October 2011 - Boost-users - lists.boost.org

Extend boost::adjacency_list
by Jupp Tscheak 03 Oct '11

03 Oct '11

Dear users of boost graph library, I would like to extend boost::adjacency_list like this code: class MyGraph: public boost::adjacency_list<boost::multisetS, boost::vecS, boost::bidirectionalS, boost::property<boost::some_vertex_property, SomeVertexProperty*>, boost::property<boost::some_edge_property, SomeEdgeProperty*> {} This works as expected. Is it possible to extend class "MyGraph" with additional vertex and/or edge properties defined? My first approach was like this code: template <class AdditionalVertexProperty = boost::no_property, class AdditionalEdgeProperty = boost::no_property> class MyGraph: public boost::adjacency_list<boost::multisetS, boost::vecS, boost::bidirectionalS, boost::property<boost::some_vertex_property, SomeVertexProperty*, AdditionalVertexProperty>, boost::property<boost::some_edge_property, SomeEdgeProperty*, AdditionalEdgeProperty> {} class ExtendedMyGraph: public MyGraph<boost::property<boost::some_add_vertex_property, SomeAddVertexProperty*> > {} This, however, led into multiple compiler errors. Does anyone know how to achieve the wanted behavior? Thanks in advance Jupp

2 1

[graph] compile breadth_first_search error.
by gongyiling 03 Oct '11

03 Oct '11

i 'm a newbie to boost graph library, I'm trying to compile a sample code as following: struct demo_visitor : public boost::default_bfs_visitor { template <typename Vertex, typename Graph> void discover_vertex(Vertex v, Graph& g) { //std::cout<<g[v].url<<std::endl; } }; int main(int argc, char* argv[]) { typedef boost::adjacency_list<boost::vecS, boost::listS, //change this to boost::vecS compile ok. boost::directedS> graph_t; typedef boost::graph_traits<graph_t>::vertex_descriptor vertex_descriptor_t; graph_t g; demo_visitor vis; boost::breadth_first_search(g, boost::vertex(0, g), boost::visitor(vis)); return 0; } And it fail to compile will the error in two_bit_color_map.hpp: can't transfer boost::detail::error_property_not_found to size_t. When I change the second type argument(VertexListS) of boost::adjacency_list from boost::listS to boost::vecS, it compile ok. Is there any implicite constraints on breadth_first_search? I also checked that boost::adjacency_list match the VertexListGraph concept both, using boost::vecS and boost::listS as it VertexListS type argument. Any response is appreacited!

3 3

[String Algo] funny warning
by Leo Cacciari 02 Oct '11

02 Oct '11

Hi, I usualy add flag -Wall to gnu g++ compilation while developing. Now, I got a funny warning while compiling the following function (I use the 1.42.0 release of boost) std::vector<std::string> split_payload(const std::string & payload_string) { std::vector<std::string> components; boost::algorithm::split(components,payload_string, boost::algorithm::is_any_of(",")); return components; } The warning I get is /usr/include/c++/4.4/bits/stl_algo.h: In function ‘std::vector<std::basic_string<char, std::char_traits<char>, std::allocator<char> >, std::allocator<std::basic_string<char, std::char_traits<char>, std::allocator<char> > > > split_payload(const std::string&)’: /usr/include/c++/4.4/bits/stl_algo.h:2178: warning: array subscript is above array bounds Someone could explain this? By the way, the function does what is supposed to do, so its not a big problem, but I'd really like to understand what the problem is... -- Leo Cacciari Aliae nationes servitutem pati possunt populi romani est propria libertas

