no access member function to the first member. The only way to access it is: my_tuple t("aaa", 10, xyz); std::string s(get<0>(t)); but as I can see the doc for multi_index it supports only access through a member function or by using the hash function which is applied to the whole object. The second scenario does not suit well, since I must pass an object instance to the hasher, where I would like to pass a name only. Should I write my own key_extractor class? Any help is appreciated. With Kind Regards, Ovanes

2 4

[iostreams] replacement of std::ifstream, review request
by Sascha Ochsenknecht 23 Apr '07

23 Apr '07

Hi, I made a library which gets an std::ifstream and read a certain file format. Now I would like to support also gzip compressed input files, but it should also support not compressed files. I used boost::iostreams to implement a new stream class which should replace the std::ifstream in my library. Following functions are called from std::ifstream: - std::ifstream(const char*) - read(...) - seek(...) - close(...) My question is, is everything done correctly or is there a more elegant way to implement this? I'm not a stream expert, so I doubt if I do it in a good way :-). Would be great if someone from the experts can have a look to it and review it. Thanks in advance, Sascha I derived my class from boost::iostreams::filtering_istream, here is my implementation: #ifndef PAT_INPUTSTREAM_H #define PAT_INPUTSTREAM_H #include <boost/iostreams/device/file.hpp> #include <boost/iostreams/filter/gzip.hpp> #include <boost/iostreams/filtering_stream.hpp> #include <string> namespace my { namespace io = boost::iostreams; /*! Check if a given file is a (gzipped-)compressed format It is using functions from zlib */ bool isCompressed(const char* p_name); /*! \class ifstream Special input stream to handle compressed files. After opening an my::ifstream, it should be checked with is_open() if it is openend correctly. */ class ifstream : public io::filtering_istream { private: std::string _name; // the filename is needed for re-opening in seek() public: ifstream(const char* p_name) : _name(p_name) { if (isCompressed(p_name)) { io::filtering_istream::push(io::gzip_decompressor()); io::filtering_istream::push(io::file_source(p_name, std::ios::binary)); } else { io::filtering_istream::push(io::file_source(p_name)); } } bool open(const char* p_name) { if (!is_open()) { _name = p_name; if (isCompressed(p_name)) { io::filtering_istream::push(io::gzip_decompressor()); io::filtering_istream::push(io::file_source(p_name, std::ios::binary)); } else { io::filtering_istream::push(io::file_source(p_name)); } } return is_open(); } void close() { if (is_open()) { while (io::filtering_istream::size()) { io::filtering_istream::pop(); } } } bool is_open() { if (io::filtering_istream::size()) { // the last components is the file device io::file_source* device = io::filtering_istream::component<io::file_source>(io::filtering_istream::size()-1); if (device) return device->is_open(); } return false; } ifstream& seekg(std::streampos p_pos) { // seek means close the stream // and read until position is reached were we // want to be if (is_open()) { close(); if (open(_name.c_str())) { const std::streampos PAT_BLOCK_SIZE = 1024; char* buf = new char[PAT_BLOCK_SIZE]; while (p_pos) { if (p_pos / PAT_BLOCK_SIZE) { read(buf, PAT_BLOCK_SIZE); p_pos -= PAT_BLOCK_SIZE; } else { read(buf, p_pos % PAT_BLOCK_SIZE); p_pos -= (p_pos % PAT_BLOCK_SIZE); } } delete [] buf; } } return *this; } }; } // end of namespace pat #endif

1 0

[program_options] missing typed_value implicit()
by Romulus 23 Apr '07

23 Apr '07

Hi everybody, I'm trying the program_options library, I'd like to create my own options groups, but when i compile (using -lboost_program_options) option_groups.cpp from the example directory, i'm told that "implicit()" method from the typed_value class doesn't exist. Btw it doesn't appear in http://www.boost.org/doc/html/typed_value.html What is "implicit()" supposed to do, is it different from "default_value()" ? How can i solve this problem ? Here is the error message: option_groups.cpp: In function 'int main(int, char**)': option_groups.cpp:42: error: 'class boost::program_options::typed_value<std::basic_string<char, std::char_traits<char>, std::allocator<char> >, char>' has no member named 'implicit' Regards -- Romuald THION Doctorant level 3.0 "la vie, c'est comme un jeu mal foutu dont on ne connait pas les règles et où il n'y a pas de sauvegardes"

2 1

Including boost threads
by Christian Mol 23 Apr '07

23 Apr '07

Hello, Probably a newbie question, but here it goes: if I include the boost thread as follows #include <boost/thread/thread.hpp> and try to compile my library in MSVC++ 2005, I get a lot of errors in c:\boost\include\boost-1_33_1\boost\bind\mem_fn_template.hpp I have included a piece of the errorlog next: 3>c:\boost\include\boost-1_33_1\boost\bind\mem_fn_template.hpp(120) : error C2917: '_av *' : invalid template-parameter 3>c:\boost\include\boost-1_33_1\boost\bind\mem_fn_template.hpp(126) : error C2146: syntax error : missing ';' before identifier 'second_argument_type' 3> c:\boost\include\boost-1_33_1\boost\bind\mem_fn_template.hpp(171) : see reference to class template instantiation 'boost::_mfi::mf1<R,T,_av>' being compiled 3>c:\boost\include\boost-1_33_1\boost\bind\mem_fn_template.hpp(126) : error C4430: missing type specifier - int assumed. Note: C++ does not support default-int 3>c:\boost\include\boost-1_33_1\boost\bind\mem_fn_template.hpp(126) : error C4430: missing type specifier - int assumed. Note: C++ does not support default-int ...... ...... >c:\boost\include\boost-1_33_1\boost\bind\mem_fn_template.hpp(581) : error C2917: '_av *' : invalid template-parameter 3>c:\boost\include\boost-1_33_1\boost\bind\mem_fn_template.hpp(581) : error C2991: redefinition of template parameter '_av' 3> c:\boost\include\boost-1_33_1\boost\bind\mem_fn_template.hpp(581) : see declaration of '_av' 3>c:\boost\include\boost-1_33_1\boost\bind\mem_fn_template.hpp(581) : fatal error C1003: error count exceeds 100; stopping compilation Could someone give me any pointers to what could be the problem? When using it in a very simple program, it does work. Including it in my quite large library it does not work. Kind regards, Christian

2 2

Boost and Intellisense...
by Graham Reitz 23 Apr '07

23 Apr '07

This is probably a little out of scope but I am hoping that someone has gotten this to work. A code line like this: boost::filesystem::path working_path_a( boost::filesystem::current_path() ); ...causes vstudio 2005 intellisense to stop working. It says this 'No additional Information Available' at the bottom. Anything before that line and intellisense works fine and anything after it doesn't. Any help would be greatly appreciated. As a relatively new user to boost having this feature really helps out. Thanks, Graham

9 10

[boost][thread] boost::thread::condition documentation error
by David Callu 23 Apr '07

23 Apr '07

Hi, I found an error in the boost::thread::condition documentation. The member-function template <typename L, typename Pr> bool timed_wait(L& lock, const xtime& xt, Pr pred) is documented like this template<typename ScopedLock, typename Pred> bool timed_wait(ScopedLock&, Pred); The "const xtime&" parameter is discard. Regards David Callu

1 0