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

[serialization] binary archiver with generic primitive types
by Florent Teichteil 31 Aug '09

31 Aug '09

Hi all, The binary_oarchive and binary_iarchive classes serialize an object to a stream whose buffer type is streambuf=basic_streambuf<char>. It means that the serialized object is translated into an array of chars. It is fine if you can work with an array containing chars. Unfortunately, some libraries need serialized objects that are translated into arrays whose type is different from char. For instance, I'm using a library that requires an array of unsigned chars. For that reason, I have defined generic (templated) tbinary_oarchive and tbinary_iarchive classes based on the ones provided by the boost serialization library. It is straightforward since the original boost serialization classes use the std::ostream type that is just an instantiation of the generic basic_ostream class with the char type (ostream=basic_ostream<char>). For instance, the tbinary_oarchive class is now defined as follows: // CODE BEGIN template <class Tdata> class tbinary_oarchive : public binary_oarchive_impl< tbinary_oarchive<Tdata>, typename std::basic_ostream<Tdata>::char_type, typename std::basic_ostream<Tdata>::traits_type > { public: tbinary_oarchive(std::basic_ostream<Tdata> & os, unsigned int flags = 0) : binary_oarchive_impl< tbinary_oarchive, typename std::basic_ostream<Tdata>::char_type, typename std::basic_ostream<Tdata>::traits_type >(os, flags) {} tbinary_oarchive(std::basic_streambuf<Tdata> & bsb, unsigned int flags = 0) : binary_oarchive_impl< tbinary_oarchive, typename std::basic_ostream<Tdata>::char_type, typename std::basic_ostream<Tdata>::traits_type >(bsb, flags) {} }; // CODE END Then, when I need to use a binary archiver with the type unsigned char, I just put the following lines in my code: // CODE BEGIN typedef boost::archive::tbinary_oarchive<unsigned char> BinaryOBArchive; typedef boost::archive::tbinary_iarchive<unsigned char> BinaryIBArchive; BOOST_SERIALIZATION_REGISTER_ARCHIVE(BinaryOBArchive) BOOST_SERIALIZATION_REGISTER_ARCHIVE(BinaryIBArchive) // CODE END This compiles fine except that I get some linker errors concerning the basic_binary_oarchive and the basic_binary_oprimitive classes. For instance: undefined reference to `boost::archive::basic_binary_oprimitive<boost::archive::tbinary_oarchive<unsigned char>, unsigned char, std::char_traits<unsigned char> >::~basic_binary_oprimitive()' It seems that parts of the serialization library's code is outside the definition classes, preventing to extend it with different binary types... Is there a workaround? Thank you very much for your patience reading this e-mail entirely, Florent.

4 4

[Graph] Error in cuthill_mckee_ordering in 1.40
by Ryan W. Frenz 31 Aug '09

31 Aug '09

I am attempting to upgrade to boost 1.40 and have found an issue with cuthill_mckee_ordering.hpp. The following code demonstrates the issue. #include <boost/graph/adjacency_matrix.hpp> #include <boost/graph/cuthill_mckee_ordering.hpp> int main() { typedef boost::property<boost::vertex_name_t, std::string> VertexProperty; typedef boost::adjacency_matrix<boost::undirectedS, VertexProperty> MyGraph; typedef boost::graph_traits<MyGraph>::vertex_descriptor Vertex; MyGraph graph(0); // Fill graph with nodes, etc.... // Perform reverse cuthill mckee... std::vector<Vertex> inv_perm(boost::num_vertices(graph)); { boost::cuthill_mckee_ordering(graph, inv_perm.rbegin()); // 'has_no_vertices' error... // do stuff with ordered graph... } return 0; } This code builds cleanly against 1.36 (my previous version) but fails when built against 1.40 with the following: ....boost/graph/cuthill_mckee_ordering.hpp(172): 'has_no_vertices': identifer not found Looking at the code, the function boost::cuthill_mckee_ordering is calling has_no_vertices (with no specified namespace). has_no_vertices (in graph_utility.hpp) lives in the boost::graph namespace. Changing the line to graph::has_no_vertices fixes the compiler error. Is this possibly a bug in 1.40 or am I doing something wrong? Thanks, Ryan

