- Boost-users - lists.boost.org

is there a function can get two file path relative path?
by k 29 Apr '11

29 Apr '11

the boost filesystem library provide relative_ path() but it only give relate root directory. or is it other function can do it?

2 1

[Boost][enable_if] SFINAE and decltype
by Germán Diago 28 Apr '11

28 Apr '11

Hi all, I'm having a hard time trying to figure out how to make a code like this to improve error messages through static_assert. I want that when the function instantiation fails, the other function's static assert fails. I don't know how to do it. I began with something like this: template <class Range, class T> Range find(Range && haystack, const T & elem) { static_assert(std::is_convertible<decltype (elem == haystack.front())>::value, "elem and haystack must be equality comparable"); while (!haystack.empty() && !(haystack.front() == elem)) haystack.popFront(); return haystack; } But that didn't work since when decltype is applied to an invalid expression, SFINAE doesn't trigger in the function body. So now I'm trying an alternative like this, but I don't know how to express it: template <class Range, class T> Range find(Range && haystack, const T & elem, decltype (elem == haystack.front()) * = 0) { while (!haystack.empty() && !(haystack.front() == elem)) haystack.popFront(); return haystack; } template <class Range, class T> Range find(Range && haystack, const T & elem, !decltype (elem == haystack.front()) * = 0) { static_assert(..., "haystack.front() and elem must be equality comparable"); return haystack; } I can't figure out how to make it work. I also tried with boost::enable_if but the problem is that a bool expression doesn't have a ::value member. Could anyone suggest me THE SIMPLEST and most readable way to make this work? Thanks in advance. P.S.: By the way, I also tried structs and partial specialization, but no luck.

2 2

Read/write with gzip
by Anders Knudby 28 Apr '11

28 Apr '11

Hello, I previously managed to write data to a gzipped file using the following code structure (it's an example of course, the "memblock" is actually filled with real data and the real filename is specified elsewhere). #include <boost/iostreams/filtering_streambuf.hpp> #include <boost/iostreams/filtering_stream.hpp> #include <boost/iostreams/copy.hpp> #include <boost/iostreams/filter/gzip.hpp> #include <boost/iostreams/device/file_descriptor.hpp> #include <boost/filesystem.hpp> namespace io = boost::iostreams; int size = 1000; char* memblock = new char [size]; io::filtering_ostream out; out.push(io::gzip_compressor()); out.push(io::file_descriptor_sink("c:/outfile.gz")); out.write (memblock, size); That works fine. I would now like to read these data again. I imagined I could do something like the following (using the same #include statements etc. as above): int size = 1000; char* memblock = new char [size]; io::filtering_istream in; in.push(io::gzip_decompressor()); in.push(io::file_descriptor_source("c:/outfile.gz")); in.read (memblock, size); That complies and runs, but it doesn't give me the real data back. What am I doing wrong? Anders

4 7

boost build problem?...
by Krzysztof 28 Apr '11

28 Apr '11

I wrote a few lines of code just for start with a library of boost (program_options). Before I did it I had built boost library as prescribed and *.a and *.so files were created in my /home/krzysztof/Documents/Projects/boost/lib directory. However, compilation fails because of undefined references of some external symbols which belongs to program_options library. Could anyone help me with it? Below, there is the code and g++ output. (Linux 64b, Qt Creator and Code::Blocks used as IDE) #include "boost/program_options.hpp" using namespace boost::program_options; using namespace std; int main(int argc, char *argv[]) { options_description opd; opd.add_options() ("help","Display help") ("input_file,if",value<string>(),"Specifies input file") ("fet_file",value<string>(),"definitions file") ("output_file,of",value<string>(),"file after edition"); } g++ -o fetxedit main.o -L/usr/lib -L/home/krzysztof/Documents/Projects/boost/lib -lQtCore -lpthread main.o: In function `main': /home/krzysztof/Documents/Projects/fetxedit-build-desktop/../fetxedit/main.cpp:8: undefined reference to `boost::program_options::options_description::m_default_line_length' /home/krzysztof/Documents/Projects/fetxedit-build-desktop/../fetxedit/main.cpp:8: undefined reference to `boost::program_options::options_description::m_default_line_length' /home/krzysztof/Documents/Projects/fetxedit-build-desktop/../fetxedit/main.cpp:8: undefined reference to `boost::program_options::options_description::options_description(unsigned int, unsigned int)' /home/krzysztof/Documents/Projects/fetxedit-build-desktop/../fetxedit/main.cpp:14: undefined reference to `boost::program_options::options_description::add_options()' /home/krzysztof/Documents/Projects/fetxedit-build-desktop/../fetxedit/main.cpp:14: undefined reference to `boost::program_options::options_description_easy_init::operator()(char const*, char const*)' ... and so on -- Regards Krzysztof Jasieński

2 1

boost::bind
by simone pilozzi 28 Apr '11

28 Apr '11

Dear Users, I am using bind to calculate the transform of an initial vector in its element by element power (to a given exponent). I used this syntax std::transform(values.begin(), values.end(), values.begin(),boost::bind(pow,_1,exponent)); but the compiler complains for with error C2780 : expects 2 arguments 3 provided... Do you have any solution/explanation? Thanks in adavnce!

