- Boost-users - lists.boost.org

[MPI] Serializing C++ objects for transmission using Python MPI module
by Adam Stephens 29 May '08

29 May '08

I'd like to write a class in C++, then use a tool like SWIG (or Boost.Python, if absolutely necessary) to make a shared library/Python module. Then, I want to write python code that uses the boost.mpi python module to pass messages containing instances of my C++ class. In other words, I want to do what's described here: "Boost.MPI also supports efficient serialization and transmission of C++ objects (that have been exposed to Python) through its C++ interface. Any C++ type that provides (de-)serialization routines that meet the requirements of the Boost.Serialization library is eligible for this optimization, but the type must be registered in advance. To register a C++ type, invoke the C++ function register_serialized. If your C++ types come from other Python modules (they probably will!), those modules will need to link against the boost_mpi and boost_mpi_python libraries as described in the installation section. Note that you do not need to link against the Boost.MPI Python extension module." ( from http://www.boost.org/doc/libs/1_35_0/doc/html/mpi/python.html#mpi.python_us… ) I've looked at the documentation for the function register_serialized, but I am evidentally using it incorrectly, because I get compile time errors. My class looks like this: #include <boost/archive/text_oarchive.hpp> #include <boost/archive/text_iarchive.hpp> #include <boost/mpi/python/serialize.hpp> class Testclass { private: static bool registered; int data; friend class boost::serialization::access; template<class Archive> void serialize(Archive & ar, const unsigned int version) { ar & data; } public: Testclass() { if(!registered) { registered = true; boost::mpi::python::register_serialized(*this); } } void set_data(const int & in) { data = in; } int get_data() { return data; } }; bool Testclass::registered = false; I might be so far out in left field that what I'm trying to accomplish here needs explanation. According to the only documentation I can find, register_serialized is defined this way: template<typename T> void register_serialized(const T & value = T(), PyTypeObject * type = 0); The parameters have this meaning: type: "The Python type associated with the C++ type T. If not provided, it will be computed from the same value value." value: "A sample value of the type T. This may be used to compute the Python type associated with the C++ type T" So, I'm trying to register my class by passing in the object pointed to by this. I figure I only need to register it once; hence the static member. When I try to compile this to an object file, I get this error: /share/home/00791/adam/boost/include/boost-1_35/boost/mpi/python/serialize.hpp(323): error: no operator "<<" matches these operands operand types are: boost::mpi::packed_oarchive << Testclass ar << value; ^ detected during: instantiation of "void boost::python::detail::direct_serialization_table<IArchiver, OArchiver>::default_saver<T>::operator()(OArchiver &, const boost::python::api::object &, unsigned int) [with IArchiver=boost::mpi::packed_iarchive, OArchiver=boost::mpi::packed_oarchive, T=Testclass]" at line 158 of "/share/home/00791/adam/boost/include/boost-1_35/boost/function/function_template.hpp" Followed by several lines of similar messages, which I can attach if anyone needs to see them. There's one final wrinkle.. I've been trying to build boost for the past few days, and I kept getting a very simliar compile error at the same point. One of the authors of Boost.MPI took a look at it and submitted this change, which I made manually on my system: http://svn.boost.org/trac/boost/changeset/45825 Only after that change was I able to build Boost. Thanks

1 0

ASIO on AIX with gcc 4.2.3/4.3 annoying warnings
by Andrej van der Zee 29 May '08

29 May '08

Hi, The ASIO library (fantastic lib by the way) produces warnings as below on AIX with gcc 4.2.3 and 4.3. Some issue with braces. Is there an easy way to get rid of those warnings? Cheers, Andrej /opt/freeware/include/boost-1_35/boost/asio/ip/address_v6.hpp: In constructor 'boost::asio::ip::address_v6::address_v6()': /opt/freeware/include/boost-1_35/boost/asio/ip/address_v6.hpp:58: warning: missing braces around initializer for 'in6_addr::<anonymous union>' /opt/freeware/include/boost-1_35/boost/asio/ip/address_v6.hpp:58: warning: missing braces around initializer for 'u_int32_t [4]' /opt/freeware/include/boost-1_35/boost/asio/ip/address_v6.hpp: In static member function 'static boost::asio::ip::address_v6 boost::asio::ip::address_v6::loopback()': /opt/freeware/include/boost-1_35/boost/asio/ip/address_v6.hpp:340: warning: missing braces around initializer for 'in6_addr::<anonymous union>' /opt/freeware/include/boost-1_35/boost/asio/ip/address_v6.hpp:340: warning: missing braces around initializer for 'u_int32_t [4]' __________________________________________________________ Sent from Yahoo! Mail. A Smarter Email http://uk.docs.yahoo.com/nowyoucan.html

