- Boost-users - lists.preview.boost.org

[mpl] Any mpl tricks to use default tags and get around function template argument deduction with function defaults?
by jesseperla 10 Jul '09

10 Jul '09

I am building a make_ function for a complicated type with many generic parameters that all have defaults. I believe the following doesn't work in C++03 (but might in C++0X?) template<class T> void myfunc(T t = default_type()) {} if called with no arguments. Here is pseudocode for what I would like to do (unworking, so please bear with any mpl mistakes) struct default_tag{}; struct default_type // metafunction to return the default type to keep it centralized. { typedef int type; type construct() //A construction function? { return type(); } }; template<class T> class myclass { myclass(const T& t){....} //Requires a type }; template<class T> myclass< mpl::if_<mpl::is_same<T, default_tag>, //returns a my_class<> default_type, //call this metafunction if default T > //otherwise pass it in. > make_myclass(const T& t = default_tag()) //Note use of the default_tag { typename myclass< mpl::if_<mpl::is_same<T, default_tag>, default_type, //call this metafunction T >::type resolved_type; //this is what we want to use //Here I am confused of how to conditionally construct a type of default_type::type to pass in? // I thought it might help to have a default_type.construct function but wasn't sure what to do? //Note that T is often too large to do a full 'copy'. It won't be an int. //Pass in t directly or a constructed version of resolved_type if T if 'default'? return my_class<resolved_type>( ...); } The usage I would like to enable: auto m1 = make_myclass(); //Uses the default auto m2 = make_myclass( default_tag() ); //Uses the default as well. Why do I need this? Will actually have a number of these tags where you may want to conditionally choose some as defaults. int n = 5; auto m3 = make_myclass(n); Any ideas on the best idioms to support this and get around the template argument detection problems? Is this possible? If not, can I drop the need for the make_myclass(); usage and still keep the ability to call with a default and conditionally constructs the default object based ont eh type of a metafunction? Thanks, Jesse

1 0

Re: [Boost-users] [optional] How to make boost::optional throw if trying to access uninitialized value
by Anders Dalvander 10 Jul '09

10 Jul '09

dariomt(a)gmail.com wrote: > Thanks for the info! > > I think it is not possible to have a specific boost::assertion_failed handler > for boost::optional and other handlers for other assertion failures in other > places, right? > > That would mean that the exception I throw from the handler cannot mention the > fact that I am accessing an uninitialized value in a boost::optional. > > Maybe I should create my own my::safe_optional based on boost::optional but > checking for uninitialized access. > > Any ideas? > You could create your own functions for this: template <typename T> T const& safe_get(boost::optional<T> const& opt) { if (!opt) throw some_exception(); return *opt; } template <typename T> T& safe_get(boost::optional<T>& opt) { if (!opt) throw some_exception(); return *opt; } Regards, Anders Dalvander

1 0

boost::asio::async_read_until composite completion condition
by Etienne Philip Pretorius 10 Jul '09

10 Jul '09

Hello List, I need to implement a completion condition for boost::asio::async_read_until where either a boost::regex is matched or boost::asio::transfer_at_least is matched. Any suggestions will be much appreciated. Kindly, Etienne

2 4

A question on data change with lambda expression. Can any one with kindness help me?
by fmingu 10 Jul '09

10 Jul '09

I am using Dev-C++ and lambda expression.I wrote the following lines: .................. typedef std::map<int, int> IIPrimemap; IIPrimemap primemapvec; ..................... void a(){ primemapvec[1]=3; bind(&IIPrimemap::value_type::second,(*bind(static_cast<IIPrimemap::iterator (IIPrimemap::*)()> (&IIPrimemap::end),var(primemapvec))))++; return; } .................... but after the program ran, the primemapvec[1] is still 3, not 4. I am puzzled about that. How can I change primemapvec to 4 ? Can anyone with kindness help me?

