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

Re: [Boost-users] [Boost.Test] valgrind complains about invalid reads
by Scott Banachowski 17 Dec '09

17 Dec '09

I agree with Peter here, the destructor for test_unit should be virtual. I find that it is being deleted when test_unit* is a stored in reference counted pointer, and at program exit it is being deleted by calling ~test_unit(). If you use a derived class (test_suite), this is leading to an error. Also, in the code I'm looking at (1.40.0), the test_suite class has a virtual destructor, but its base class test_unit does not. Not making the base class virtual seems to be defeating the purpose. Scott

2 1

Interfaces in C++
by Jerry Jeremiah 16 Dec '09

16 Dec '09

Hello, I am really interested in interfaces ever since Google's new Go language brought them to my attention. Who would have guessed they are a really old thing? Joel posted some code based on fusion that does interfaces at http://lists.boost.org/Archives/boost/2005/06/89418.php but it is based on a version of fusion before it became a separate library. I have fiddled with it trying to get it to compile against the trunk but I guess I just don't understand enough about fusion to get it going. :-( If Joel or some other fusion guru could tell me how to change it to get it to compile with the current version of fusion I would really appreciate it. Besides getting the ability to play with interfaces hopefully I will gain more in depth knowledge of how to use fusion... Jerry

1 0

boost graph library, default_dfs_visitor, dfs
by List User 16 Dec '09

16 Dec '09

Hi Please see the code below. class my_visitor: public default_dfs_visitor { public: my_visitor(){} void finish_vertex(Vertex u, const Graph & g); vector<Vertex> getFinishingOrder(void){return finishingOrder;} //private: vector<Vertex> finishingOrder;//stores the vertices in the finishing order }; template<typename Vertex, typename Graph> void my_visitor::finish_vertex(Vertex u, const Graph & g){ finishingOrder.push_back(u); } main.cpp has the following: my_visitor vis; depth_first_search(g, visitor(vis));//to get the finishing order on vertices cout << "size of fin vector = " << vis.finishingOrder.size() << endl; My problem: size of the finishingOrder vector is zero outside the my_visitor::finish_vertex method. I checked its size inside the my_visitor::finish_vertex method and its value is increasing... Does I miss something in the understanding of the visitor concept? Thanks suresh

4 3

[GIL] Change layout type of a pixel...
by Eloi Du Bois 16 Dec '09

16 Dec '09

Hi all, I would like to know how to change the layout of a given pixel type. For example, I would like to do something like: // Pixel = rgb16_pixel_t for example. template<class Pixel> void foo(Pixel & A) { typedef typename pixel<hsl_t, Pixel::bits>::type hsl_pixel_t; hsl_pixel_t a; color_convert(A, a); } Thank you by advance, Eloi.