1 0

Similar problem to: boost vs libboost naming problem with bjam compile for vc90
by k.m.buckley＠lancaster.ac.uk 29 May '08

29 May '08

Hi there, found the thread entitled boost vs libboost naming problem with bjam compile for vc90 http://www.nabble.com/boost-vs-libboost-naming-problem-with-bjam-compile-fo… and think it is pointing me in the right direction however the circumstances here are slightly different to the circumstances in the thread and to the potential solution to that specific problem, namely the info at the Quantlib site. OK, here's the scoop: 1) Not usually a Windows user at all but the folk I am working for require a cross-platform solution to some already existing codebase 2) Part of the legacy code uses Boost and so we are sticking with it for now. 3) Of course it all works on a couple of UNIX-like platforms but then it would wouldn't it. 4) Am having to use MS VisStudio 2008 (aka MSVS 8) under XP 5) Don't want to use the MSVS GUI client for anything, so have GNU Fortran with various "proper" Makefiles flaoting around for some earlier stuff which has worked well and it seems that everything should be possible from the command line here. 5) Have an unzipped Boost 1.35 and a downloaded bjam.exe 6) Have been able to build and install the various bits of boost into C:Boost with the following process cd C:\local\boost_1.35_0 bjam --with-thread which builds the following in this subdir bin.v2/libs/thread/build/msvc-8.0/release/threading-multi boost_thread-vc80-mt-1_35.dll boost_thread-vc80-mt-1_35.dll.manifest boost_thread-vc80-mt-1_35.dll.rsp boost_thread-vc80-mt-1_35.exp boost_thread-vc80-mt-1_35.lib exceptions.obj exceptions.obj.rsp thread.obj thread.obj.rsp tss_dll.obj tss_dll.obj.rsp tss_pe.obj tss_pe.obj.rsp then remembered I wanted the static stuff so went: bjam link=static --with-thread which builds the following in this subdir bin.v2/libs/thread/build/msvc-8.0/release/link-static/threading-multi exceptions.obj exceptions.obj.rsp libboost_thread-vc80-mt-1_35.lib libboost_thread-vc80-mt-1_35.lib.rsp thread.obj thread.obj.rsp tss_dll.obj tss_dll.obj.rsp tss_pe.obj tss_pe.obj.rsp and, wanting to try out the regex example, I then tried bjam link=static --with-regex install which builds/makes its stuff (only static) in bin.v2/libs/regex/build/msvc-8.0/release/link-static/threading-multi c_regex_traits.obj c_regex_traits.obj.rsp cpp_regex_traits.obj cpp_regex_traits.obj.rsp cregex.obj cregex.obj.rsp fileiter.obj fileiter.obj.rsp icu.obj icu.obj.rsp instances.obj instances.obj.rsp libboost_regex-vc80-mt-1_35.lib libboost_regex-vc80-mt-1_35.lib.rsp posix_api.obj posix_api.obj.rsp regex.obj regex.obj.rsp regex_debug.obj regex_debug.obj.rsp regex_raw_buffer.obj regex_raw_buffer.obj.rsp regex_traits_defaults.obj regex_traits_defaults.obj.rsp static_mutex.obj static_mutex.obj.rsp usinstances.obj usinstances.obj.rsp w32_regex_traits.obj w32_regex_traits.obj.rsp wc_regex_traits.obj wc_regex_traits.obj.rsp wide_posix_api.obj wide_posix_api.obj.rsp winstances.obj winstances.obj.rsp though this latter command then installs, not only the regex static stuff, in C:\Boost\lib but also the thread static (I assume static anyway) stuff AND all the header-only libraries in C:\Boost\include\boost-1_35 libboost_regex-vc80-mt-1_35.lib libboost_regex-vc80-mt.lib libboost_thread-vc80-mt-1_35.lib libboost_thread-vc80-mt.lib which is not what I actualy wanted as i would have assumed that the --with-regex would limit the operations to just the regex stuff, but there you go. However, when I come to compile and link the regex example with the documentation directed command cl /EHsc /I C:\Boost\include\boost-1_35 regtest.cpp ^ /link /LIBPATH: C:\Boost\lib I am informed that the file libboost-regex-vc80-mt-s-1_35.lib cannot be opened I understand (because I RTFM) where the "-s" would/should come from but so far I can't work out how to produce such a file. Would I be better off just creating a list of the object files that the bjam link=static exercise creates and add them to a list to be linked to any examples I wish to compile statically - or is there a fix in the Boost develeopment pipeline that I could apply to my current setup ? Hoping I have not missed the obvious or previously solved, Kevin

