January 2009 - Boost-users - lists.preview.boost.org

Help with a boost::lambda problem and boost::array?
by jesseperla＠gmail.com 23 Jan '09

23 Jan '09

Does anyone have an idea why the following doesn't compile: using lambda::_1; using lambda::_2; boost::function<double (const boost::array<double, 2>&)> f = lambda::ret<double>(_1[0] * _1[1]); But the following does as a double-check: using lambda::_1; using lambda::_2; boost::function<double (const double&, const double&)> f2 = lambda::ret<double>(_1 * _2); Thanks, Jesse

5 8

WinCE family
by Marat Abrarov 22 Jan '09

22 Jan '09

Does anybody can describe what happens with the development process of support of WM 5-6x and WinCE 5-6? What is the current state of affairs? It would make boost more popular if boost supports (either particular, but official) such modern and famous mobile platforms like Symbian (S60), WM (5 and above), WinCE (5 and especially 6), Android. I understand that it's rather difficult to support such a different case of platforms, but boost (for me) ~ Standard C++ Library, and if Qt supports S60 3rd edition, WM 5-6, WinCE 5-6, Linux Embedded - why boost can't? Best regards, Marat Abrarov.

3 2

boost::interprocess shared memory performance
by Andy Wiese 22 Jan '09

22 Jan '09

This is my first experience with using shared memory for anything more than trivial IPC. Thanks again to Ion Gaztañaga for getting the library working on FreeBSD. I'm developing initially on OS-X 10.5.6 with boost_trunk, but the eventual target platform is FreeBSD--I'm just an old mac programmer used to cushy development tools. I originally used managed_mapped_file, and got everything working, but I was disappointed by performance. Profiling with shark, I found that the application was spending a lot of time in msync, which according to man(2) is used to synchronize mapped memory with the filesystem. It makes sense to me that the interprocess containers I was creating could have a lot of file I/O overhead, and managed_mapped_file was probably not a good choice. So I rewrote the code to use managed_shared_memory instead of managed_mapped_file, thinking that it would eliminate the file I/O and therefore be faster. However, I am surprised that it is not much faster, and when I profile with Shark on os-x I see that it is still spending a lot of time in msync, specifically whenever a managed_shared_memory object is destroyed. (in boost::interprocess::mapped_region::flush(unsigned long, unsigned long), which is called within basic_managed_shared_memory's destructor) Does managed_shared_memory really need to call msync? I see that I should optimize my code to cache the managed_shared_memory objects so that fewer create/deletes are necessary, but this is still going to happen fairly frequently and I wonder if this expensive msync call is necessary. In the tradition of coder forums everywhere, someone will probably ask what I'm trying to accomplish and whether there may be a better way. Suggestions welcome. I'm writing a little cgi driven database utility that queries data stored in a filesystem directory using a simple query language. I would like to keep indexes of the data to speed query resolution. The utility is old-school cgi, so all its resources (such as indexes) have to be instantiated into memory each time the cgi process is started. I could write indexes to files, but then I incur a de/serialization overhead that is expensive. My intention was to keep the indexes as ready-to-use interprocess::maps in shared memory, to be used by all invocations of the cgi. It works, but the performance of the shared memory is poor enough that I'm not getting much increase over just doing a brute force search through the datafiles. All suggestions appreciated! Andy

4 5

[MPL] meta-function returning the arity of lambda expression
by Joel Falcou 22 Jan '09

22 Jan '09

Is there any meta-function that returns the arity of a MPL meta-function as an integral constant ? -- ___________________________________________ Joel Falcou - Assistant Professor PARALL Team - LRI - Universite Paris Sud XI Tel : (+33)1 69 15 66 35

2 2

Transferring files across a network.
by Andrew Maclean 22 Jan '09

22 Jan '09

We have a situation where there are several computers that have a folder where files are periodically deposited. Currently, in each computer, this folder is scanned periodically and the files are parsed by a program that then sends the SQL data to a central machine where there is postgresql server running. The files are generally small because they are compressed. But the uncompressing and subsequent parsing into SQL strings dramatically inflates the amount of data that is sent across the network. What I am thinking of doing is sending the all the files to a directory on the central machine then parsing them there for subsequent inclusion into the database. So I need a program to periodically scan the folder on the computer and send these files to the central machine. On the central machine I need a similar program to listen for incoming files and put them into a directory for subsequent processing. The listner on the central machine will need to process files arriving simultaneously from several machine. There is no issue with file name conflicts as the names are unique. Can anyone give me any pointers? I think I should use boost asio but I don't know where to start with it. The tutorial and examples, while good haven't provided me with an insight yet! The main requirement I have is that it should be cross-platform. I looked at cURL/cURLpp but there are immense difficulties building this in a windows environment. I would like to stick with boost if possible. Thanks Andrew -- ___________________________________________ Andrew J. P. Maclean Centre for Autonomous Systems The Rose Street Building J04 The University of Sydney 2006 NSW AUSTRALIA Ph: +61 2 9351 3283 Fax: +61 2 9351 7474 URL: http://www.acfr.usyd.edu.au/ ___________________________________________