2 2

fast_pool_allocator, std::multi_map and purge_memory
by Soeren Meyer-Eppler 31 Aug '09

31 Aug '09

Hi, I have written a small test program to measure the performance benefit of using a boost::fast_pool_allocator vs the default system allocator in conjunction with a std::multi_map. Memory overhead is ~16% lower and the boost allocator is ~4x faster - sweet. However, I can't release the allocated memory via the singleton_pool interface as described in the documentation. In my test program quoted below the call to purge_memory always returns 0 and according to window's task manager the memory has not been released. An even simpler test just using an allocator for ints and allocating/relasing/purging 10.000.000 ints works as expected. What am I doing wrong? cheers, Sören PS: I'm using boost 1.39 with MSVC2008 #include <map> #include <iostream> #include <boost/pool/pool_alloc.hpp> int main( int, char ** ) { typedef boost::fast_pool_allocator< std::pair< int, void * > > TPoolAllocator; typedef std::multimap< int, void *, std::less< int >, TPoolAllocator > TMap; { TMap map; for ( int i = 0; i < 10000000; ++i ) map.insert( std::make_pair( rand() % 100, (void *)0 )); } std::cout << boost::singleton_pool< boost::fast_pool_allocator_tag, sizeof( TPoolAllocator::value_type ) >::purge_memory(); return 0; }

2 2

Cannot file tickets for 1.40
by Roland Bock 31 Aug '09

31 Aug '09

Hi, I would like to file a few tickets for version 1.40, but TRAC does not know version 1.40 yet. Regards, Roland

2 1

Linker errors with boost's threading and python libs, OSX
by Shamyl Zakariya 31 Aug '09

31 Aug '09

Hi, Yesterday I upgraded boost to 1.4, from 1.36. A big jump, but hey why not. I'm running OSX, 10.6 with the most recent dev tools ( Xcode 3.2, GCC 4.2, etc ) and used the bjam build system as described on the website. Vanilla install to /usr/local/ giving me the static and dynamic libs in /usr/local/lib. Two (related) questions: 1) My project ( built in Xcode ) uses mostly header-based libs, but does link against libboost_thread.a -- and now it throws all sorts of linker errors when it worked just fine linking against 1.36 For example: "boost::thread::~thread()", referenced from: void boost::checked_delete<boost::thread>(boost::thread*)in QTFrameTransfer.o /// many many more And for reference, I really *am* linking against libboost_thread.a 2) My related followup question -- I've *never* been able to link against libboost_python.a. So in the past I added the full source from $BOOST/libs/python/src to a sub-target of my project and manually build my own libboost_python.a which I then linked against. That worked fine. But I don't understand why I can't link against /usr/ local/lib/libboost_python.a I have to assume something's wonky with the way bjam is building these libraries. But I'm following the out-of-the-box install instructions. Finally, I assume I could just build my own local boost the way I used to build boost's python. But I'd rather use the versions in /usr/local/ lib... shamyl zakariya - squamous and rugose

3 2

Advise sought for container type
by Archie14 30 Aug '09

30 Aug '09

given the class like the one below: template <typename T> class animal { public: animal (const char* ofname) : _name(ofname) {} private: typedef T _T; std::string _name; _T behavior; } class MeatEater {}; class GrassEather {}; animal<MeatEater> bear ("bear"); animal<GrassEater> rabbit ("rabbit"); How to store "bear" and "rabbit" instances in a vector? Advise is greatly appreciated.

3 4

Chain of Responsibility Pattern in a boost library
by Ryan McConnehey 30 Aug '09

30 Aug '09

Is there a boost library that implements the Chain of Responsibility Pattern? Ryan

1 0

[mapreduce] Prim Calculator
by Christian Henning 30 Aug '09

30 Aug '09

Hi there, first of all, I'm not sure if boost user list should be used for boost-sandbox libraries. Anyone? Now, I would like create a new example for the mapreduce lib. How can I, for instance, setup a system that computes the prim numbers for 0 to 1,000,000,000. For this, I need to create the input for the map function on the fly, rather than reading some files from the hard drive. How is that possible? The documentation is a bit spare so far. Thanks, Christian