1 0

[Thread] (Newb) shared memory in multithread process
by gabe_rosser 10 Jul '09

10 Jul '09

Dear All, Apologies for my ignorance but I would be extremely grateful if someone would kindly point me in the right direction. I am trying to parallelize my Monte Carlo simulation. The bottleneck in the process is the generation of Normally distributed random numbers (using boost.random libraries), so I was hoping to substantially improve the efficiency of the process by setting 2 or 3 cores of my quad core processor to the the task of generating these numbers while one core gets on with the rest of the algorithm. The simplest way I can envisage doing this is by allocating shared memory that is simultaneously written (by rand number generating threads) and read (by main thread), with some method of ensuring that the main thread doesn't read too far ahead (perhaps locking or just a simple while loop?). Boost.thread alone doesn't seem to be enough - an attempt to do exactly this resulted in a substantial slow down (with or without locking - without may result in undefined behaviour but I tried it anyway to see if it was causing the slowdown). Do I need to investigate shared pointers? I really need some large container in shared memory which I can fill in any order, much as would be the case if I were appending to and reading from a file, but hopefully faster! Any suggestions please? Sorry for the lack of understanding! Thanks, Gabe -- View this message in context: http://www.nabble.com/-Thread--%28Newb%29-shared-memory-in-multithread-proc… Sent from the Boost - Users mailing list archive at Nabble.com.

3 4

[Boost.Program_options] extra_parser for config files
by Marti Cuquet 10 Jul '09

10 Jul '09

Hi, I want to implement the --switch / --no-switch behavior. I defined a bool option "switch" and then I parse the command line with an extra parser that sets "switch" to "false" if it finds a --no-switch. myopts.add_options() ("switch", po::value<bool>()->implicit_value(true), "Use switch.") ; po::store (po::command_line_parser(argc, argv).options(all).extra_parser(parse_no_options).run(), vm); The extra parser is: std::pair<std::string, std::string> parse_no_options (const std::string& s, const std::string name) { if (s.find("--switch") != s.npos) { if (s.substr (2, 3) == "no-") return std::make_pair (s.substr (5), std::string("false")); else return std::make_pair (s.substr (2), std::string("true")); } return std::make_pair (std::string(), std::string()); } This works fine for the command line. I also need to implement this for the config file, but I do not know how to pass the extra parser to the parse_config_file. Is there a way? Thanks, Marti

2 1

[Threads] Test for running thread
by Krzysiek 10 Jul '09

10 Jul '09

Hi, I use following code to test running thread. bool isRunning = !thread.timed_join ( boost::posix_time::milliseconds ( 0 ) ); Is it safe and portable? Thanks, Krzysiek

1 0

[interprocess] Container
by Moritz 10 Jul '09

10 Jul '09

Hi, is there a possibility to put something like this into the shared_memory? class A { int a; double d; // ... boost::interprocess string s; } Thanks in advance.

2 3

[asio] Is io_service::run() reentrant?
by bearclaw 10 Jul '09

10 Jul '09

Hi, I cannot figure out from the documentation if I can call io_service::run() from an asynchronous handler, called itsels by run() of the same io_service. If it is possible, should I call reset() or not? I tried and it seems to work under linux, but I need to call run() in a loop as it returns prematurely, even with an io_service::work(), and with/without a reset(). The reason I need to do this is because my asynchronous handlers are calling legacy code that needs to perform synchronous operations. Using synchronous socket operations while this lasts as it was suggested to handle this kind of situations would be possible, but a lot more complicated in my setup. A pseudocode example of what I am trying to do in case I'm not clear: io_service io; tcp::socket s(io); onRead2() { // We got our data, stop inner io::run invocation io.stop(); } onRead(socket& s) { s.write("request"); //we need the result of the request synchronously // I know this looks silly, but it makes more sense in my real-world // use case where the async_read is burried deep. s.async_read(&onRead2); // optional io.reset() here, has no effect. io.run(); // <-- ends prematurely } main() { <connect s > s.async_read(&onRead) io.run(); } Regards, Matthieu

1 0

shared_ptr: struct of 3 shared_ptrs: is operator= for struct required?
by Hicham Mouline 10 Jul '09

