June 2009 - Boost-users - lists.boost.org

String, vector type with stack-allocated internal array (and dynamic array fallback)?
by Boris Dušek 01 Jul '09

01 Jul '09

Hello, for performance reasons, I would like not to have to have a std::wstring (1 allocation/1 deallocation of its character array) in one routine and instead have a wchar_t[MAXSIZE]. But I also would like my code to be correct for sizes bigger than MAXSIZE. Is there some type (in boost or elsewhere, preferable for both basic_string and vector) that defaults to stack-allocated array of some size (preferably configurable by a template parameter), and in case of overflow (i.e. push_back or insert when it's full), would allocate a dynamically-allocated array? This is how it works on MSVC with their STL basic_string implementation, just the array size is unconfigurable (16 bytes including null-terminating byte). I would obviously choose the MAXSIZE so that 99.9% of cases are handled by the stack-allocated array. Thanks, Boris

2 3

[asio] async_read and 100% CPU
by Etienne Philip Pretorius 01 Jul '09

01 Jul '09

Hello list, Could someone please shed some light on this subject. It seems that me read handler keeps on being called even when there has been no data transferred... Here is the accept handler that starts off the read handler (All methods are part of the same class, namely 'client'): /* Accept Handler */ void handle_accept( const boost::system::error_code& error ) { client* c; switch(error.value()) { case boost::system::posix_error::success: { c = new client(); socket.async_receive( boost::asio::buffer( buffer ),boost::bind( &client::handle_read, this, boost::asio::placeholders::error, boost::asio::placeholders::bytes_transferred ) ); break; } case boost::system::posix_error::interrupted: case boost::system::posix_error::connection_aborted: { c = this; break; } default: { throw error; } } acceptor.async_accept( c->socket, boost::bind( &client::handle_accept, c, boost::asio::placeholders::error ) ); }; And here is the read handler.... /* Read Handler */ void handle_read( const boost::system::error_code& error, std::size_t bytes_transferred ) { switch(error.value()) { case boost::system::posix_error::success: { if(bytes_transferred > 0) { std::copy( buffer.begin(), buffer.end(), std::back_insert_iterator<std::string>(data) ); break; } } case boost::system::posix_error::interrupted: { socket.async_receive( boost::asio::buffer( buffer ),boost::bind( &client::handle_read, this, boost::asio::placeholders::error, boost::asio::placeholders::bytes_transferred ) ); return; } default: { throw error; } } //TODO: parse contents.. }; Using epoll interface, as I am running Ubuntu Jaunty. Calling io_service.run() from the main thread. Thank you, Etienne Pretorius.

4 7

[serialization] Multiple archives in same stream
by Chard 01 Jul '09

01 Jul '09

Pardon me if this issue has been discussed earlier, but is there a particular technical issue why multiple archives cannot be serialized to (or, specifically, de-serialized from) the same stream? (I'm still on 1.38, so slap me if this works with 1.39+). Given the following example: #include <boost/archive/text_oarchive.hpp> #include <boost/archive/text_iarchive.hpp> #include <boost/serialization/serialization.hpp> void test_ser_app() { { { std::ofstream file("ser.txt"); archive::text_oarchive arch(file); std::string s("Hello World"); arch & s; size_t pos = file.tellp(); // [1] } { std::ofstream file("ser.txt", std::ios::app | std::ios::ate); archive::text_oarchive arch(file); std::string s("Goodbye World"); arch & s; } } // [2] { std::string s1, s2; std::ifstream file("ser.txt"); { archive::text_iarchive arch(file); arch & s1; } size_t pos = file.tellg(); // [3] try { archive::text_iarchive arch(file); arch & s2; } catch (archive::archive_exception &e) { std::cout << e.what() << std::endl; } } } The stream position reported at [1] is (correctly) 42, and examination of the text file (at [2]) shows the two archives serialized okay. However, the stream position at [3] appears to be at the end of the stream. Does the serialization read in a big chunk, regardless of whether it will be processing it all? If I hack in a set stream position at [3], then the library quite happily reads the second archive. Is it possible to have the library leave the stream pointer at the correct end position?

