December 2010 - Boost-users - lists.preview.boost.org

Sparse Matrix Serialization
by Luke Bloy 24 Dec '10

24 Dec '10

I'm interested in serializing some boost compressed row matrices. Based on this mailing list thread from 2008 http://lists.boost.org/MailArchives/ublas/2008/08/2974.php it seemed that dense matrix support has been implemented so I had hoped that sparse matrix support was there as well. However when i attempt to compile the attached code I get a number linking errors of the type BoostTest3.cxx:(.text+0xf1e): undefined reference to `boost::serialization::typeid_system::extended_type_info_typeid_0::extended_type_info_typeid_0(char const*)' BoostTest3.cxx:(.text+0xf38): undefined reference to `boost::serialization::typeid_system::extended_type_info_typeid_0::type_register(std::type_info const&)' BoostTest3.cxx:(.text+0xf42): undefined reference to `boost::serialization::extended_type_info::key_register() const' BoostTest3.cxx:(.text+0x1000): undefined reference to `boost::serialization::typeid_system::extended_type_info_typeid_0::~extended_type_info_typeid_0()' The only boost library i'm attempting to link to is -lboost_serialization-mt Am I missing something or has this functionality not been implemented? If it isn't implemented can someone propose a work around? Could i serialize the unbounded arrays that make up the core of the sparse matrix representation on there own? Thanks Luke

2 1

filesystem v2 / v3 and symlilnks
by Jayesh Chaudhary 24 Dec '10

24 Dec '10

I was looking for some documentation but could not find any. Can boost filesystem resolve symlinks? Thanks, Aashish

2 2

[boost.filesystem3] directory name from a directory iterator
by Sean Farrow 23 Dec '10

23 Dec '10

Hi: I have the following code to iterate through sub directories: for (boost::filesystem::directory_iterator end, dir(DirectoryToIterate); dir != end; ++dir ) { if (boost::filesystem::is_directory(dir->status())) TempVec.push_back(boost::filesystem::path(dir->path()).generic_wstring()); } I'm getting the following output which I expect: C:/Documents and Settings/All Users/Application Data/Freedom Scientific/JAWS/12.0/Settings/enu I have two questions: Firstly how do I just obtain the last directory in this case enu (with out any slashes)? And secondly, is there a method of the path object that gives me the path in windows form (i.e with the \ character replacing the / character.) Any help appreciated. Cheers Sean.

2 1

Re: [Boost-users] [GIL] gray pixel iterator in generic algorithm
by Robert Lupton the Good 23 Dec '10

23 Dec '10

Dear Lubomir, This must be quite common (I have the same problem, wanting unscaled values, in astronomical images). It was quite a lot of work to implement all the types I needed (short, int, float, double) as unscaled types that properly convert (i.e. don't, 1234 -> 1234.0). I can share the code if you like, but I wrote it long ago when I was learning both gil and C++. Should this be easy? It would be really nice if gil supported unscaled types out of the box. R > I'm performing quantitative analysis with a physically significant > gray scale. So I don't expect to do any color conversion. > > On Dec 22, 2010, at 2:23 PM, Lubomir Bourdev wrote: > >> Hi Curtis, >> >> GIL's default float type is not "float" but a special float wrapper >> to ensure that the range is 0..1 >> The built-in float has a range from -infinity to infinity and that >> does not work well when converting a value from another channel type. >> If you want an image with a standard float you can create it like >> this (untested) >> >> typedef pixel<float,gray_layout_t> gray_float_pixel_t; >> typedef image<gray_float_pixel_t,false> gray_float_image_t; >> typedef gray_float_image_t::view_t gray_float_view_t;

