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

Header only libraries, why not all?
by Sachin Garg 02 Feb '10

02 Feb '10

Header-only libraries are great as that eliminates the need for building the libraries and then linking with them. But not all libraries in boost are header only, which is an inconvenience. Is this only because of 'legacy', and with time all libraries will move towards being header only? Or is there some technical requirement due to which some libraries just can't ever be header-only? SG

6 7

boost any questions and praise
by Alan Tennant 02 Feb '10

02 Feb '10

Firstly. Thank you very much team Boost, it looks like it will be very useful and solve many of my coding problems. I can create an any boost::any a1; I can assign something to it a1 = string1; I can get it back out later std::cout<<any_cast<std::string>(a1); A boost any has a "type()" method. However this produced an error. std::cout<<any_cast<a1.type()>(a1);

4 7

String Pools: Reusing Memory
by Bilokon, Paul 02 Feb '10

02 Feb '10

Dear all, First, may I blaspheme by invoking Java. In Java, a pool of strings, initially empty, is maintained privately by the class String. This is essentially a case of the Flyweight design pattern (http://en.wikipedia.org/wiki/Flyweight_pattern) where memory usage is minimised by sharing as much data as possible with other similar objects. Thus if we have two different instances of a string "foo" we shall reuse a single area of memory. In Java there is a concept of string internisation: http://java.sun.com/j2se/1.4.2/docs/api/java/lang/String.html#intern%28%29 My question is, what is the best way to implement this in C++? Is the answer to this a combination of std::string and boost::fast_pool_allocator, as discussed here: http://www.gamedev.net/community/forums/topic.asp?topic_id=534428 Or are there better ways? Perhaps there is something like an intern_string? What would be the best practice where memory usage is an issue? Many thanks, Paul

2 1

Re: [Boost-users] [fusion] xlC 6.0 building math lib - Errors
by Jerry Jeremiah 02 Feb '10

02 Feb '10

On 1 Feb 2010, Radha wrote: > > Hi Team, > > I am trying to compile boost math library using xlC version 6 on Aix 5.3. > > I get the following error: > "./boost/fusion/sequence/intrinsic/begin.hpp", line 54.88: 1540-0408 (S) The base class "apply<const boost::fusion::tuple<double,double,double,boost::fusion::void_,boost::fusion::void_,boost::fusion::void_,boost::fusion::void_,boost::fusion::void_,boost::fusion::void_,boost::fusion::void_>" is declared but not defined. > ... > "libs/math/build/../src/tr1/cbrt.cpp", line 16.11: 1540-0700 (I) The previous message was produced while processing "cbrt(double)". > > Is there any patch i need to apply? > > Any help would be much appreciated! > > All my search on this error didn't get me anywhere; if there is prior > post on same topic and solution please let me know! > > Thanks, > Radha On 2/1/2010 9:30 PM, John Maddock wrote: > > I've changed the subject line on this one, as this is really a problem > with Boost.Fusion rather than the Math lib as such, Joel? > > John. > On 02 Feb 2010, Joel de Guzman wrote: > > Alas, I don't have access to the platform/compiler :-( I'll need someone > to help me with this. The errors don't make much sense as it seems. > > Regards, > > Joel de Guzman > I have access to this platform. When I tried to compile the boost libraries with xlC 6 I got lots of template related errors in the math library (the first library bjam tries to build) but I didn't see his fusion errors. xlC 6 is not very standards compliant - upgrading to xlC 10 would really help if that is an option. But like I said, I can't get that far even - my attempt at compiling the trunk dies with a similar error to his "declared but not defined" error: vacpp.compile.c++ bin.v2/libs/math/build/vacpp/release/threading-multi/assoc_legendre.o "./boost/math/special_functions/gamma.hpp", line 1084.16: 1540-0130 (S) "L::g" is not declared. ... "libs/math/build/../src/tr1/assoc_legendre.cpp", line 16.28: 1540-0700 (I) The previous message was produced while processing "assoc_legendre(unsigned int, unsigned int, double)". "./boost/math/special_functions/gamma.hpp", line 176.17: 1540-0130 (S) "L::lanczos_sum" is not declared. ... "libs/math/build/../src/tr1/assoc_legendre.cpp", line 16.28: 1540-0700 (I) The previous message was produced while processing "assoc_legendre(unsigned int, unsigned int, double)".

2 2

[multi_array] deduce logical shape of "parent" multi_array from view
by alfC 02 Feb '10

02 Feb '10

Hi, I am programming and adaptor of the HDF5 library to handle, to some extend, multi_arrays in an elegant way. It turns out that when I try to save a multi_array_view, with strides and offsets, etc. HD5F still needs to know the logical size (shape) of the parent array (the one that has contiguous elements in memory). The paradox is that in order to save an array view I need to know the original array. The question is, is there an (elegant) way to reconstruct the shape of the original array just from the multi_array_view. For moments it seems imposible because array_view doesn't know even the dimensionality of the "parent" array (or it does?). The task is further complicated because many members that may have this information are actually private or protected within multi_array_view. This is an example of what I want: multi_array<2> A(extents[10][10]); multi_array_view<1> Aview = A[3][indeces[range(0,10,2)]]; desired_function(Aview); --> returns 10, 10. Thanks, Alfredo