2 1

STLPort Install Before Building Boost 1.39.0
by biolaser 01 Jul '09

01 Jul '09

I'm on Windows XP SP3, msvc71 SP1, and I want to install STLport 5.2.1 before I install/build Boost 1.39.0. The STLport distribution install file (configure.bat) has two options I need help with: "--with-static-rtl" "-with-dynamic-rtl" Does Boost care whether I enable the static (libc.lib family) or whether I enable the dynamic (msvcrt.lib family) C/C++ library when linking with STLport? Then, there is the "-use-boost <boost install path>" option - I assume I should use this STLport option for Boost 1.39.0 - correct? Finally, is there anything I need to do in my building/installing of Boost itself to make sure it knows I have STLport installed? Thanks much! William Stroupe

2 1

[TypeTraits] Is it possible to detect that type is not exist and use default type?
by Sakharuk, Vlad (GMI Development) 30 Jun '09

30 Jun '09

Here is example: class one { public: typedef someclass internal; }; class two { // No internal class }; class default_internal{ }; I would like to be able to write something if it possible: template<class T, class B = if_exist(T::internal) T::internal else default_internal > class proc { }; So I can use proc with both proc<one> One; proc<two> Two; without specifying derived type? Thanks! \/. -------------------------------------------------------------------------- This message w/attachments (message) may be privileged, confidential or proprietary, and if you are not an intended recipient, please notify the sender, do not use or share it and delete it. Unless specifically indicated, this message is not an offer to sell or a solicitation of any investment products or other financial product or service, an official confirmation of any transaction, or an official statement of Merrill Lynch. Subject to applicable law, Merrill Lynch may monitor, review and retain e-communications (EC) traveling through its networks/systems. The laws of the country of each sender/recipient may impact the handling of EC, and EC may be archived, supervised and produced in countries other than the country in which you are located. This message cannot be guaranteed to be secure or error-free. References to "Merrill Lynch" are references to any company in the Merrill Lynch & Co., Inc. group of companies, which are wholly-owned by Bank of America Corporation. Securities and Insurance Products: * Are Not FDIC Insured * Are Not Bank Guaranteed * May Lose Value * Are Not a Bank Deposit * Are Not a Condition to Any Banking Service or Activity * Are Not Insured by Any Federal Government Agency. Attachments that are part of this E-communication may have additional important disclosures and disclaimers, which you should read. This message is subject to terms available at the following link: http://www.ml.com/e-communications_terms/. By messaging with Merrill Lynch you consent to the foregoing. --------------------------------------------------------------------------

3 4

Serialization of a universal data type
by Gokulakannan Somasundaram 30 Jun '09

30 Jun '09

Hi, I have started using Boost Serialization in one of my project and thanks for the library. I may be asking a question that might have got repeated, but i did my search on the forums but couldn't get the relevant answer. I have a data-type which is more like a universal datatype(unsigned long). Consider a 64 bit system and if the value to be stored in this datatype is greater than 8 bytes, then it is stored as value. If the value to be stored is greater than 8 bytes, then it is stored elsewhere and the pointer is stored here. By using other meta-data information, the data is extracted from this data-type. To be exact, i am referring to the datum implementation in open-source database postgresql. If i have to serialize this data-type, i need more info from other variables which store the metadata. The functions, i am trying to implement in Boost serialization are readDatum and _outDatum(similar to load/save). I don't see, how i can do that in Boost serialization, especially because of the signature of the load/save function. It only takes the version and the object to be serialized. How can i pass the other meta-data? I could always create a new data-type(say SerializableDatum) with all the required info, but then this data-type is being created only for the purpose of serialization. Is there any better way to achieve the same result? Thanks, Gokul.

2 2

Using boost::date_time library with wince
by Matt Schuckmann 30 Jun '09

30 Jun '09

Does anybody have any experience with using boost::date_time (boost version 1.37 or later) with windows ce (aka Windows Mobile 5 or Windows Mobile 6) Specifically I'm trying to figure out the best way to get around the missing functions std::localtime() and std::gmtime(). Yes as hard as it is to believe Microsoft choose to leave these 2 simple functions out of wince, imagine that. I'm actually running into this because I'm trying to use boost::thread which uses mostly the header only parts of date_time. The problem crops up when c_time.hpp is included. Thanks in advance for any help Matt Schuckmann