2 22

[proto] folding a proto tree in reverse
by Maurizio Vitale 30 Aug '09

30 Aug '09

I need to traverse a proto tree from right to left so that given the statement: (a, b, c) = expr; I can assign in order c then b and lastly a. A fragment that does the order a,b,c (and work, but not very efficiently, which is why I want to go the other way) is: template<typename E> struct assign_aux { assign_aux (const E& rhs) : m_rhs (rhs) {} typedef int result_type; template<typename T> int operator()(const T& t, int offset) const { int next_offset = offset - actual_width_transform () (t); int value = systemc_eval (m_rhs.range (offset,next_offset+1)); const_cast<T&>(t) = value; return next_offset; } const E& m_rhs; }; template <typename E> struct assign<E, proto::tag::comma> { template<typename RHS> const E& doit (const E& e, const RHS& rhs) { int width = actual_width_transform () (e); fusion::fold(proto::flatten (const_cast<systemc_expression&>(e)), width-1, assign_aux<RHS> (rhs)); return e; } }; template<typename E> const systemc_expression& operator = (const E& e) const { BOOST_MPL_ASSERT ((proto::matches<systemc_expression, lhs_systemc_grammar>)); return assign<systemc_expression>().doit (*this, e); return *this; } I've tried to use fusion::reverse_view to invert the fusion sequence, but failed. I also suspect that the resulting code wouldn't be as good as it could be, which in my application is critical. So I looked at alternatives and tried proto::reverse_fold_tree. Here's my attempt: struct assign_terminal : proto::callable { typedef int result_type; template<typename T, typename O, typename R> int operator () (T& t, O o, const R& r) const { int width = actual_width_transform () (t); const_cast<T&>(t) = r.range (1,0); return o+width; } }; template <typename E> struct assign<E, proto::tag::comma> { template<typename RHS> const E& doit (const E& e, const RHS& rhs) { proto::reverse_fold_tree<_, int() , assign_terminal (_, proto::_state, proto::_data)>() (e, rhs); return e; } }; template<typename E> const systemc_expression& operator = (const E& e) const { BOOST_MPL_ASSERT ((proto::matches<systemc_expression, lhs_systemc_grammar>)); return assign<systemc_expression>().doit (*this, e); return *this; } But this doesn't seem to pass the data (rhs) to assign_terminal::operator () (expr, state, data). There're really very few examples on how to use proto::fold_tree and even fewer using the data parameter, so I don't have much to go on. Would anybody whose name is Eric tell me why I'm this stupid? TIA, Maurizio

3 3

boost::multi_index_container vector-like O(1) time indexing question
by zap.foster 30 Aug '09

30 Aug '09

I originally posted this to the wrong group, but I intended it for this group. I have a boost::multi_index question. My current container has looked like the following for a very long time: typedef std::map<double,int> mtype_t; What I am looking for is a way to index the mtype_t::value_type.first value so that, syntactically, it appears like a vector. For example, the container I need would let me do something like: { container_t c; c.insert( container_t::value_type( keycount_t( 1.414,88) )); c.insert( container_t::value_type( keycount_t( 3.141,77) )); c.insert( container_t::value_type( keycount_t( 23.000,99) )); for(int index=0; index<c.size(); ++index) printf("%d\n",c[index].second); } You can see that I want to be able to index into the container much like one would do with a vector. But I want my indexing to be O(1) and not O(n). Now I still need to be able to iterate the container, and I need to be able to do efficient insertions, and yes, the contents of the container will get bigger and smaller and the values vary quite a lot. A basic sketch of the multi_index container (but without the vector-like index) would be: typedef struct { double key; int count; } keycount_t; struct by_key{}; typedef boost::multi_index_container< keycount_t, boost::multi_index::indexed_by < boost::multi_index::ordered_unique < boost::multi_index::tag<by_key>, boost::multi_index::member< keycount_t,double,&keycount_t::key > > > > container_t; This gives me the basics. But I cannot think of a way to add a secondary index that would give me the O(1) vector-like indexing. A use-case of the grammar I am after would be something like the following: void useless_function(container_t& c) { for(container_t::iterator i = c.begin(); i!=c.end(); ++i) print("key=%f\tval=%d\n",i->first,i->second); for(index=0; index<c.size(); ++index) print("key=%f\tval=%d\n",c[index].first,c[index].>second); } Here's WHY I need this: I have a stream of floating point values coming in (fixed to 5 significant digits). I need to keep count of how often any one of the floating point values is repeated. Very often I need to respond to requests as to the contents of the container at the Nth position. The slow O(n) method would be: const keycount_t& get_nth_record(int index,const container_t& c) { container_t::const_iterator i; for(i=c.begin(); i!=c.end(); ++i) if(--index==0 ) break; return( *i ); } But again, I need O(1) time, so I need some sort of index added to my container_t. Has anyone done this sort of thing before, and met the O(1) positional index requirement? Many thanks! -zap

