September 2010 - Boost-users

Re: [Boost-users] Problem using serialization::xml_archive withpolymrphic classes
by przemyslaw.sliwa＠uk.bnpparibas.com 03 Sep '10

03 Sep '10

Thanks for reply, All classes a, b, c and d are actually defined within one dll (one module). I forgot to mention it. I can do this following thing: d* pD = mew d(); and then ar & boost::serialization::make_nvp("item", pD); but c* pD = new D(); ar & boost::serialization::make_nvp("item", pD); does not work at all! It throws this strange type of exceptions which I cannot handle at all. Another problem is that if I do the first version the member m_key variable of true_type in the call const boost::serialization::extended_type_info * true_type = i.get_derived_extended_type_info(t); is a null pointer which is not right. Do you have any idea what could be wrong? Thanks a lot for help ___________________________________________________________ This e-mail may contain confidential and/or privileged information. If you are not the intended recipient (or have received this e-mail in error) please notify the sender immediately and delete this e-mail. Any unauthorised copying, disclosure or distribution of the material in this e-mail is prohibited. Please refer to http://www.bnpparibas.co.uk/en/information/legal_information.asp?Code=ECAS-… for additional disclosures.

2 1

Boost.Functional/Hash
by Andrej van der Zee 03 Sep '10

03 Sep '10

Hi, I have a question that I hope can be answered by using Boost. I have the following datastructure: struct ip_info { const u_int8_t _ifile_index; const unsigned long _src_ip; const unsigned long _dst_ip; const u_int16_t _id; const u_int16_t _length; const u_int16_t _src_port; const u_int16_t _dst_port; const u_int32_t _seq; const u_int32_t _ack_seq; }; The object is identified by all its members. Now I want to count duplicates. But I have MANY of these objects, I mean billions. I have to reduce the memory footprint of my program, because it soon grows over 4GB and boom! So I rather not keep all the objects in memory and why would I, because I do not need them. I just need to count duplicates. My idea was to hash such objects to a smaller values and use that hash value as key_type in a map. The value_type of the map would be an integer. This something like std::map<int32_t, u_int16_t> or something. Then I though maybe use Boost.Functional/Hash, but are there any guarantees that the hash value is unique? How unique are they? Are there better alternatives? Thank you, Andrej

4 3

[BGL] Using pointers as indexes in graphs
by Bruno Barberi Gnecco 02 Sep '10

02 Sep '10

Hi, I want to store pointers in a graph. Can I somehow use their pointers as indexes? Here's the code I'm trying to execute. It does not work correctly, however; apparently graph add other nodes in the vector storage (to fill the holes?). I suppose I could use listS instead of vecS, but then I can't get the topological_sort code to work (I know I have to provide some form of indices, but I don't understand how to do that). Any suggestions to solve this? Thanks! #include <iostream> #include <boost/graph/graph_traits.hpp> #include <boost/graph/adjacency_list.hpp> #include <boost/graph/topological_sort.hpp> class Plugin { public: // some data here int x; Plugin(int a) : x(a) {} }; typedef std::list< std::pair<Plugin *, Plugin *> > EdgeList; typedef boost::adjacency_list<boost::setS, boost::vecS, boost::directedS, Plugin *> PipelineGraph; typedef boost::graph_traits<PipelineGraph>::vertex_descriptor PipelineVertex; void graph(EdgeList edges) { PipelineGraph graph; for (EdgeList::iterator it(edges.begin()); it != edges.end(); it++) { PipelineVertex a(boost::vertex(reinterpret_cast<unsigned long> (it->first), graph)); PipelineVertex b(boost::vertex(reinterpret_cast<unsigned long> (it->second), graph)); boost::add_edge(a, b, graph); } // sort by topology, so we run the plugins in the correct order std::vector<PipelineVertex> sorted; boost::topological_sort(graph, std::back_inserter(sorted)); std::cout << sorted.size() << std::endl; // REVERSE topological order std::vector<PipelineVertex> vl(sorted.rbegin(), sorted.rend()); // loop for (std::vector<PipelineVertex>::iterator it(vl.begin()); it != vl.end(); it++) { Plugin *p = reinterpret_cast<Plugin *>(*it); std::cout << p->x << std::endl; } } int main() { EdgeList edges; Plugin *a = new Plugin(1); Plugin *b = new Plugin(2); Plugin *c = new Plugin(3); Plugin *d = new Plugin(4); edges.push_back(std::pair<Plugin*, Plugin*>(a, b)); edges.push_back(std::pair<Plugin*, Plugin*>(b, c)); edges.push_back(std::pair<Plugin*, Plugin*>(b, d)); graph(edges); return 0; }