1 0

[Build] First Boost build experience
by Robert Jones 23 Dec '10

23 Dec '10

I have been using Boost at work for quite a while, so I'm no stranger to Boost, or this list, and need no convincing about Boost's general high quality and utility. However, I've recently installed Boost on my home PC, and found the experience surprisingly difficult, and to be honest found that the "Getting Started" instructions are a wee bit poor. I'm installing on Fedora 13, 32 bit, on fairly old hardware, with Gnu C++ 4.5.1, which I've also just built from source for the first time, so maybe there's something I've missed from that process which is having some impact on my Boost experience. Here's a few observations... * First time round I figured that the right place to unpack the zip archive was to the place I ultimately wanted the include files to be (say /usr/local/include), because it's only when you get to the build binaries stuff that the whole copy/build thing is mentioned. Ok, so I should have read all the instructions first, but maybe this could be made clearer earlier. * Having come finally to step 5, the whole 'path/to/installation' thing appears, so I backtracked and put my zip expansion somewhere else. I'm keen to set this up so that I can support several boost versions, so am using the --prefix argument to bootstrap. I think I also need to supply this argument to bjam, at least it seemed to work that way. * That seemd to work ok, although there were some failures for things like MPI and zip, which are, I now realise, ok, but disconcerting. Maybe the instructions could say a few words about the log output expected. I know this is mentioned in the Custom Build stuff, but I'm following the easy idiot's guide. (Appropriate you may think, and you may very well be right!) * Then I came to link the example regex program... what is all that stuff about library naming? I didn't seem to have libraries with names like that at all, my regex libraries were libboost_regex.{a,so,so.1.45.0}, so I used that, and I still don't know what all the stage stuff is about - it seemed to work without it, but maybe there's something that's going to bite me. * Finally my link was failing with "`GLIBCXX_3.4.14' not found", which Googling seemed to suggest was because I need Fedora 14 libstdc++. So I installed that, and now it seems to work.. so far! Ok, so I got there, as I'm sure most people do, but it's not exactly a shrink-wrapped install experience, and a good deal harder than installing on Windoze. I wonder how many people are put off by this. Hmmm... reading that back it sounds like a lot of minor whinging, which is a shame 'cos I think Boost is 'triffic. Thx - Rob.

2 1

Quickbook: source-mode teletype highlighting as c++
by Wilhelm 22 Dec '10

22 Dec '10

Hi, experimenting with source-mode teletype seems to have no effect: all imports are highlighted as c++. -- Wilhelm

3 2

Quickbook: source-mode for java
by Wilhelm 22 Dec '10

22 Dec '10

Hi, are there efforts to add a java source-mode to quickbook? -- Wilhelm

2 1

[gil] New format guidelines about custom Device
by Mateusz Loskot 22 Dec '10

22 Dec '10

Hi, I'm implementing new image format support for Boost.GIL based on the new IO implementation which has been reviewed [1] lately. [1] http://lists.boost.org/boost-announce/2010/11/0273.php What would be the guidelines regarding custom Device implementation? Currently, only TIFF format implements it. When it is required to implement custom Device for the IO? When recommended as a good practice? I think it would be good to have such details documented. Best regards, -- Mateusz Loskot, http://mateusz.loskot.net Charter Member of OSGeo, http://osgeo.org Member of ACCU, http://accu.org

2 3

[interprocess] Unmap resource of the object constructed from managed_mapped_file
by Ryan47＠cgcg.com.tw 22 Dec '10

22 Dec '10

Hi, I have a managed_mapped_file and use it to dynamically construct lots of interprocess::vector<size_t> objects(all the objects and elements are resided in the mapped file). Now I'm looking for a way to keep the run-time memory consumption below a certain amount. Is it possible to unmap the resource of the vector object? Although all the resources would be freed when the managed_mapped_file is destroyed, but the cost of highly frequently mapping/un-mapping seems too expensive for my purpose. Is there any way to workaround this? Any suggestion is welcome. Thanks in advance. Regards, Ryan

2 1

Re: [Boost-users] pthread_exit() from boost::thread segfaults
by Bogdan Cristea 22 Dec '10

22 Dec '10

There is no reason for your snippet to fail. I have tested on openSUSE 11.3, 64 bits with gcc 4.5.0 and libboost_thread 1.42. Try to remove from the thread starter function the call to pthread_exit() and see if you still get the error.

3 4