2 4

[iostream] gzip filter buggy on windows
by Bo Jensen 28 Apr '11

28 Apr '11

In my quest of making gzip work for my application on windows, I found the following bug : int main(int argc, char *argv[]) { std::ofstream strm("test.gz"); boost::iostreams::filtering_ostream sink; /* Push gzip compressor */ sink.push(boost::iostreams::gzip_compressor()); /* Push file writer */ sink.push(strm); #if 0 /* This does not work */ sink <<" OBJ N"<< std::endl; #else /* This works */ sink <<" OBJ N"<< std::endl; #endif } The first string results in a invalid gzip file i.e invalid compressed data--length error. I am aware this might be a bug in zib lib. It works on without any problems on linux. Does anyone use gzip on windows on a regular basis ?

2 4

Request for conditional index feature in boost multi-index
by Gokulakannan Somasundaram 28 Apr '11

28 Apr '11

Hi, First of all thanks for the great library, which had reduced our development time by a lot. In boost multi-index, we currently have almost all the indexing concepts integrated. Hence i am requesting for a new feature called conditional indexes, where in an index/view inside multi-index will only index a portion of the entries inside the entire multi-index(Say something that satisfies a functor). This is in resemblance to the partial indexes in Postgresql. I feel that would provide elegant solutions to lot of issues without forming round about solutions. For example i have a requirement where in there are two sets of data with the same search criteria, but different eviction criteria Say i have two sets of data with same search condition, but one set needs a MRU(Most Recently Used) list of 100 and the other set requires a MRU of 200. Say the entry is like this class Student { int student_no; char sex; std::string address; }; The search criteria is student_no, but for sex='m', we need MRU of 200 and for sex='f', we need a MRU of 100. Now i have a solution where in i introduce a new ordered index to maintain ordering. For example the IndexSpecifierList will be something like this typedef multi_index_container< Student, indexed_by< ordered_unique< member<Student, int, &Student::student_no> >, ordered_unique< composite_key< member<Student, char, &Student::sex>, member<Student, int, &Student::sex_specific_student_counter> > > > > student_set Here i have to introduce a new counter called sex_specific_student_counter, and during the eviction time, i need a logn search. Insertion also involves a extra logn complexity. Now instead if i have two sequenced indexes (like mentioned in the MRU example) in the place of second ordered index, but both with a condition on indexing only a few items, this would have looked more elegant and with better performance. Thanks in advance, Gokul.

3 3

stop thread before socket destructor
by Gianni Ambrosio 28 Apr '11

28 Apr '11

Hi All, I modified a boost example to implement a tcp server that's able to accept clients on a thread. Unfortunately I get the following message: "The I/O operation has been aborted because of either a thread exit or an application request". I realized that the problem is related to the fact the server destructor is called before the handle_accept occurs. Now, how could I stop the thread in the server destructor so that the handle_accept is not called anymore? Here is the test code I used: typedef boost::shared_ptr<boost::asio::ip::tcp::socket> tcp_socket_ptr; class stream_handler { public: stream_handler() : acceptor_(io_service_) { boost::asio::ip::tcp::endpoint endpoint = boost::asio::ip::tcp::endpoint(boost::asio::ip::tcp::v4(), 32123); acceptor_.open(endpoint.protocol()); acceptor_.set_option(boost::asio::ip::tcp::acceptor::reuse_address(true)); acceptor_.bind(endpoint); acceptor_.listen(); new_socket = tcp_socket_ptr(new boost::asio::ip::tcp::socket(io_service_)); acceptor_.async_accept(*new_socket, boost::bind(&stream_handler::handle_accept, this, boost::asio::placeholders::error)); thread_ = boost::thread(boost::bind(&boost::asio::io_service::run, &io_service_)); } ~stream_handler() { acceptor_.close(); } void handle_accept(const boost::system::error_code& err) { if (err) { std::cout << err.message() << std::endl; } else { new_socket = tcp_socket_ptr(new boost::asio::ip::tcp::socket(io_service_)); acceptor_.async_accept(*new_socket, boost::bind(&stream_handler::handle_accept, this, boost::asio::placeholders::error)); } } private: io_service io_service_; tcp::acceptor acceptor_; boost::thread thread_; tcp_socket_ptr new_socket; }; int _tmain(int argc, _TCHAR* argv[]) { stream_handler sh; return 0; } Regards Gianni

2 2

mpi skeleton/content does not work with stl complex
by Tobias Stollenwerk 27 Apr '11