2 1

Intercommunicators with Boost.MPI ?
by Ashika Umanga Umagiliya 30 Aug '09

30 Aug '09

Greeting all, I am completly new to Boost and this is my very first post here . I have used Boost.Serialization and Boost.MPI to send user-defined-datatypes across MPI processes. (I use OpenMPI 1.3.3 ) Now, I want to spawn processes (MPI_Comm_spawn) and send across these objects among newly created child processes. How can I handle Intercommunicators with Boost.MPI ? Thanks in advance, umanga

1 0

circular_buffer: How to make push_back() increase capacity?
by Robert Dailey 30 Aug '09

30 Aug '09

Hello, I have a list of colors that I need to be able to cycle through, either by rotating the container or by using an iterator that will rotate (i.e. return to the beginning when I increment at the end, and vice-versa). The goal is to be able to iterate through the container in a circular way and process all of the colors repeatedly. I chose a boost::circular_buffer for this job, I assumed it was the perfect tool for the job. It works great, except I would like to be able to call push_back() to add all of my colors to the buffer without having to call set_capacity() on it first. It's a minor thing, but omitting the call to set_capacity() would make it slightly more scalable/manageable since I won't have to keep incrementing the size of the capacity each time I want to add colors to the buffer. How can I achieve this? --------- Robert Dailey

4 3

[boost-users][Variant] Sequence instantiation
by Ryan McConnehey 30 Aug '09

30 Aug '09

I have a sequence of types defined with the Loki library. I believe the equivalent would be boost::mpl::vector. The sequence is used to cause compile time asserts for unsupported types being passed to template functions. Boost variant can be created with a given list of types. typedef boost::variant< int, std::string, short, long > type_list; | <boost/variant.html>| Is it possible to create a variant from a sequence by either passing the sequence or iterating through each value in the sequence? This would reduce the code duplication of having one sequence for assert checking and the other for variants. Ryan

3 12

Using BGL::read_graphml()
by Ramapriya Rangachar 30 Aug '09

30 Aug '09

Hello Boost users, I have managed to build the latest version (1.40.0) under Windows/MSVC. Could someone kindly point me to a working example that complements the link below, for read_graphml()? http://www.boost.org/doc/libs/1_40_0/libs/graph/doc/write_graphml.html Best regards, Ram _________________________________________________________________ Windows Live: Make it easier for your friends to see what you’re up to on Facebook. http://windowslive.com/Campaign/SocialNetworking?ocid=PID23285::T:WLMTAGL:O…

1 0

[asio] DTR serial port option?
by Robert Dailey 30 Aug '09

30 Aug '09

Hello, I'm rewriting a C# program in C++. This C# program previously was communicating with a device via serial port. One thing in particular it was doing was setting the DTR flag to enabled. The MSDN article on this option is here: http://msdn.microsoft.com/en-us/library/system.io.ports.serialport.dtrenabl… <http://msdn.microsoft.com/en-us/library/system.io.ports.serialport.dtrenabl…>I do not see an option in the Boost.Asio documentation that is identical to this C# option. Can anyone comment on this? How can I enable DTR through Boost Asio's serial_port? --------- Robert Dailey

4 10

Re: [Boost-users] [Graph] Problems with adjacency_matrix
by Juan Antonio Farré Basurte 29 Aug '09