6 11

[Program Options] Streaming of actual option values.
by Rodolfo Federico Gamarra 22 Jan '09

22 Jan '09

Hi all!! I'm using <boost/program_options.hpp> to allow options in my program to be configured using 3 flavors: - Command line values - Config-file values ( parse_config_file). - Default values ( value<double>(&myDouble)->default_value(3.3) ) prioritized in the given order and put all together in the same variables_map. Upon finalization I'd like to generate a config-file (as required by the 2nd flavor) with the actual values used in the run. So far I've come with a solution that involves iterating over the map, and performing the outstreaming by myself. Something in the line of for(variables_map::const_iterator i = vm.begin(); i != vm.end();++i) { std::cout << i->first << " " << i->second.as<std::string>() << std::endl; } What is cumbersome of this approach is that, actually, not all my args are "std::string", so in reality the call to as<>() ends with bad_any_cast in some cases: it seems that there's no operator << defined for i->second or its underlying "boost::any". So I'd have to switch-case over all the types I'm using for the values. Second, I have to format the output in accordance to parse_config_file by myself. Is there any better solution to this? Thanks a lot in advance. -- Rodolfo Federico Gamarra

1 0

[BGL] - filtered_graph EdgePredicate performance
by Michael Fawcett 22 Jan '09

22 Jan '09

filtered_graph documentation says: [1] The reason for requiring Default Constructible in the EdgePredicate and VertexPredicate types is that these predicates are stored by-value (for performance reasons) in the filter iterator adaptor, and iterators are required to be Default Constructible by the C++ Standard. http://www.boost.org/doc/libs/1_37_0/libs/graph/doc/filtered_graph.html I only noticed this after scouring the documentation for hints as to why dijkstra_shortest_paths was taking 2 hours with a filtered_graph compared to 2 seconds with my adjacency_list. The reason for this huge difference was my filter's EdgePredicate had a boost::unordered_set member that returned true if the edge was in it. Changing that member to a pointer to an unordered_set reduced running time back down to the 2 second range. As I now understand it, every time a filtered graph iterator adaptor is incremented, it actually incrememts in a loop while the EdgePredicate returns false (or end is reached). This implies that the EdgePredicate should be extremely lightweight, since it is stored by-value in the iterator. What is the general recommendation when writing filter_graph predicates? My use case is filtering out all edges that reside inside a polygonal region. To do this, I do the intersection, then store the edges that are inside that region in an unordered_set. This requires that my EdgePredicate store a pointer to the graph and a pointer to the unordered_set since I can't hash an edge_descriptor without a reference to the graph (to get the target and source). That still seems big to me, given that it will be copied per iterator increment. On top of that, to ensure proper release of the unordered_set, I need to use shared_ptr or something similar, instead of just having the predicate own the unordered_set like it logically should, introducing yet more overhead in reference counting. The footnote states that it is stored by value for performance reasons. What types of predicate implementations were in mind when that was decided? I can't see a realistic filtering predicate that could say true/false with no member variables - at the very least it must store a pointer to the graph so that it can use the edge_descriptor. Is there a different, preferred way of writing EdgePredicates? Is there a way to get target/source from an edge_descriptor without the graph instance? Thanks, --Michael Fawcett

2 1

[GIL] Raw C pointer from any_image
by Olivier Tournaire 21 Jan '09

21 Jan '09

Hi all, I am wondering if it is posible to get a raw C pointer from an any_image / any_image_view. I tried this (taken from the tutorial): typedef boost::mpl::vector<gray8_image_t, rgb8_image_t> my_img_types; any_image<my_img_types> runtime_image; jpeg_read_image(filename.toStdString().c_str(), runtime_image); unsigned char* data = interleaved_view_get_raw_data( any_image_view<my_img_types>(runtime_image) ); And got the following errors: /usr/include/boost/gil/extension/io/../dynamic_image/../../image_view_factory.hpp:95: error: 'value' is not a member of 'boost::gil::view_is_mutable< .................. > error: no matching function for call to 'interleaved_view_get_raw_data(boost::gil::any_image_view< .................. > As I am really not familiar with the MPL, I am certainly doing something bad. Hope you could help. Regards, Olivier -- Olivier Tournaire MATIS - Institut Géographique National 73, Ave de Paris 94165 St Mandé cedex, France tel: (+33) 1 43 98 80 00 - 71 25 fax: (+33) 1 43 98 85 81