10 Jul '09

Hello Copying structs by default means memberwise copy. If I have a struct S { shared_ptr<T1> ptr1; shared_ptr<T2> ptr2; shared_ptr<T3> ptr3; }; and i have 2 instances s1 and s2 s1=s2 will call shared_ptr<T.> 's operator=(), right? Regards,

2 1

[optional] How to make boost::optional throw if trying to access uninitialized value
by dariomt＠gmail.com 10 Jul '09

10 Jul '09

Hi list, I've seen in the docs that boost::optional asserts if you try to access an uninitialized value. Is it possible to make it throw an exception instead? Even in release mode? TIA

3 2

How can I save the values in lambda expression? Can any one with kindness help me?
by fmingu 10 Jul '09

10 Jul '09

I am using Dev-C++ with boost lambda expression, I wrote: .................... int nnumber; int switchvalue; typedef std::map<int, int> IIPrimemap; IIPrimemap primemapvec; # define BetaKGre2 1 # define EqualToOne 2 #define ProductOfKPrime 3 ..................... if_then_else(var(nnumber)==1,var(switchvalue)=EqualToOne, if_then_else(constant(count_if(primemapvec.begin(),primemapvec.end(), bind(&IIPrimemap::value_type::second,_1)>2)>0), var(switchvalue)=BetaKGre2,var(switchvalue)=ProductOfKPrime)); std::cout<<" the var(switchvalue) is "<<var(switchvalue)<<"\n"; .............. and the program ran and showed: the var(switchvalue) is without any expression of var(switchvalue). I do not know how to save the values in lambda expression and where is wrong? Can I use *var(switchvalue) instead of var(switchvalue)? Can anyone with kindness help me?

4 6

bootstrap.sh generates incorrect libdir in project-config.jam
by Andreas Klöckner 10 Jul '09

10 Jul '09

Hi all, When I ran bootstrap.sh --prefix=$HOME/pool, it did generate a broken project- config.jam--it used option.set libdir : /lib ; instead of option.set libdir : /home/andreas/pool/lib ; Has anybody else run into the issue described here? Any chance for fixed tarballs? Thanks, Andreas

2 1

[Boost.Program_options] zero or one tokens with zero_tokens()?
by Marti Cuquet 09 Jul '09

09 Jul '09

Hello, I am trying to use an option that can have either zero or one arguments. I have seen an example in the documentation claiming that the following code would allow the user to provide either zero or one tokens to option 'verbose', but instead it throws an exception "extra parameter" when one token is given. So the question is, does zero_tokens() allow the user to specify a value? options_description desc; desc.add_options() ("help", "produce help message") ("compression", value<string>(), "compression level") ("verbose", value<string>()->zero_tokens(), "verbosity level") ("email", value<string>()->multitoken(), "email to send to") ; Thanks, Marti

2 2

[Tuple] Addressing by named tag?
by Robert Jones 09 Jul '09

09 Jul '09

Am I correct in believing that tuples have to be accessed by element index, as in tuple.get<0>() Multi-Index container has a really nice facility to use named (well typed really) tags in these kind of expressions, but it seems tuples do not. Anyone fancy adding it to tuple? - Rob.

2 3

boost programmer_option.hpp
by Narendra R 09 Jul '09

09 Jul '09

Hi, i was trying to use boost program option I can compile the code written using the program_option.hpp, but i am getting linker error can any one help me? Using gcc compiler

2 1

Possible to put path to boost library somewhere else than Jamroot file?
by s seelenluft 09 Jul '09