2 4

[GIL]
by Eloi Du Bois 16 Dec '09

16 Dec '09

Hi all, Sorry to bother you with my problems but I am experiencing a new probable bug: I am trying to manage with data that are set as follow: P;P;P;P;...... Where P is a rgb pixel (10 bits per components) packed on an uint32 as follow: 00RGB;00RGB;00RGB;00RGB........ I mean there is two bits unused on each 32 bits pixel. So I tried to set up a view using packed_pixel and a bug occurs when using copy_and_convert_pixels... The very strange thing is that it is working when using color_convert on two pixels... I think there may exist something I didn't understood... Anyway, I give you the buggy code: /// 10 bits rgb packed to uint32_t starting at the 2nd bit typedef const packed_channel_reference<uint32_t, 22, 10, true> rgb10_packed_channel0_t; typedef const packed_channel_reference<uint32_t, 12, 10, true> rgb10_packed_channel1_t; typedef const packed_channel_reference<uint32_t, 02, 10, true> rgb10_packed_channel2_t; typedef mpl::vector3<rgb10_packed_channel0_t, rgb10_packed_channel1_t, rgb10_packed_channel2_t> rgb10_packed_channels_t; typedef packed_pixel<uint32_t, rgb10_packed_channels_t, rgb_layout_t> rgb10_packed_pixel_t; typedef view_type_from_pixel<rgb10_packed_pixel_t>::type rgb10_packed_view_t; // image data of dimension: {w,h} = {1,0} uint32_t *data=new uint32_t[1]; data[0] = 0xFFFFF00C; rgb10_packed_view_t vw10 = interleaved_view( 1, 0, (rgb10_packed_pixel_t*)( data ), sizeof( uint32_t ) ); rgb10_packed_pixel_t pix10 = *vw10.row_begin(0); rgb16_pixel_t pix16; // This is working color_convert(pix10, pix16); std::cout << "Right {R,G,B}={" << get_color(pix16, red_t()) << "," << get_color(pix16, green_t()) << "," << get_color(pix16, blue_t()) << "}" << std::endl; // Assert that are ok assert( get_color( pix16, red_t() ) == 0xFFFF ); assert( get_color( pix16, green_t() ) == 0xFFFF ); assert( get_color( pix16, blue_t() ) == 0x00C0 ); rgb16_image_t img16(vw10.width(), vw10.height()); rgb16_view_t vw16(view(img16)); // This isn't working with packed_view copy_and_convert_pixels(vw10, vw16); pix16 = *vw16.row_begin(0); std::cout << "Wrong {R,G,B}={" << get_color(pix16, red_t()) << "," << get_color(pix16, green_t()) << "," << get_color(pix16, blue_t()) << "}" << std::endl; // Assert that doesn't pass assert( get_color( pix16, red_t() ) == 0xFFFF ); assert( get_color( pix16, green_t() ) == 0xFFFF ); assert( get_color( pix16, blue_t() ) == 0x00C0 ); Hope this will help, Eloi.

3 5

boost.plot: integrating plotting in a C++ native application: general questions about which tools to use
by Hicham Mouline 16 Dec '09

16 Dec '09

Hello, I am writing an application with a native C++ computation engine (library) that generates plotable data 2d and 3d, and a plotting engine. I am trying to determine what tools/libraries to use for the plotting engine. My graphs are all time series in the x axis (typically dates but maybe day numbers). I have a y axis and sometimes a z axis as well. Data is available in std::vector/list< pair<> > or std::vector/list< boost::tuple<> >, sometimes std::map<k,v>, perhaps later boost::multi_array<2> or <3> as well, or in fact just native arrays of pairs or boost::tuples. Decisions are made by the users, tweaking parameters and seeing how the plots change. The application user interface will be as: 1. win command line and just the plotting being graphical 2. Linux command line and just the plotting being graphical 3. win .net interface - input of parameters to the computing engine, and the plotting being graphical 4. Linux with X (maybe gtk/qt) - input of parameters to the computing engine, and the plotting being graphical And the results of the computation will then need to be plotted. So far, I have seen these solutions: ***1 - With 1,2,3 and 4, I can see a write the numbers to a file, call the external process on windows that runs gnuplot that runs a script file to plot the file. Both 2D (y(x)) and 3D(z(x,y)) visualizations are pretty flexible, in particular 3D visualizations allow to rotate the graph and view from different angles. But this process seems tortuous, gnuplot accepts piped data on linux but not on windows without mingw/cigwin. It would be preferable to not generate intermediary files, especially if that is very large. There is also a gnuplot c++ interface, but that doesn't avoid running the gnuplot external process. ***2 - I've come across boost.plot and it generates svg files which have only 2D if I understood correctly, SVG doesn't express 3D, and then use a browser (chrome) or that tool that uses SVG as its native format, or use ImageDisplay for e.g.( this comes with c++ dev library, perhaps that helps to plot directly the svg without running the process externally ) Cons: no 3d (rotating and different angles) and still require external programs. ***3 - I don't know .net but it seems c# is more convenient for the GUI part (inputting parameters, selecting options), running the c++ computation engine, and .net probably has some plotting part (don't know about 3D plots and rotating though) Cons: non portable solution - how would c# plotting interact with native c++ stl? ***4 - gtk/qt : does it have plotting features? It is ported to linux and win, does the windows port requires cygwin/mingw? How does it interact with STL containers? The solution is ideally the quickest-to-deploy and most portable, with windows not requiring cygwin/mingw, and not needing to run an external process. The need for portability comes from hardware requirements machines with a large number of processor cores run linux, while smaller boxes run win. I appreciate there are a lot of questions here, that deal with many topics and I thank everyone for answers or suggestions, Regards,