29 Aug '09

> > > Hello, > I have a problem with adjacency_matrix template. > It claims to implement VertexAndEdgeListGraphConcept, BidirectionalGraphConcept and MutablePropertyGraphConcept. > But when I try to compile, it complains about using the next functions with an adjacency_matrix: > > num_vertices(graph) > vertices(graph) > num_edges(graph) > edges(graph) > source(edge_descriptor, graph) > target(edge_descriptor, > graph) > > and also with > > get(&V::property, graph), where V is a vetex bundled-properties type get(&E::property, > graph), where E is an edge bundled-properties type > > (these get functions should return a property map for the corresponding bundled property > and they work correctly for adjacency_list, just like the rest). > > Is this a bug or an error in the documentation? > > Thanks a lot, > > Juan > Mmmm... I just tried to include adjacency_matrix.hpp and now I don't get errors about those functions anymore. It's strange to have to inclue this header in a template library that doesn't explicitly deal with any concrete implementations of graph concepts. In fact, I don't need to include adjacency_list.hpp for it to work with adjacency_list. I just need to include graph_concepts.hpp and graph_traits.hpp. Anyway, I'm still getting lots of errors in a code that perfectly compiles and works with adjacency lists. If anybody is interested, I can send the listing of compiler errors. Thanks again, Juan

9 20

[1.40] Release of PDF docs for Boost-1.40 now available.
by John Maddock 29 Aug '09

29 Aug '09

For those who prefer printable docs, PDF builds of the docs from Boost-1.40 are now available from the usual sourceforge download site: https://sourceforge.net/projects/boost/files/ (look under the boost-docs subtree). Direct links are as follows: zip (19.2Mb): https://sourceforge.net/projects/boost/files/boost-docs/1.40.0/boost_pdf_1_… tgz (19.7Mb): https://sourceforge.net/projects/boost/files/boost-docs/1.40.0/boost_pdf_1_… 7z (17.9Mb): https://sourceforge.net/projects/boost/files/boost-docs/1.40.0/boost_pdf_1_… Please note that not all libraries have documentation suitable for automatic PDF generation yet, libraries covered are: Accumulators Any Array Asio Bimap Bjam Boost.Build Range Boostbook TR1-complex number algorithms Concepts Config Conversion Date_time Foreach Function FunctionTypes Fusion Hash Interprocess Intrusive Iterator Lambda Math (special functions and distributions) Math-gcd MPI Octonion Optional Phoenix ProgramOptions Proto Python tutorial Quaternion Quickbook Ref Regex ScopeExit Signals Signals2 StaticAssert StringAlgo Thread TR1 Tribool TypeTraits Typeof Units Unordered Variant Xpressive Enjoy! John Maddock.

2 2

ASIO - Using IO service from main program in a .so loaded with dlopen, dlsym, etc.
by J.W.F. Thirion 29 Aug '09

29 Aug '09

Dear Community, I've got an application that loads a .so using dlopen, dlsym, etc. It also starts up an IO service pool and manages them. The application implements a socket server, that starts a thread for each incoming connection. Inside each of the threads started, a .so function is called based on information received on the socket. Inside the function of the .so, I want to connect to a remote server using the IO service in the main program (that called the dlopen, etc.). It seems, however, that if I use an IO service not created inside the .so, the handle_connect never gets called by ASIO. Is anyone aware of this problem. I suppose I'm just doing something stupid w.r.t. shared object usage (maybe permissions, etc.). Or is the ASIO library not initialised inside the DLL? When I create an IO service pool inside the .so, everything works, but obviously this is not ideal as the pool starts threads that execute the run() method on the io service. Interesting thing is that I can actually pass a session object (from the socket server accept operation) from my main program to the .so function and have messages read and written on the socket. But when I try and create a new connection using the session's IO service (e.g. from a sock.get_io_service() call), then I never get the handle_connect. I've even tried: #pragma GCC visibility push(default) #include <boost/asio.hpp> #include <boost/bind.hpp> #include <boost/thread.hpp> #pragma GCC visibility pop around the boost includes, but no luck. Any help on this would be greatly appreciated. Kind regards, Derik Thirion