09 Jul '09

Hi, my question is whether I can put the path to the boost library in a another file than the Jamroot file (eg, boost-build.jam)? I run run longish simulations on multiple workstations, essentially Debian and Mac OS X machines. For good or bad reasons, the path to the boost library is differs between machines. I give the path via an <include> in the Jamroot file. Thus the Jamroot file has to be different on each machine. However, my code is synchronized via Subversion between the machines which poses a problem for the Jamroot file. Thus, I either can only do one-way synchronisation from one master computer and edit the Jamroot file on the other machines locally (marking the conflict as resolved after the first update after this edit) or exclude the Jamroot file from Subversion. Since I probably do modify the Jamroot file from time to time, it would be nice to include it in Subversion. Currently, I solve the problem by sticking to one-way synchronisation but if there would be another place to put the path, this would be nicer. Thanks. Markus

4 7

[BGL] Large graphs
by Sebastian Weber 09 Jul '09

09 Jul '09

Hi! I am using the BGL to represent a fractal structure which I iteratevley grow. As I intend to handle large structures I usually preallocate the required space for the final fractal in the beginning. My construction process then uses the graph object to put in new edges into the same graph object. Now, if I always use the same graph object things get messed up for large fractals. However, if I use 2 graph structures, one which represents the last and one the comming generation, everything is ok (which doubles memory requirements for nothing). I have no clue what the problem here is. For small fractals both schemes produce correct fractals, but not for large ones. I suspect that there are some really strange problems here with allocated memory going on. BTW, is there a way to preallocate the required memory for the edges of a adjacency_list graph? I do know ahaed how many edges in total and how many edges per vertex will be in the final graph, so that I could preallocate memory if it is possible. Preallocating vertices is easy... In case someone may have an idea, but needs the code I am having trouble with, no problem, just contact me. Thanks in advance, Sebastian

2 1

[graph] named vertices?
by Noah Roberts 09 Jul '09

09 Jul '09

I notice that in the changes for the newest boost that "named vertexes" is now supported in adjacency_list. I can't find anything on this feature though. I have the book, have searched the web. Only thing I find is comments in adjacency_list.hpp that maybe refers to a NamedGraph concept? I don't see that in the docs either: http://www.boost.org/doc/libs/1_39_0/libs/graph/doc/table_of_contents.html Could someone give me an overview of what this feature is and why I might use it? Probably better...point me to some place that documents it?

2 1

[boost-build][doc][boostbook][doxygen] Problems with using boost documenting tools
by Andrey Torba 09 Jul '09

09 Jul '09

I've attached simple project with documentation. There are two problems:1. doxygen generates documentation with full paths to headers <D:/c++/my_project/include/my_project/h1.hpp> instead of <my_project/h1.hpp> As you can see in attachment, i configure doxygen by defining STRIP_FROM... directives, where $(TOP)/include = D:/c++/project/include: <doxygen:param>STRIP_FROM_PATH=$(TOP)/include <doxygen:param>STRIP_FROM_INC_PATH=$(TOP)/include TOP variable i define in Jamroot as: path-constant TOP : . ; What do i do wrong? 2. Paths to images looks like <img src="../../doc/html/images/next.png" alt="Next">. Path goes up by two directories and then down, it seem through the Jamroot I need simple <img src="images/next.png" alt="Next"> How to generate forward relative paths? -- Regards, Andrey

1 0

write_graphviz requires not only VertexListGraph, but also EdgeListGraph
by Richard Ulrich 09 Jul '09

09 Jul '09

Hi everybody, this is my first post to the boost mailing list. I'm using boost_1_39 I wanted to use write_graphviz today with my custom graph adapter that implements VertexListGraph as well as IncidenceGraph. The documentation states, that write_graphviz only requires a VertexListGraph. But inside, it uses an edge_iterator, and calls edges(g), both of which belong to the EdgeListGraph. Am I right here, or did I miss something? If yes, should I implement a specialization that only calls out_edges() for every vertex, and thus a VertexListGraph would be sufficient? With kind regards Richard