3 2

Pointer serialization
by niranjan bangera 21 Jan '09

21 Jan '09

hi am facing isssue in pointer serialization: version:1.36.0 linux : g++ 4.1.1 Where in sample ex shows; header file my.h contains: ------------------------- class second { int var1; float var2; template<class Archive> void serialize(Archive & ar, const unsigned int file_version) { ar & var1; ar & var2; } }; class first { second so; template<class Archive> void serialize(Archive & ar, const unsigned int file_version) { ar & so; } }; mymain.cpp ---------- #include "my.h" first *ptr; { std::ofstream ofs("filename"); boost::archive::text_oarchive oa(ofs); oa << ptr: } here in while compiling this(my.cpp) i am getting these errors -------------------------------------------------------------- /project_new/boost_1_36_0/include/boost-1_36/boost/archive/detail/oserializer.hpp:154: undefined reference to `boost::archive::detail::archive_pointer_oserializer<boost::archive::text_oarchive>::~archive_pointer_oserializer()' /project_new/boost_1_36_0/include/boost-1_36/boost/archive/detail/oserializer.hpp:154: undefined reference to `boost::archive::detail::archive_pointer_oserializer<boost::archive::text_oarchive>::~archive_pointer_oserializer()' /project_new/boost_1_36_0/include/boost-1_36/boost/serialization/extended_type_info_typeid.hpp:76: undefined reference to `boost::serialization::detail::extended_type_info_typeid_0::extended_type_info_typeid_0()' /project_new/boost_1_36_0/include/boost-1_36/boost/serialization/extended_type_info_typeid.hpp:78: undefined reference to `boost::serialization::detail::extended_type_info_typeid_0::type_register(std::type_info const&)' /project_new/boost_1_36_0/include/boost-1_36/boost/serialization/extended_type_info_typeid.hpp:78: undefined reference to `boost::serialization::detail::extended_type_info_typeid_0::type_register(std::type_info const&)' /project_new/boost_1_36_0/include/boost-1_36/boost/serialization/extended_type_info_typeid.hpp:78: undefined reference to `boost::serialization::detail::extended_type_info_typeid_0::type_register(std::type_info const&)' /project_new/boost_1_36_0/include/boost-1_36/boost/serialization/extended_type_info_typeid.hpp:78: undefined reference to `boost::serialization::detail::extended_type_info_typeid_0::type_register(std::type_info const&)' /project_new/boost_1_36_0/include/boost-1_36/boost/serialization/extended_type_info_typeid.hpp:78: undefined reference to `boost::serialization::detail::extended_type_info_typeid_0::type_register(std::type_info const&)' ...................... /project_new/boost_1_36_0/include/boost-1_36/boost/archive/detail/oserializer.hpp:94: undefined reference to `boost::archive::detail::basic_oarchive::end_preamble()' /project_new/boost_1_36_0/include/boost-1_36/boost/archive/detail/oserializer.hpp:94: undefined reference to `boost::archive::detail::basic_oarchive::end_preamble()' /project_new/boost_1_36_0/include/boost-1_36/boost/archive/detail/oserializer.hpp:94: undefined reference to `boost::archive::detail::basic_oarchive::end_preamble()' Any suggestions will be accepted. Regards Niranjan

2 1

[Units] Can't assign equation
by Daniel Dilts 21 Jan '09

21 Jan '09

I'm using Boost.Units in a program for school to make certain that my equations are consistent (and that I didn't screw them up entering them). In the next Velocity equation, first get assigned with no problems, second causes a compiler error that I believe is saying that the constructor for quantity<velocity> is explicit. Why does first work without a conversion and second doesn't? Also, how do I get second to work? I assume this will require some type of explicit conversion, but wrapping the right-hand side with quantity<velocity>(...) doesn't work. const quantity<dimensionless> pi = 3.14159265358979323846264338327950288419716939937510582097494459230781640628620899862803482534211706798; const quantity<time> deltaTime = .1 * second; const quantity<length> particleDiameter = .001 * meter; const quantity<dimensionless> dragCoefficient = 0.5; const quantity<dimensionless> specificGravityGold = 19.32; const quantity<acceleration> gravity = 9.80665 * meters_per_second_squared; const quantity<length> initialHeight = 0 * meters; const quantity<time> initialTime = 0 * seconds; const quantity<velocity> initialVelocity = root<2>(4.0/3 * specificGravityGold * gravity * particleDiameter / dragCoefficient); const quantity<mass_density> densityWater = 1000 * kilograms_per_cubic_meter; const quantity<mass_density> densityGold = 19300 * kilograms_per_cubic_meter; const quantity<mass> particleMass = 4.0 / 3.0 * pi * pow<3>(particleDiameter / 2.0) * densityGold; quantity<velocity> nextVelocity(quantity<velocity> old) { quantity<velocity> first = (pi / 6.0 * pow<3>(particleDiameter) * gravity * (densityWater - densityGold)) * deltaTime / particleMass; quantity<velocity> second = (pi / 8.0 * densityWater * dragCoefficient * old * root<2>(pow<2>(old))) * deltaTime / particleMass; ... }

2 1