3 3

Is there a limit for the number of threads created by thread_group.create_thread() ?
by he xiang 29 May '08

29 May '08

Hi, all When I use thread_group to create 400 threads , it throws a boost::thread_resource_error ? Is there a limit for the number of threads created by thread_group.create_thread() ? Thanks so much. Andy. Best Regards.

2 1

boost::is_const does not recognize constant type
by Becher, Jochen 29 May '08

29 May '08

Hi, The following type expression evaluates to the false_ type: boost::is_const<std::map<std::string,std::string>::const_iterator::refer ence>::type It should evaluate to the true type because ...::reference is const std::pair<const std::string,std::string> Does anybody know the reason why? I compile this using MSVC 2005 (SP1). Regards, Jochen Becher Jochen Becher Software Architect Stryker Navigation Stryker Leibinger GmbH & Co. KG Boetzinger Strasse 41 79111 Freiburg, Germany t: +49 761 4512 463 f:: +49 761 4512 449 463 jochen.becher(a)stryker.com Stryker Leibinger GmbH & Co. KG Sitz: Freiburg - Handelsregister: AG Freiburg HRA 4375 Komplementaerin: Stryker Beteiligungs GmbH Geschaeftsfuehrer: Dr. Heinrich W. Dreyer, Klaus Welte Sitz: Freiburg - Handelsregister: AG Freiburg HRB 5441 Bankverbindung: Bank of America N.A. Frankfurt 17 362 010 - BLZ 500 109 00, IBAN: DE04 5001 0900 0017 3620 10 Diese Nachricht und etwaige Dateianhaenge sind vertrauliche Informationen, die besonderen Geheimhaltungsvereinbarungen oder -vorschriften unterliegen koennen. Falls Sie nicht der beabsichtigte Empfaenger der Nachricht sind, werden Sie gebeten, den Absender telefonisch oder durch E-Mail zu unterrichten und die Nachricht und etwaige Dateianhaenge vollstaendig und endgueltig zu loeschen. Die Nachricht und ihr Inhalt duerfen in diesem Fall nicht kopiert, ausgedruckt oder weitergeleitet werden. This message and any attachment are confidential and may be privileged or otherwise protected from disclosure. If you are not the intended recipient, please telephone or e-mail the sender and delete this message and any attachment from your system. If you are not the intended recipient you must not copy this message or attachment or disclose the contents to any other person.

3 4

using program_options
by Ravi Chemudugunta 29 May '08

29 May '08

Hello, I have the following questions, I have breifly scoured the archive but haven't quite found what I want, so if this already been covered please point me in the direction. desc.add_options() ("port", po::value<string>(), "port to connect to"); I have looked at modifiers for value_semantic, one_token() multi_token() is there such a concept as required ? such that for e.g. desc.add_options() ("port", po::value<string>()->required(), "port to connect to"); causes the parser to 'die' if the value port isn't defined ? currently I am using vm.count("port") to deduce whether it has been defined. My other question is, why does the type have to be specified when extracting something from the variable map ? for e.g. vm["port"].as<string>() ... when it has already been defined in the configurations options. I ask this because I have a function that I am using: bool check_key(po::variables_map &vm, string key, bool required, bool print) { if (required && !(vm.count(key))) { cerr << "fatal: " << key << " not defined." << endl; exit(1); } if (print && vm.count(key)) { cout << key << ": " << vm[key].as<string>() << endl; } } This incidentally solves my first problem of having a key(option) defined or failing if it does not exist in the variable map, but ... say if I have a option defined as po::value<int>() inside option_description then my my function fails ... as it tries to interpret it as a string ? Is there a better way to do this ? cheers, ravi