2 2

implementing a new Range
by Alex Flint 29 Aug '09

29 Aug '09

Hi there, I'm using a math library that provides a Vector class, but it does not have begin() or end() method, or any other normal container methods. I would like to pass it to functions that take a Range object as a parameter. How can I do this? Cheers, Alex

6 10

[Fwd: Re: [interprocess] shared_memory_object design]
by Ion Gaztañaga 29 Aug '09

29 Aug '09

2 1

pkg-config and FreeBSD
by Roman Shmelev 29 Aug '09

29 Aug '09

Hi! I've downloaded boost 1.38 from the official site and successfully build it. I was able to compile the program that uses regexes. OS is FreeBSD 7.2. gcc is 4.2.1. The question is: how can I do something like this: pkg-config --libs boost-regex I'm new to FreeBSD, so I guess I really miss something basic. Googling for subject gave me a lot but not instructions how to do what I want. Thank you

3 10

Boost C++ Math Toolkit
by Paul A. Bristow 29 Aug '09

29 Aug '09

I am pleased to announce publication of the Boost Math Toolkit PDF documentation as a book (in CD format initially). Boost C++ Math Toolkit, ISBN 0-9504833-2-X John Maddock, Paul A. Bristow, Hubert Holin, Xiaogang Zhang, Bruno Lalande, Johan Råde, Gautam Sewani, Thijs van den Berg Publisher: hetp Publication date: 1 Sep 2009 Suggested Classification 519.2-dc22 If you are using this software, and writing academic or scholarly articles, you can reference it as above. (If you an author you can add it to your CV ;-) In the unlikely event that anyone would like this as a CD, I can of course provide it (at a modest cost) - but then you can download everything for free from www.boost.org or www.sourceforge.net. The CD also includes a Windows Installer program for the Distribution Explorer (but you can also download from http://distexplorer.sourceforge.net/) Please tell me if you would like to be able to buy a printed version. Thank you. Paul --- Paul A. Bristow Prizet Farmhouse Kendal, UK LA8 8AB +44 1539 561830, mobile +44 7714330204 pbristow(a)hetp.u-net.com

1 0

[BGL] possible Bug in property handling
by Ralf Goertz 29 Aug '09

29 Aug '09

Hi, I'm new to BGL (and generic programming) so this may well not be a bug but I don't know how this could be intended. Consider the following program: #include <string> #include <iostream> #include <boost/graph/graph_traits.hpp> #include <boost/graph/properties.hpp> #include <boost/graph/adjacency_list.hpp> using namespace boost; typedef property<vertex_name_t, std::string> VertexProp; //this crashes typedef adjacency_list < vecS, vecS, undirectedS,VertexProp, no_property > graph_t; //this doesn't //typedef adjacency_list < vecS, listS, undirectedS,VertexProp, no_property > graph_t; typedef graph_traits < graph_t >::vertex_descriptor vertex_descriptor; typedef property_map<graph_t, vertex_name_t>::type id_name_map_t; typedef graph_traits<graph_t>::vertex_iterator vi_t; std::pair<vi_t,vi_t> vi; int main(){ graph_t g; vertex_descriptor vd; id_name_map_t name_map; vd=add_vertex(g); put(name_map,vd,"foo"); for (vi=vertices(g); vi.first!=vi.second;++vi.first) { std::cout<<get(name_map,*vi.first)<<std::endl; } return EXIT_SUCCESS; } From what I read in the BGL book the storage type of the edges shouldn't matter. But here using vecS gives a segfault whereas listS works fine. This is the backtrace of a gdb session: Program received signal SIGSEGV, Segmentation fault. #0 0x00007ffff79316e6 in std::string::assign(char const*, unsigned long) () from /usr/lib64/libstdc++.so.6 #1 0x00000000004016ef in boost::put<boost::vec_adj_list_vertex_property_map<boost::adjacency_list<boost::vecS, boost::vecS, boost::undirectedS, boost::property<boost::vertex_name_t, std::string, boost::no_property>, boost::no_property, boost::no_property, boost::listS>, boost::adjacency_list<boost::vecS, boost::vecS, boost::undirectedS, boost::property<boost::vertex_name_t, std::string, boost::no_property>, boost::no_property, boost::no_property, boost::listS>*, std::string, std::string&, boost::vertex_name_t>, std::string&, unsigned long, char [4]> ( pa=@0x7fffffffdcc0, k=0, v=@0x403693) at /usr/include/boost/property_map.hpp:321 #2 0x0000000000400ee8 in main () at bug.cc:26 Version information boost-devel 1.39.0 $: g++ -v Using built-in specs. Target: x86_64-suse-linux Configured with: ../configure --prefix=/usr --infodir=/usr/share/info --mandir=/usr/share/man --libdir=/usr/lib64 --libexecdir=/usr/lib64 --enable-languages=c,c++,objc,fortran,obj-c++,java,ada --enable-checking=release --with-gxx-include-dir=/usr/include/c++/4.3 --enable-ssp --disable-libssp --with-bugurl=http://bugs.opensuse.org/ --with-pkgversion='SUSE Linux' --disable-libgcj --disable-libmudflap --with-slibdir=/lib64 --with-system-zlib --enable-__cxa_atexit --enable-libstdcxx-allocator=new --disable-libstdcxx-pch --enable-version-specific-runtime-libs --program-suffix=-4.3 --enable-linux-futex --without-system-libunwind --with-cpu=generic --build=x86_64-suse-linux Thread model: posix gcc version 4.3.2 [gcc-4_3-branch revision 141291] (SUSE Linux)