2 3

how to get std::min_element() working with boost::multi_array?
by Valmor de Almeida 02 Sep '10

02 Sep '10

Hello, Could someone help getting std::min_element() (or a variation of) to work with multi_array? boost::multi_array<int, 2> A; // code for dimensioning A and filling in; deleted cout << *min_element(A.begin(),A.end()) << endl; Thanks, -- Valmor

1 1

Looking for Halfedge or other topologic representation library
by Lang Stefan 02 Sep '10

02 Sep '10

Hello everyone, While looking for an existing implementation of halfedge data structures or similar topological representations of 2-dimensional manifolds I found one article pointing to boost:graph. I've checked the documentation, but it seem that this library is way to generic/abstract for my purpose, and yet seems to miss a number of important functions that probably only make sense when working with a topological representation of a two-dimensional oriented manifold. Conceptually, Half-edge data structures would fit my needs very well, and I've found several sites with half-done or half-documented implemetations. But those that actually had code were either too specific to be useful for my purposes, or half of the site couldn't be retrieved, probably because the relevant pages are no longer hosted. So I was wondering, if there is, maybe, another library within boost that is dealing more with the topological aspect of graphs, whether or not based on boost::graph? What I am looking for is a topological framework for oriented surfaces that lets me retrieve adjacency information regarding faces, edges and vertices, disregardful of whether or not the set of surfaces represent the full surface of a 3D object or just a part of it. I also need to directly navigate between the objects defining the actual 3D points, curves, and surfaces, and the corresponding topological objects. But that, I hope, is just a minor detail. I suspect it won't be so hard to implement myself, but building on a well-designed library is so much less error prone... Cheers, Stefan

3 2

Re: [Boost-users] does pool::release_memory() really require an ordered
by Lang Stefan 02 Sep '10

02 Sep '10

-----Ursprüngliche Nachricht----- Message: 5 Date: Wed, 1 Sep 2010 17:32:46 -0400 From: Mark Rajcok <mrajcok(a)gmail.com> To: boost-users(a)lists.boost.org Subject: [Boost-users] does pool::release_memory() really require an ordered pool? Message-ID: <AANLkTim0t9KTDJowTbn2viBu+RSTqnwfKA1c7qRbv_X2(a)mail.gmail.com> Content-Type: text/plain; charset=ISO-8859-1 On page http://www.boost.org/doc/libs/1_44_0/libs/pool/doc/interfaces/pool.html there is this precondition for release_memory(): "t must be ordered". Does that mean that ordered_malloc() must always be used, if I want to use release_memory()? Looking at the implementation for release_memory(), it appears to be iterating through chunks in each block, which I wouldn't expect if ordered_malloc() were always used. What I'm really after: I want to use object_pool, but periodically I'd like to reclaim/release any memory blocks that are no longer in use. In my application, I may have traffic spikes, where I allocate a lot of memory, but then later (after spiking) I'd like to release it. (The object pools will persist for the life of the application, and there will probably always be some "active objects" -- i.e., allocated from the pool and not free()d yet.) (I'm hoping the documentation may be out of date, and release_memory() can be used on pools that don't use ordered_malloc() and on object_pools.) -- Mark ------------------------------ Mark, I cannot check the repository right now, but last time I did there were no changes to the version of boost::pool since I last tried to work with it, about three years ago. Back then I found that memory wasn't released at all due to a relevant part of the code being commented out. But even uncommenting it didn't help all that much: just like you, I found that the effort of releasing a single object was, essentially, proportional to the size of the pool, at best at an order of O(log(N)), maybe even worse. So just using standard new and delete operators would at some point outperform this pool implementation! I've done my own implementation in the meantime, with all operations at constant time, and using shared pointers to objects from the pool, which would make sure unused memory blocks can be released ASAP. Never got around to actually use it though, so I cannot say for sure if it has errors still. The tests and benchmarks I ran did look rather good, but I found that using malloc/free instead of new/delete got me about the same performance boost, which is why I ultimately discarded it. (malloc and free were already being used anyway, not that I was happy with it. But unfortunately I ran out of arguments to replace them...) Cheers, Stefan