3 2

Null shared pointer confusion
by Carl Hetherington 09 Jul '09

09 Jul '09

Hi, In the following program... #include <boost/shared_ptr.hpp> #include <boost/weak_ptr.hpp> class A {}; int main () { boost::shared_ptr<A> x; // A: bad_weak_ptr thrown // boost::shared_ptr<A> x ((A *) 0); // B: works boost::weak_ptr<A> y (x); boost::shared_ptr<A> z (y); } ... I observe that with line A included (but not line B), a bad_weak_ptr is thrown by the line which declares z. With line B included (but not line A), all is well. >From looking at the boost source, it seems that line A results in a shared_ptr to nothing (use_count() == 0), and line B results in a shared_ptr to 0 (use_count() == 1). I can work around this, I'm sure. Can anyone clarify my muddled thinking and explain why boost is behaving correctly in this situation (if indeed it is!) Many thanks Carl

2 1

[serialization] 1.34.1 -> 1.35 vector bug
by Josh Kelley 08 Jul '09

08 Jul '09

I'm upgrading from C++Builder 2007 with Boost 1.34.1 to C++Builder 2009 with Boost 1.35. (I realize that this is a few versions behind, but C++Builder compatibility seems problematic enough that I'm going to try sticking with 1.35, which is what the vendor officially supports. Or would people here recommend using a later version?) I'm running into problems loading XML archives containing vectors that were created by Boost 1.34.1. This appears to be the same problem that others have reported: https://svn.boost.org/trac/boost/ticket/1774 https://svn.boost.org/trac/boost/ticket/2271 http://lists.boost.org/boost-users/2009/05/48292.php If I understand those pages correctly, to work around this problem, I should be able to copy the relevant changes from Boost 1.39.0's boost/serialization/vector.hpp and boost/serialization/array.hpp, plus change BOOST_SERIALIZATION_VECTOR_VERSION from 3 to 4, and my copy of 1.35 will work both with 1.34.1 and with 1.36 and above. Is this correct? Thanks. Josh Kelley

1 0

boost::random distribution is not working within a custom class
by Oliver Fochler 08 Jul '09

08 Jul '09

Hey there, I'm having a problem with the boost::random library and I can't figure out what's going on, so maybe you guys might help me out. Here's the scenario (code below): I wrote a small class that should encapsulate a random number generator (floating point, uniform in [0,1) ) so that I ultimately can switch between different approaches (boost and non-boost) without the rest of my code ever caring. Within this class I have a boost::mt19937 object and a pointer to a boost::uniform_01<boost::mt19937&, double> as private members. Within the constructor I actually create an instance of uniform_01 via new and initialize it with the mt19937 RNG. Now within the constructor (see cout statement) the ran_boost->operator()() or (*ran_boost)() work perfectly and yield some random double between 0 and 1. However, when using the randomNumberGenerator object created in main.cpp I'm struck with segfault. Doing some backtrace it turns out that it fails with in boost::random::mersenne_twister<unsigned int, 32, 624, 397, 31, 2567483615u, 11, 7, 2636928640u, 15, 4022730752u, 18, 3346425566u>::operator() ( this=0x89485ed18949ed31) at /home/fochler/usr/include/boost/random/mersenne_twister.hpp:275 If I try to use global object ran2 (declared extern in random2.h and defined in random2.cpp) I also get a segfault, this time with in boost::random::detail::pass_through_engine<boost::random::mersenne_twister<unsigned int, 32, 624, 397, 31, 2567483615u, 11, 7, 2636928640u, 15, 4022730752u, 18, 3346425566u>&>::base (this=0x0) at /home/fochler/usr/include/boost/random/detail/pass_through_engine.hpp:42 42 base_type& base() { return helper_type::ref(_rng); } I don't really know (a) what is going on and what I've messed up and (b) whether these two errors (using the local and the global object) are connected. It would be great if somebody could help me out. Cheers, Oliver Some minimal example ----------------- random2.h ----------------- #ifndef RANDOM_NG_H #define RANDOM_NG_H #include <stdint.h> #include <boost/random.hpp> class randomNumberGenerator { public: randomNumberGenerator( const uint32_t s ); double operator()() { return ran_boost->operator()(); } private: boost::mt19937 rng_mt_boost; boost::uniform_01<boost::mt19937&, double> * ran_boost; }; extern randomNumberGenerator ran2; #endif ----------------- random2.cpp ----------------- #include <iostream> #include <boost/random.hpp> #include "random2.h" randomNumberGenerator ran2; randomNumberGenerator::randomNumberGenerator( const uint32_t s ) { boost::uniform_01<boost::mt19937&, double> * ran_boost = new boost::uniform_01<boost::mt19937&, double>( rng_mt_boost ); rng_mt_boost.seed( s ); std::cout << ran_boost->operator()() << std::endl; } ----------------- main.cpp ----------------- #include "random2.h" int main(int argc, char **argv) { randomNumberGenerator ran3( 1234567 ); double a = ran2(); double b = ran3(); return 0; }