1 0

Using boost with windows mobile 5 or 6.
by Matt Schuckmann 30 Jun '09

30 Jun '09

I'm currently trying to use boost in my Windows Mobile 5/6 (aka WinCE) project and I'm having some real problems. Note the project is a port of a C++ library that is currently used in a Windows application and a Linux application. First let me say I'm using boost 1.37 and Visual Studio 2005 (VC 8), I'm not currently using STLPort. The specific libraries I'm trying to use are Thread, State machine and to some extent Date Time (mostly because Thread makes use of date_time) I'm not finding much of concrete info on how to go about doing thiss. The most help has come from the VeecoFTC regression build setup but that just got me to the point of building the boost library for the thread project. However, when I try to use thread in my project I run into problems specificly: * Wince does not implement std::localtime or std::gmtime which are used by the boost versions of localtime and gmtime in c_time.hpp. (Thread appears to depend on this via the date_time library) * Wince does not implement the standard library file locale. Most of the problems caused by this are solved be defining BOOST_NO_STD_LOCAL however the files compare.hpp and replace.hpp in boost/algorithm/string/ both don't respect the BOOST_NO_STD_LOCAL and they use std::locale a lot. Has anyone else gotten round these problems and if so what did you do? I've never been clear on why many people appear to be using STLPort with boost with winCE (or at least the VeecoFTC build is) is that something I should be doing? If so can I do it just for my WinCE builds? Does anybody know if I'd have more luck with boost 1.39 or even upgrading to Visual Studio 2008 (VC 9)? Thanks Matt Schuckmann

2 3

[boost][graph] problems using reverse_graph adaptor
by Andrey Torba 30 Jun '09

30 Jun '09

I have a Graph type which satisfies concepts: boost::function_requires<boost::GraphConcept<Graph> >(); boost::function_requires<boost::IncidenceGraphConcept<Graph> >(); boost::function_requires<boost::BidirectionalGraphConcept<Graph> >(); boost::function_requires<boost::VertexListGraphConcept<Graph> >(); boost::function_requires<boost::AdjacencyGraphConcept<Graph> >(); when i make a reverse adaptor: boost::make_reverse_graph(g); i receive such errors: 'edges_size_type' : is not a member of 'boost::graph_traits<Graph>' 'edge_property_type' : is not a member of 'Graph' 'vertex_property_type' : is not a member of 'Graph' Why does reverse_graph requires edges_size_type? edges_size_type isn't in BidirectionalGraph, VertexListGraph, PropertyGraph. it is in EdgeListGraph. reverse_graph requires Graph to be `BidirectionalGraph and optionally VertexListGraph and PropertyGraph`. What does `optionally` mean? My graph is not a PropertyGraph. I want just revert edges. -- Regards, Andrey

2 1

[Interprocess] Custom naming for the managed_external_buffer
by Thomas d'Erceville 30 Jun '09

30 Jun '09

Hello, I'm using a mapped file in order to share my objects between many processes and I already have my own kind of key to identify them. So instead of linking them to a <char *> in the function construct(), I would like to use my own type (providing convenient functions such as less(), operator==, ...) So I mean, instead of : MyObject * my_object = mapped_file.construct<MyObject>("object_name"); I want to known if there is a way to write something like : MyKey *object_key*(...) MyObject * my_object = mapped_file.construct<MyObject>(*object_key*); I'm taking a look at the section "Building custom indexes<http://www.boost.org/doc/libs/1_39_0/doc/html/interprocess/customizing_inte…>" but I am not sure to be able to solve my problem with that. Could someone confirm? Best regards, Thomas 2009/6/29 David Larsson <laserallan(a)gmail.com> > Hi! > I'm outputting the progress from an externally executed command line > application using the windows API call OutputDebugString. > This means that I need to convert all line breaking characters to flush > commands to make sure the text is outputted when a new line is received > rather than when some underlying structure finds it suitable to linebreak. > I've created a sink that writes to OutputDebugString and it works great, > however unless I manually insert lots of std::endl (or flushing some other > way) I won't get any output until the buffer is full. > > Is it possible to write an iostream filter that forces a flush on newlines > or is this something I need to handle before feeding data into my stream? > > /David > > > > _______________________________________________ > Boost-users mailing list > Boost-users(a)lists.boost.org > http://lists.boost.org/mailman/listinfo.cgi/boost-users >

2 1