1 0

[BGL] Segmentation fault in clear_vertex
by Daniel Trebbien 02 Sep '10

02 Sep '10

I am attempting to modify my implementation of the Stoer–Wagner min-cut algorithm to modify the input graph. In the code sample below, t and s are vertex_descriptors that were calculated by a call to boost::detail::stoer_wagner_phase: boost::tie(s, t, w) = boost::detail::stoer_wagner_phase(g, assignments, weights, pq); assert(s != t); // ... std::multimap<vertex_descriptor, edge_descriptor> sOutEdges; BGL_FORALL_OUTEDGES_T(s, e, g, UndirectedGraph) { const vertex_descriptor u = target(e, g); if (t != u) { sOutEdges.insert(std::make_pair(u, e)); } } BGL_FORALL_OUTEDGES_T(t, e, g, UndirectedGraph) { const vertex_descriptor u = target(e, g); const weight_type eWeight = get(weights, e); typename std::multimap<vertex_descriptor, edge_descriptor>::iterator sOutEdgesIt, sOutEdgesEnd; boost::tie(sOutEdgesIt, sOutEdgesEnd) = sOutEdges.equal_range(u); if (sOutEdgesIt != sOutEdgesEnd) { for(; sOutEdgesIt != sOutEdgesEnd; ++sOutEdgesIt) { put(weights, sOutEdgesIt->second, get(weights, sOutEdgesIt->second) + eWeight); } } else if (t != u) { // need to update e to set the source to s edge_descriptor ePrime; boost::tie(ePrime, boost::tuples::ignore) = add_edge(s, u, g); put(weights, ePrime, eWeight); } } clear_vertex(t, g); The segmentation fault occurs in the call to clear_vertex, but does not occur if I comment out the line that adds an edge to the graph g. The type of g, by the way, is: boost::adjacency_list<boost::listS, boost::vecS, boost::undirectedS, boost::no_property, boost::property<boost::edge_weight_t, int> > Why does adding an edge between vertices s and target(e, g), both of which are not equal to t, cause a segmentation fault? For reference, gdb says that the fault occurs on line 999 of boost/graph/detail/adjacency_list.hpp: // O(E/V * E/V) template <class Config> inline void clear_vertex(typename Config::vertex_descriptor u, undirected_graph_helper<Config>& g_) { typedef typename Config::global_edgelist_selector EdgeListS; BOOST_STATIC_ASSERT((!is_same<EdgeListS, vecS>::value)); typedef typename Config::graph_type graph_type; typedef typename Config::edge_parallel_category Cat; graph_type& g = static_cast<graph_type&>(g_); typename Config::OutEdgeList& el = g.out_edge_list(u); typename Config::OutEdgeList::iterator ei = el.begin(), ei_end = el.end(); for (; ei != ei_end; ++ei) { detail::erase_from_incidence_list (g.out_edge_list((*ei).get_target()), u, Cat()); g.m_edges.erase((*ei).get_iter()); *// line 999* } g.out_edge_list(u).clear(); } Program received signal SIGSEGV, Segmentation fault. std::_List_node_base::unhook (this=0x3d7a38) at ../../../../gcc-4.4.0/libstdc++-v3/src/list.cc:134 134 ../../../../gcc-4.4.0/libstdc++-v3/src/list.cc: No such file or director y. in ../../../../gcc-4.4.0/libstdc++-v3/src/list.cc (gdb) bt #0 std::_List_node_base::unhook (this=0x3d7a38) at ../../../../gcc-4.4.0/libstdc++-v3/src/list.cc:134 #1 0x0045fa26 in std::list<boost::list_edge<unsigned int, boost::property<boost ::edge_weight_t, int, boost::no_property> >, std::allocator<boost::list_edge<uns igned int, boost::property<boost::edge_weight_t, int, boost::no_property> > > >: :_M_erase (this=0x22fa9c, __position=...) at c:/mingw/bin/../lib/gcc/mingw32/4.4.0/include/c++/bits/stl_list.h:1424 #2 0x0045fa0c in std::list<boost::list_edge<unsigned int, boost::property<boost ::edge_weight_t, int, boost::no_property> >, std::allocator<boost::list_edge<uns igned int, boost::property<boost::edge_weight_t, int, boost::no_property> > > >: :erase (this=0x22fa9c, __position=...) at c:/mingw/bin/../lib/gcc/mingw32/4.4.0/include/c++/bits/list.tcc:111 #3 0x00413ec9 in boost::clear_vertex<boost::detail::adj_list_gen<boost::adjacen cy_list<boost::listS, boost::vecS, boost::undirectedS, boost::no_property, boost ::property<boost::edge_weight_t, int, boost::no_property>, boost::no_property, b oost::listS>, boost::vecS, boost::listS, boost::undirectedS, boost::no_property, boost::property<boost::edge_weight_t, int, boost::no_property>, boost::no_prope rty, boost::listS>::config> (u=5, g_=...) at c:/mingw/bin/../lib/gcc/mingw32/4.4.0/../../../../include/boost/graph/det ail/adjacency_list.hpp:999 #4 0x0041b0cc in boost::detail::stoer_wagner_min_cut<boost::adjacency_list<boos t::listS, boost::vecS, boost::undirectedS, boost::no_property, boost::property<b oost::edge_weight_t, int, boost::no_property>, boost::no_property, boost::listS> , boost::adj_list_edge_property_map<boost::undirected_tag, int, int&, unsigned i nt, boost::property<boost::edge_weight_t, int, boost::no_property>, boost::edge_ weight_t>, boost::associative_property_map<std::map<int, bool, std::less<int>, s td::allocator<std::pair<int const, bool> > > >, boost::shared_array_property_map <unsigned int, boost::vec_adj_list_vertex_id_map<boost::no_property, unsigned in t> >, boost::d_ary_heap_indirect<unsigned int, 4u, boost::shared_array_property_ map<unsigned int, boost::vec_adj_list_vertex_id_map<boost::no_property, unsigned int> >, boost::shared_array_property_map<int, boost::vec_adj_list_vertex_id_map <boost::no_property, unsigned int> >, std::greater<int>, std::vector<unsigned in t, std::allocator<unsigned int> > > > (g=..., weights=..., parities=..., assignments=..., pq=...) ...