1 0

Callback that calls a callback using boost::bind
by Christopher Pisz 02 Oct '11

02 Oct '11

I want to make a callback that also calls back. I cannot get the code to compile. I read on the boost documentation about apply()., but the example is unclear and I am not sure if that is what I am looking for. /// Smart Pointer typedef boost::shared_ptr<TcpServerSideConnection> SmartPtr; /// Callback function pointer type for TcpListener notification that our execution of /// TcpServerSideConnection::OnConnectionAccepted has completed typedef boost::function<void(TcpServerSideConnection::SmartPtr newConnection, const boost::system::error_code & error)> ConnectionAcceptedListenerCallback; /// Callback function pointer type for TcpServerSideConnection notification of /// a TcpListener::Listen IO completion typedef boost::function<void(ConnectionAcceptedListenerCallback listenerCallback, const boost::system::error_code & error)> ConnectionAcceptedCallback; TcpListener::SmartPtr me = boost::dynamic_pointer_cast<TcpListener, TcpBaseSocket>(shared_from_this()); TcpServerSideConnection::ConnectionAcceptedListenerCallback listenerCallback = boost::bind(&TcpListener::OnConnectionAccepted, me, _1, _2); TcpServerSideConnection::ConnectionAcceptedCallback acceptCompleteCallback = boost::bind(&TcpServerSideConnection::OnConnectionAccepted, newConnection, listenerCallback, boost::asio::placeholders::error); acceptor_.async_accept(newConnection->GetSocket(), acceptCompleteCallback); Gives me a compile error in bind.hpp with 100s of pages. What am I doing wrong here? Does anyone have an example of a callback that callsback using boost::bind?

2 1

shared_from_this - derived classes (yes, I googled)
by Christopher Pisz 02 Oct '11

02 Oct '11

Please examine the following: //----------------------------------------------------------------------------- void TcpListener::Listen() { TcpServerSideConnection::SmartPtr newConnection = TcpServerSideConnection::Create(ioService_); TcpBaseSocket::SmartPtr ptr = shared_from_this(); TcpListener::SmartPtr me = boost::dynamic_pointer_cast<TcpListener, TcpBaseSocket>(ptr); /// Asynchronously accept one connection request when it occurs acceptor_.async_accept(newConnection->GetSocket(), boost::bind(&TcpListener::OnConnectionAccepted, me, newConnection, boost::asio::placeholders::error)); } It throws an exception, unwinds my stack and destroys my object on the assignment of shared_from_this. I forced at least one shared pointer to be created, by making factory methods from base on up the class hierarchy. I read on google that you can only use shared_from_this on ONE class type in a class hierarchy. I did not understand the details of why. following a peer's advice, I enabled_shared_from_this on the base class and attempted to cast it in the derived class methods, getting the result that I did. I am trying to use a smart pointer here, following the example code for boost::asio::ip::tcp, assuming that I want to ensure the instance of 'ths' is alive between the time of the async call and the time it completes! I do not see a way to do this when using a class hierarchy. My base classes do similar async calls and try to handle it similarlly, so I cannot enable_shared_from_this only on the most derived. As there is no guarentee that only the most derived will be instantiated. I have more than 3 classes in the inheritance hierarchy, any of which could be instantiated. TcpBase - holds the socket and a smart pointer to the io_service TcpConnection - Adds Write and Read methods TcpClientConnection - Adds Connect method TcpServerConnection - Does not have connect, but can only be instantiated by a TcpListener - Accepts connections and creates TcpServerConnections Any of these, except the very base can be instantiated. How do I work around this problem?

2 5

identity function
by Krzysztof Żelechowski 02 Oct '11

02 Oct '11

I use the following function on a regular basis: template < class P_T > P_T const &same (P_T const &p) { return p; } I use it for two purposes: 1. to pass an identity function to an API that expects a unary function, 2. to help the compiler identify the right chain of type conversions. The standard uses static_cast for (2), which could be used but it disables some compile-time verification so I would rather not use static_cast when not absolutely necessary. I have been unable to find a corresponding feature in Boost (although the value factory comes close). Do you think adding such a function to Boost utilities would be worthwhile? Chris

5 4

[boost-newbie] g++ Version 3 (MinGW) and compiling boost
by Robert Hartmann 01 Oct '11