1 0

multicast example build problem
by Gregory Gorbatov 29 May '08

29 May '08

Trying to build multicast boost example on SunOS 5.10 Generic_118833-36 sun4u sparc SUNW,Sun-Fire Got the foolwing errors. Any help appreciated cd /ilx/src/dbs/boost_1_35_0/libs/asio/example/multicast /ilx/src/dbs/boost_1_35_0/tools/jam/src/bin.solaris/bjam --toolset=sun sun.compile.c++ ../../../../bin.v2/libs/asio/example/multicast/sun/debug/threading-multi/receiver.o "../../../../boost/asio/detail/socket_ops.hpp", line 202: Error: Cannot assign void* to char*. "../../../../boost/asio/detail/socket_ops.hpp", line 213: Error: Cannot assign void* to char*. "../../../../boost/array.hpp", line 75: Error: Too few arguments for template std::reverse_iterator<boost::T*>. "../../../../boost/array.hpp", line 289: Where: While specializing "boost::array<boost::T, boost::N>". "../../../../boost/array.hpp", line 289: Where: Specialized in non-template code. "../../../../boost/array.hpp", line 76: Error: Too few arguments for template std::reverse_iterator<const boost::T*>.

3 3

Off topic: Interface / design question
by Khandelwal, Amit 28 May '08

28 May '08

Background: I am/was planning to profile the thread libraries from boost and ACE. During this experiment, I wish we had an interface class defined for threads. Also, I wish boost and ACE had implemented those interface classes. It would have helped someone like me profile different C++ libraries with ease. I was going through the header files of boost and I don't see a interface classes being used. I am trying to understand was there a conscientious decision made to not define them. I am trying to understand the reason behind those decisions. It will help newbie's like me to understand C++ better and write better code. Thanks, Amit - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - This message is intended only for the personal and confidential use of the designated recipient(s) named above. If you are not the intended recipient of this message you are hereby notified that any review, dissemination, distribution or copying of this message is strictly prohibited. This communication is for information purposes only and should not be regarded as an offer to sell or as a solicitation of an offer to buy any financial product, an official confirmation of any transaction, or as an official statement of Lehman Brothers. Email transmission cannot be guaranteed to be secure or error-free. Therefore, we do not represent that this information is complete or accurate and it should not be relied upon as such. All information is subject to change without notice. -------- IRS Circular 230 Disclosure: Please be advised that any discussion of U.S. tax matters contained within this communication (including any attachments) is not intended or written to be used and cannot be used for the purpose of (i) avoiding U.S. tax related penalties or (ii) promoting, marketing or recommending to another party any transaction or matter addressed herein.

1 0

How to tell the boost version in a C++ program?
by Peng Yu 28 May '08

28 May '08

Hi, I want to know the version number of boost in order to adapt my program different. Is there anything like that defined in boost? What would be the macro name for the version number? Thanks, Peng

2 1

multi_array vs vector vs map vs native arrays
by Hicham Mouline 28 May '08

28 May '08

Hello, I have an algorithm that interpolates a value out of tabulated data. Performance tests show that using double algo( const double* data [2], int datasize, double x) is faster than double algo( const std::vector<double>&, const std::vector<double>&, double x) and double algo( const boost::multi_array<double, 2>&, double x) by 30% or so. So I chose the first one. However, I have a requirement to add points to the tabulated data dynamically. For this, the best container would have been std::map<double, double> with which insertion is part of map's interface. However, the algorithm requires _random_ iterators to the container to proceed, so map doesn't workout but vector and multi_array do. What would be the best compromise? Regards,

1 0