3 5

boost::asio::io_service and posix message queues
by Sergio Martinez 28 Aug '09

28 Aug '09

I'm working on a project where we want a single thread to asynchronously interact with posix message queues, TCP sockets and drivers. From what I've read, boost::asio::io_service provides a common framework dealing with asynchronous IO. It looks like there's a asio library for TCP and posix::stream_descriptor could be used for interacting with a driver file descriptor. What I don't see is support for posix message queues. The io_service documentation states the following: "The io_service class also includes facilities intended for developers of custom asynchronous services.". I was thinking of adding a custom service for posix message queues. Has anyone already done this? Is there any fundamental reason why it could not be done? (i.e. message queues are not byte streams, etc.) Thanks, Sergio

1 1

[proto] I'm lost
by Mail Delivery System 28 Aug '09

28 Aug '09

I need a transform property_b that evaluates another transform property_a and returns what property_a returns+1. Simple enough, here's what I do: struct property_a : proto::or_< proto::when< proto::terminal<_>, mpl::int_<42>()> > {}; struct property_b : proto::or_ < proto::when< _, mpl::next<property_a(_) >() > > {}; The complete file is attached at the end of this message as "Working example". But when I move to my real application (of which I attach a very condensed version as "Non-working example") I cannot manage to make it compile. The code in #ifdef BUG...#endif gives an error, the #else branch doesn't (but doesn't compute what I need, which makes the fact it compiles secondary). I cannot see differences, other than my terminals now being in their own domain. H E L P TIA, Maurizio --=-=-= Content-Type: text/x-c++src Content-Disposition: inline; filename=test_transforms.cpp Content-Description: Working example #include <iostream> #include <boost/mpl/int.hpp> #include <boost/mpl/assert.hpp> #include <boost/mpl/min_max.hpp> #include <boost/proto/core.hpp> #include <boost/proto/context.hpp> #include <boost/proto/transform.hpp> namespace mpl = boost::mpl; namespace proto = boost::proto; using proto::_; template<typename T> struct data { typedef T type; }; proto::terminal<data<mpl::int_<42> > >::type a; struct property_a : proto::or_< proto::when< proto::terminal<_>, mpl::int_<42>()> > {}; struct property_b : proto::or_ < // proto::when< _, property_a> proto::when< _, mpl::next<property_a(_) >() > > {}; int main(int argc, char *argv[]) { std::cout << "property A = " << property_a () (a) << std::endl; std::cout << "property B = " << property_b () (a) << std::endl; return 0; } /// Local Variables: /// mode:c++ /// comment-column:60 /// fill-column:150 /// compile-command:"g++ -I. -I./boost -o test_transforms test_transforms.cpp" /// c-macro-cppflags:"-C -I. -I./boost" /// c-backslash-column:120 /// c-backslash-max-column:149 /// End: --=-=-= Content-Type: text/x-c++src Content-Disposition: inline; filename=width.cpp Content-Description: Non-working example //* Includes #include <iostream> //#include <iomanip> //#include <sstream> #include <limits.h> #define BOOST_PROTO_MAX_LOGICAL_ARITY 15 #include <boost/proto/proto.hpp> #include <boost/utility.hpp> // for enable_if #include <boost/mpl/min_max.hpp> #include <boost/type_traits/is_same.hpp> #include <boost/type_traits/is_arithmetic.hpp> #include <boost/integer_traits.hpp> #include <boost/mpl/assert.hpp> #include <boost/typeof/typeof.hpp> namespace quicksilver { namespace mpl = boost::mpl; namespace fusion = boost::fusion; namespace proto = boost::proto; using proto::_; template<typename E> class my_expression; template<typename R> struct data; template<typename B> struct representation_traits; template<int L, int R> struct bounds_c; template<typename E> int left (const E&); // template<typename E> struct result_of_expression; // template<typename E> struct representation_of_expression; struct actual_width_transform; struct unspecified {}; struct builtin_types_grammar : proto::or_ < proto::terminal<bool>, proto::terminal<int>, proto::terminal<unsigned int>, proto::terminal<long>, proto::terminal<unsigned long>, proto::terminal<long long>, proto::terminal<unsigned long long> > {}; struct my_grammar : proto::or_ < proto::terminal<data<_> >, proto::unary_expr<_, my_grammar>, proto::binary_expr<_, my_grammar, my_grammar>, builtin_types_grammar > {}; struct my_domain : proto::domain<proto::generator<my_expression>, struct my_grammar> {}; template <typename Expr> struct my_expression : proto::extends<Expr, my_expression<Expr>, my_domain> { typedef proto::extends<Expr, my_expression<Expr>, my_domain> base_type; my_expression (Expr const& expr_ = Expr()) : base_type (expr_) {} }; struct left_bound_transform : proto::or_< proto::when<proto::terminal<data<_> >, mpl::int_<42>() > > {}; struct actual_width_transform : proto::or_< //#define BUG #ifdef BUG mpl::next<left_bound_transform (_)>() #else left_bound_transform #endif > {}; template<typename L, typename R, typename I=short> struct bounds { typedef L meta_left; }; template<int L, int R> struct bounds_c : bounds<mpl::int_<L>, mpl::int_<R> > {}; template<typename Bounds> struct representation_traits { typedef Bounds bounds; }; template<typename R> struct data : R::bounds { typedef R representation; }; template<typename R> struct number : my_expression<typename proto::terminal<data<R> >::type> { typedef my_expression<typename proto::terminal<data<R> >::type> base_type; number () : base_type (base_type::proto_base_expr::make (data<R> ())) {} }; } // end of namespace quicksilver #include <boost/test/included/unit_test_framework.hpp> using namespace boost::unit_test; using namespace quicksilver; void test_width_transform () { typedef number<representation_traits<bounds_c<6,0> > > number_type; number_type i7; BOOST_MPL_ASSERT ((proto::matches<number_type, left_bound_transform>)); BOOST_MPL_ASSERT ((proto::matches<number_type, actual_width_transform>)); std::cout << "left (i7) = " << left_bound_transform () (i7) << std::endl; BOOST_ASSERT (left_bound_transform () (i7) == 6); std::cout << "width (i7) = " << actual_width_transform () (i7) << std::endl; } test_suite* init_unit_test_suite (int, char**) { test_suite* test= BOOST_TEST_SUITE( "Quicksilver" ); test->add (BOOST_TEST_CASE (&test_width_transform)); return test; } /// Local Variables: /// mode:c++ /// comment-column:60 /// fill-column:150 /// compile-command:"g++ -I. -I./boost -o width width.cpp -lgmpxx -lgmp" /// c-macro-cppflags:"-C -I. -I./boost" /// c-backslash-column:120 /// c-backslash-max-column:149 /// End: --=-=-=--

3 2