01 Oct '11

Hi there, does anyone could say me how can I compile Boost.Filesystem with MinGW: Is something wrong with my b2-call see below, or is my project-config.jam buggy? ========================= import option ; using gcc ; using python : 2.7 # version : E:/Programme/Python27 # cmd-or-prefix : E:/Programme/Python27/include # includes : E:/Programme/Python27/libs # libraries ; using mpi : : <find-static-library>mpi <library-path>"E:/Programme/MPICH2/lib" <include>"E:/Programme/MPICH2/include" : "\"E:\\Programme\\MPICH2\\bin\\mpiexec\"" ; option.set keep-going : false ; ========================== g++ (GCC) 3.4.2 (mingw-special) Copyright (C) 2004 Free Software Foundation, Inc. This is free software; see the source for copying conditions. There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. gcc (GCC) 3.4.2 (mingw-special) Copyright (C) 2004 Free Software Foundation, Inc. This is free software; see the source for copying conditions. There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. I got as error message for Boost.Filesystem: ./boost/filesystem/v3/config.hpp:49:5: #error Configuration not supported: Boost.Filesystem V3 and later requires std::wstring support That means Boost.Filesystem V2 should work? Can I compile it without other boost things? I run set BUILDPATH=G:/Downloads/Libraries/boost_1_47_0/ownbuild set INSTALLPATH=G:/Downloads/Libraries/boost_1_47_0/distrib b2 --prefix=%INSTALLPATH%/libs/boost-1.47.0/distrib ^ --toolset=gcc ^ --build-dir=%BUILDPATH% ^ --layout=versioned ^ --variant=debug,release ^ --runtime-link=static ^ --link=shared,static ^ --threading=multi ^ --without-mpi ^ --without-python stage >build.log 2>&1 And in build.log I found many errors after waiting some Minutes: ...failed updating 26 targets... ...skipped 30 targets... ...updated 842 targets... If I doesn't use the options --without-mpi --without-python than about 30 seconds later I got this output in build.log: ====================== G:/Downloads/Libraries/boost_1_47_0/tools/build/v2/build\virtual-target.jam:1079: in virtual-target.register-actual-name from module virtual-target error: Duplicate name of actual target: <pstage\lib>mpi.pyd error: previous virtual target { common%common.copy-mpi.pyd.PYTHON_EXTENSION { gcc%gcc.link.dll-mpi.pyd.PYTHON_EXTENSION { gcc%gcc.compile.c++-python\collectives.o.OBJ { python/collectives.cpp.CPP } } { [...] error: created from ./stage-proper error: added properties: <debug-symbols>off <define>NDEBUG <inlining>full <library>object(file-target)@4418 <library>object(file-target)@4420 <library>object(file-target)@4488 <library>object(file-target)@4490 <library>object(file-target)@4602 <library>object(file-target)@4604 <library>object(file-target)@4617 <library>object(file-target)@4619 <library>object(searched-lib-target)@4507 <optimization>speed <runtime-debugging>off <variant>release <xdll-path>/G:/Downloads/Libraries/boost_1_47_0/ownbuild/boost/bin.v2/libs/mpi/build/gcc-mingw-3.4.2/release/threading-multi <xdll-path>/G:/Downloads/Libraries/boost_1_47_0/ownbuild/boost/bin.v2/libs/python/build/gcc-mingw-3.4.2/release/threading-multi <xdll-path>/G:/Downloads/Libraries/boost_1_47_0/ownbuild/boost/bin.v2/libs/serialization/build/gcc-mingw-3.4.2/release/threading-multi error: removed properties: <debug-symbols>on <inlining>off <library>object(file-target)@2472 <library>object(file-target)@2474 <library>object(file-target)@2542 <library>object(file-target)@2544 <library>object(file-target)@2656 <library>object(file-target)@2658 <library>object(file-target)@2671 <library>object(file-target)@2673 <library>object(searched-lib-target)@2561 <optimization>off <runtime-debugging>on <variant>debug <xdll-path>/G:/Downloads/Libraries/boost_1_47_0/ownbuild/boost/bin.v2/libs/mpi/build/gcc-mingw-3.4.2/debug/threading-multi <xdll-path>/G:/Downloads/Libraries/boost_1_47_0/ownbuild/boost/bin.v2/libs/python/build/gcc-mingw-3.4.2/debug/threading-multi <xdll-path>/G:/Downloads/Libraries/boost_1_47_0/ownbuild/boost/bin.v2/libs/serialization/build/gcc-mingw-3.4.2/debug/threading-multi G:/Downloads/Libraries/boost_1_47_0/tools/build/v2/build\virtual-target.jam:490: in actualize-no-scanner from module object(file-target)@5073 G:/Downloads/Libraries/boost_1_47_0/tools/build/v2/build\virtual-target.jam:135: in object(file-target)(a)5073.actualize from module object(file-target)@5073 G:/Downloads/Libraries/boost_1_47_0/tools/build/v2\build-system.jam:749: in load from module build-system G:\Downloads\Libraries\boost_1_47_0\tools\build\v2/kernel\modules.jam:283: in import from module modules G:\Downloads\Libraries\boost_1_47_0\tools\build\v2\kernel\bootstrap.jam:142: in boost-build from module G:\Downloads\Libraries\boost_1_47_0\boost-build.jam:17: in module scope from module ============================== Please, can any one help me? Best regards, Robert