5 6

[boost-users][asio] async operations order.
by Germán Diago 16 Dec '09

16 Dec '09

Hello. I'm implementing a fully asynchronous server and I would like to know the order of async operations. The server receives, first of all, a size, and later the data. The onSizeRead handler allocates the memory required to store the data that is coming after the async_read. So I would like to know if this buffer will be available when onDataRead is called. so I get the size and data with two async_read operations. I would like to know if the relative order of the async calls is respected, since I'm reading from the same socket for both operations and I need the resources dictated by the first read before using the second one.

2 1

Compiler crash with Phoenix and PropertyTree.
by Surya Kiran Gullapalli 16 Dec '09

16 Dec '09

Hello all, I'm seeing a compiler crash (MSVC 8.0) with the following piece of code. I'm using Boost.Phoenix along with Boost.PropertyTree namespace pt = boost::property_tree ; pt::ptree tree ; boost::phoenix::bind (&pt::ptree::put<string, bool>, tree, string("Temp.Flag"), true) ; Attached is the build log. Actually, this is a test program, I wrote to reproduce a C4180 warning, but I see a crash here (actually I see C4180 also) Let me know if I'm doing something wrong. Thanks, Surya

4 7

is visitor object destroyed immediately after the execution of depth_first_search?
by Suresh Kumar 16 Dec '09

16 Dec '09

Hi, my idea is to get the finishing order of vertices in dfs by pushing the vertices in a vector and return it later through a function return value from my visitor class object. //code snippet int main() { my_visitor vis; depth_first_search(g, visitor(vis)); cout << "size of fin vector = " << vis.finishingOrder.size() << endl; //i get zero here!!! } class my_visitor: public default_dfs_visitor { public: my_visitor(){} void finish_vertex(Vertex u, const Graph & g){finishingOrder.push_back(u);}; vector<Vertex> getFinishingOrder(void){return finishingOrder;} private: vector<Vertex> finishingOrder; }; My problem: size of the finishingOrder vector is zero outside the my_visitor::finish_vertex method. I checked its size inside the my_visitor::finish_vertex method and its value is increasing every time a vertex is finished. Does I miss something in the understanding of the visitor concept? Thanks suresh -- R. Suresh Kumar PhD Student, Vislab #216, Engineering II Bldg University of California, Riverside, CA 92521

2 2

[phoenix] A simple lambda func as boost::function
by Richard Vock 16 Dec '09

16 Dec '09

Hi boost gurus, after some time and a lot of pain I somehow managed to work with boost.lambda in a way that actually outweighs the costs (code elegance versus time spent compiling it), but I miss some features (like local variables) I see in that cute phoenix documentation and a lot of mailing list threads which all seem to propose using phoenix rather than lambda (and it indeed looks a lot better). But whenever I try to use it, I somehow fail at the most simple tasks. Take for example this minimalistic code example here (note the use of #define USE_LAMBDA): http://pastebin.com/m3f590a01 Both versions (with and without USE_LAMBDA defined) compile well on the most recent stable g++ compiler, but the phoenix version does not output anything. Not at all. Also note, that this exact lambda expression is mentioned here: http://tinyurl.com/ybmp2lw (well, without the << "\n" but that doesn't change the problem) Could someone please enlighten me by telling me what I have to do here? I should mention I am bound to boost::function, not some funky phoenix::actor derivations... Greets, Richard Vock

4 9