27 Apr '11

Hello all, It seems that sending stl complex data via the skeleton/content mechanism does not work. Here is a simple code that shows th problem: #include <boost/mpi.hpp> #include <boost/serialization/complex.hpp> #include <boost/serialization/vector.hpp> #include <iostream> namespace mpi = boost::mpi; using namespace std; int main(int argc, char* argv[]) { mpi::environment env(argc, argv); mpi::communicator world; int NN=world.size(); int myid=world.rank(); if (myid == 0) { int N = 10; vector <complex<double> > l(N); for (int p=1; p!=NN; p++) { int taskid=1; world.send(p, 0, taskid); world.send(p, 1, mpi::skeleton(l)); } mpi::content c = mpi::get_content(l); for (int n = 0; n!=l.size() ; n++) { l[n]=complex<double>(1.0,1.0); } for (int p=1; p!=NN; p++) { world.send(p, 1, c); } } else { vector <complex<double> > l; mpi::content c; world.recv(0, 1, mpi::skeleton(l)); c = mpi::get_content(l); world.recv(0, 1, c); for (int n=0; n!=l.size(); n++) { cout << l[n] << " "; } cout << endl; } } The entries of the vector l are not (1.0,1.0) but they seem to be uninitialized. This only occurs if one uses complex data types AND the skeleton/content mechanism. Does anyone know, if this is a build in problem, or if I am doing something wrong? Thanks in advance, Tobias

1 0

questions on the BFS example of Parallel BGL
by Da Zheng 27 Apr '11

27 Apr '11

Hello, I'm testing PBGL on my cluster. I have 4 nodes and each node has a dual-core processor. I changed the BFS example a little bit. The code generates a graph with RMAT, and it does BFS 10 times, starting from different nodes. I first run the program in a sequential mode, and the graph has 1,000,000 vertices. Then I run it in the cluster. The result is shown below, but I don't see much speedup. The running time is reduced from 14 seconds to 9.5 seconds. Can the result be right? I see very good speedup in the PBGL document. Another problem is that BFS starting from some vertices only needs very little time. I can imagine that it traverses only very few vertices. Is there any way I can count how many vertices are visited from the specific root vertex? My code is also shown below. Thanks, Da [zhengda@amdahl36 pbgl]$ ./bfs ~/bfs.out 1000000 generate graph takes 162.336 seconds to generate graph start BFS random v: 1395111516 takes 14.3496 seconds for BFS search random v: 488302563 takes 0.029389 seconds for BFS search random v: 72509101 takes 14.1103 seconds for BFS search random v: 1835953408 takes 0.028544 seconds for BFS search random v: 951813172 takes 14.7907 seconds for BFS search random v: 1708481694 takes 14.2254 seconds for BFS search random v: 305355333 takes 14.1587 seconds for BFS search random v: 1627009882 takes 0.02749 seconds for BFS search random v: 1769559585 takes 0.027495 seconds for BFS search random v: 101184463 takes 0.02756 seconds for BFS search [zhengda@amdahl36 pbgl]$ mpirun -np 8 --hostfile ~/.mpi_hostfile ./bfs ~/bfs.out 1000000 generate graph generate graph generate graph generate graph generate graph generate graph generate graph generate graph takes 136.623 seconds to generate graph start BFS random v: 1395111516 takes 9.6362 seconds for BFS search random v: 488302563 takes 0.017129 seconds for BFS search random v: 72509101 takes 9.49864 seconds for BFS search random v: 1835953408 takes 0.031777 seconds for BFS search random v: 951813172 takes 9.53063 seconds for BFS search random v: 1708481694 takes 9.45427 seconds for BFS search random v: 305355333 takes 9.58886 seconds for BFS search random v: 1627009882 takes 0.018723 seconds for BFS search random v: 1769559585 takes 0.020844 seconds for BFS search random v: 101184463 takes 0.016004 seconds for BFS search boost::minstd_rand gen; std::cout << "generate graph" << std::endl; struct timeval start_time; struct timeval end_time; gettimeofday (&start_time, NULL); Graph g(RMATGen(gen, nvertices, 16 * nvertices, 0.57, 0.19, 0.19, 0.05), RMATGen(), nvertices); gettimeofday (&end_time, NULL); if (process_id(process_group(g)) == 0) { std::cout << "takes " << (end_time.tv_sec - start_time.tv_sec) + ((double) end_time.tv_usec - start_time.tv_usec) / 1000000 << " seconds to generate graph" << std::endl; std::cout << "start BFS" << std::endl; } for (int i = 0; i < 10; i++) { int rv = gen(); if (process_id(process_group(g)) == 0) { std::cout << "random v: " << rv << std::endl; } graph_traits<Graph>::vertex_descriptor start = vertex(rv % nvertices, g); // Compute BFS levels from vertex 0 property_map<Graph, vertex_distance_t>::type distance = get(vertex_distance, g); put(distance, start, 0); gettimeofday (&start_time, NULL); breadth_first_search (g, start, visitor(make_bfs_visitor(record_distances(distance, on_tree_edge())))); gettimeofday (&end_time, NULL); if (process_id(process_group(g)) == 0) { std::cout << "takes " << (end_time.tv_sec - start_time.tv_sec) + ((double) end_time.tv_usec - start_time.tv_usec) / 1000000 << " seconds for BFS search" << std::endl; } }

1 0