1 1

does pool::release_memory() really require an ordered pool?
by Mark Rajcok 02 Sep '10

02 Sep '10

On page http://www.boost.org/doc/libs/1_44_0/libs/pool/doc/interfaces/pool.html there is this precondition for release_memory(): "t must be ordered". Does that mean that ordered_malloc() must always be used, if I want to use release_memory()? Looking at the implementation for release_memory(), it appears to be iterating through chunks in each block, which I wouldn't expect if ordered_malloc() were always used. What I'm really after: I want to use object_pool, but periodically I'd like to reclaim/release any memory blocks that are no longer in use. In my application, I may have traffic spikes, where I allocate a lot of memory, but then later (after spiking) I'd like to release it. (The object pools will persist for the life of the application, and there will probably always be some "active objects" -- i.e., allocated from the pool and not free()d yet.) (I'm hoping the documentation may be out of date, and release_memory() can be used on pools that don't use ordered_malloc() and on object_pools.) -- Mark

2 1

Struggling to implement bzip2 on windows
by Jared James Thompson 02 Sep '10

02 Sep '10

Hi All, I'm a VS2008 user, trying to use the ESBTL library for a project, which implements Boost to do many operations. ESBTL #includes <boost/iostreams/filter/bzip2.hpp>, which has been disabled for the Boostpro installation of windows. Quite simply, I'm struggling to get bzip2 installed and set up so that boost runs with bzip2 enabled. Has anybody done this before? Can anyone offer me more information than "compile from source?" best, Jared -- Jared James Thompson Department of Medicinal Chemistry and Molecular Pharmacology Laboratory for Computational Drug Design and Biology RHPH 504C Heine Pharmacy Building 575 Stadium Mall Drive West Lafayette, IN 47907-2091

3 3

lamba: QStringList <<
by Hicham Mouline 02 Sep '10

02 Sep '10

Hello, the following for_each call with _1 fails to compile #include <QStringList.h> #include <boost/lambda/lambda.hpp> using namespace boost::lambda; QStringList items_list; std::for_each(begin_items(),end_items(), items_list << _1 ); QStringList has these << operators defined: QStringList& operator<< ( const QString& str ); QStringList& operator<< ( const QStringList& other ); the type of item iterator returned by begin_items() is: const char* const* QString has a ctor: QString (const char * str) The compilation error is: error C2678: binary '<<' : no operator found which takes a left-hand operand of type 'const QStringList' (or there is no acceptable conversion) I don't see why it tries to find a << that takes a ***const*** QStringList left-hand operand? Regards,

2 1