2 2

question regarding Depth First Search algorithm
by Pablo Fleurquin 01 Oct '11

01 Oct '11

Hi, My name is Pablo Fleurquin.I am a student of M.Sc in Physics at the UIB (Palma de Mallorca, Spain) and I'm also working in the IFISC institute (Institute for Interdisciplinary Science and Complex Systems). I have a question regarding the Depth First Search algorithm. Ihave beenlooking at the Boost library page but I'm a newbie with this stuff. *Having a directed graph G (that could be a connected graph or not), I want my program to find the cycles within the graph and remove all the edges that are part of them. It doesn't matter which cycle is erased first as long as the result is the graph G without edge that formely were part of a cycle (in other words, the remaining edges are those that doesn´t form cycles) [attached is example1]* Intuitively, I am sure that using the depth_first_search() function and recalling it recursively for each "new" graph G "with less" cycles will finally arrive to graph G with no cycles. The problem is that I have the idea, but my knowledge of c++ is not so vast to implement this code. I would be gratefully, if you could show me a code for this implementation or a function that does that. Thank you. Pablo Fleurquin P.S: Example 2 is a situation that might happen. If the code can be implemented to eliminate these 2 cycles that share a link it would be greate, if not it will work fine as well.

3 13

class initialization problem
by Christopher Pisz 01 Oct '11

01 Oct '11

I have a Server class inherited from a Listener class, with the latter being bases on the boost::asio example code. My problem is that the Listener class requires am boost::asio::io_service & in order to initialize its member variables. I wanted the Server class to be responsible for starting a thread, owning the io_service, and running it. It seems I cannot do that! class Server : public Listener { public: Server() : Listener(ioService_) { } // Snip protected: boost::asio::io_service ioService_; }; class Listener { public: Listener(boost::asio::io_service & ioService) : socket_(ioService) { } // Snip protected: boot::asio::ip::tcp::socket socket_; }; The above throws an unhandled exception in the initialization of the derived class. I suppose this is because ioService_ is not constructed yet? Since the C++ rules, if I remember correctly, say the base class gets constructed before the members of the derived? So, I tried: Server() : ioService_(), Listener(ioService_) { } to try and force it, but it still throws an exception. The only way to get it to not throw an exception is to create the instance to io-service in main() and then pass it as a reference into the class constructor. If I do that, then there is no way for any class in my inheritance hierarchy to own the io_service. As such, I cannot start and stop it either. Is there any way to get ownership of the io_service in my class hierarchy?

3 2