3 4

Serialize a Tree
by 张炜 08 Jul '09

08 Jul '09

i want to serialize a self-defined tree to a binary file and vice-versa. The data structure is as follows: class Node { public: Node(); Node(int type,const string &name); ~Node(); string getNodeName(); int getNodeType(); Node *getNodeParent(); bool setNodeName(string &name); bool setNodeType(int type); bool setNodeParent(Node *parent); ............................................................. private: friend class boost::serialization::access; template<class Archive> void serialize(Archive & ar, const unsigned int version) { ar & type; ar & name; } int type; string name; Node *parent; public: vector<Node *> children; // children list of current node vector<AttribNode *> attribList; // attribute list of current node }; class AttribNode { public: AttribNode(); AttribNode(string attribName,int attribType,const char* value); string getAttribName(); int getAttribType(); private: friend class boost::serialization::access; template<class Archive> void serialize(Archive & ar, const unsigned int version) { ar & name; ar & type; ar & asciiValue; ar & binSize; } string name; int type; string asciiValue; long binSize; char *binValue; }; the difficult is how to serialize the relation between different Nodes.and the binValue. while in the memory, there has a tree.the save function is as follows: bool ConfigRegistry::saveConfigTreeToLocal() { ofstream ofs("D:\\Config.bin",ios::binary); boost::archive::binary_oarchive oa(ofs); queue<Node *> nodeQueue; nodeQueue.push((Node *)root); while(!nodeQueue.empty()) { Node * temp = nodeQueue.front(); for (int i = 0;i < temp->children.size();i++) { nodeQueue.push(temp->children.at(i)); } oa << *temp; int attribNum = temp->attribList.size(); oa << attribNum; for(int i =0;i < temp->attribList.size();i++) { oa << *(temp->attribList.at(i)); if (temp->attribList.at(i)->getAttribType() == ATTRIBUTE_TYPE_BINARY) { char * ptr = temp->attribList.at(i)->binValue; long size = temp->attribList.at(i)->binSize; oa << ptr[size]; } } int childrenNum = temp->children.size(); oa << childrenNum; nodeQueue.pop(); } return true; } bool ConfigRegistry::getConfigTreeFromLocal() { ofstream ifs("D:\\Config.bin",ios::binary); boost::archive::binary_oarchive ia(ifs); int childNum = 0; int attribNum = 0; queue<Node *> nodeQueue; root = (void *)new Node(); Node *curNode = (Node *)root; nodeQueue.push(curNode); while(!nodeQueue.empty()) { ia >> *(curNode); ia >> attribNum; while(attribNum--) { AttribNode * attribNode = new AttribNode(); ia >> *attribNode; if (attribNode->getAttribType() == ATTRIBUTE_TYPE_BINARY) { attribNode->binValue = new char[attribNode->binSize]; ia >> attribNode->binValue; } else { attribNode->binValue = NULL; attribNode->binSize = 0; } curNode->attribList.push_back(attribNode); } ia >> childNum; while (childNum--) { Node * newNode = new Node(); curNode->addChild(newNode); nodeQueue.push(newNode); } nodeQueue.pop(); curNode = nodeQueue.front(); } return true; } the saveConfigTreeToLocal() is okay. but the getConfigTreeFromLocal() is error.y? how can i serialize the self-defined tree above?

2 1