2 2

Absolute library path references - boost 1.41
by Rui Pedro Fernandes 02 Feb '10

02 Feb '10

Hi, I'm building Boost 1.41 in HP-UX with gcc and the HP-UX linker, and I'm getting problems similar to these: http://old.nabble.com/HPUX-aCC:-Howto-avoid-building-boost-libraries-contai… Namely, the problem is that the path to the boost system lib is keep as an absolute in lib filesystem. Removing the explicit dependency of lib filesystem on lib system solves the problem, but I then have to add the system lib as an explicit dependency on projects using lib filesystem (which does not happen in other platforms). Perhaps the proposed fix should be extended to gcc in HP-UX? Also, linking project executables with libs that are part of the same project (even if you install the build) has the executables keep the static depend paths on the project libs. So, if I define exe myexe : myexe.cpp /project//liba ; I will get myexe depending statically on the path where liba was built against the root of the project. if the liba project was in ROOT/libs/liba/bin/(...), I will have myexe statically depend on ROOT/libs/liba/bin/(...)/liba.so, even if I bjam install . Any ideas? Thank you

1 0

using a condition_variable with a scoped_try_lock
by Richard Connon 02 Feb '10

02 Feb '10

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 When I try to call wait() on a boost::condition variable with a boost::scoped_try_lock as the argument I get the following error... XXXXXXX.cpp:XX: error: ‘boost::unique_lock<boost::mutex>’ is an inaccessible base of ‘boost::detail::try_lock_wrapper<boost::mutex>’ Does this mean I can't use a condition variable with a try lock? -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.14 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/ iQEcBAEBAgAGBQJLZ3f9AAoJEO6BIH4o6iWz110IAKxqpv3zNQWp8KmJZtC4JqOu 07uh94q+TWLzVgdqENEFhO9h6M64crLyEcmLgvRLM7wSMYpL1FifyGVDmLcmZ1qR +nGCeOfwJEsqQlNNlp2GmAW44jABFRgY8eb15B1QYKYHe1uhaZxnLd6NZplHOWEZ fJZylFlobrUUDvYDZKntmIzgCpAefMkcWN6GX1Dgm2W9LJGnOkkDmFO7N645eYfN 02a8DkxjixHRz2nvfAu7VRX5ECpGweGCiZ05JSOIP1ymyVQNlyIJS5a+nzsDDAgg SDpbLSaHtdBFeVEkwHKgQcyvurrtqdkTOo/oaJzwExUDzDVNh7CwHFv0oeEXSeU= =JrhX -----END PGP SIGNATURE-----

2 1

[proto] Wrapper compile error
by David A. Greene 02 Feb '10

02 Feb '10

Eric, Here's a message I've owed you for a while. I mentioned a compile error with the ADL example for expression wrappers. I've attached a testcase. Uncommenting the commented bits and compiling with g++ results in an error message which leads me to believe that somewhere in the guts of proto there's an expectation that wrappers have one template argument. Usually this gets solved by using metafunction classes. Or, more likely, I'm just doing something wrong. :) -Dave

2 1

Help with custom classes PLEASE!!
by Daniel Veneros 02 Feb '10

02 Feb '10

Hi, I am trying to create a shared memory segment where I can share a STRING between 2 process, I tried using the following code: *In the "main" process:* #include <boost/interprocess/managed_shared_memory.hpp> #include <boost/interprocess/containers/vector.hpp> #include <boost/interprocess/allocators/allocator.hpp> #include <cstdlib> #include <string> #include <iostream> using namespace boost::interprocess; typedef allocator<std::string, managed_shared_memory::segment_manager> ShMemAllocator; typedef vector<std::string, ShMemAllocator> MyVector; int main(int argc, char** argv) { struct shm_remove { shm_remove() {shared_memory_object::remove("MySharedMemory"); } ~shm_remove() {shared_memory_object::remove("MySharedMemory"); } }remover; managed_shared_memory segment(create_only, "MySharedMemory", 65536); const ShMemAllocator alloc_inst(segment.get_segment_manager()); MyVector *myVector = segment.construct<MyVector>("MyVector" )(alloc_inst); myVector->push_back("a"); int stop; //this line helps me wait until i run the other process std::cin >> stop; return (EXIT_SUCCESS); } *at the "client" process:* #include <boost/interprocess/managed_shared_memory.hpp> #include <boost/interprocess/containers/vector.hpp> #include <boost/interprocess/allocators/allocator.hpp> #include <cstdlib> #include <string> #include <iostream> using namespace boost::interprocess; typedef allocator<std::string, managed_shared_memory::segment_manager> ShMemAllocator; typedef vector<std::string, ShMemAllocator> MyVector; int main(int argc, char** argv) { managed_shared_memory segment(open_only, "MySharedMemory"); MyVector *myVector = segment.find<MyVector>("MyVector").first; std::cout << "myVector size: " << myVector->at(0) << std::endl; return (EXIT_SUCCESS); } When I run both I receive an error "Segmentation Fault, Core Dumped"... Does anyone know how to solve this? Thanks Dann

2 5

Re: [Boost-users] generate structs with all combinations of members from a given struct
by Hicham Mouline 02 Feb '10

02 Feb '10

----- Original Message ----- From: Hicham Mouline To: boost-users(a)lists.boost.org Sent: Monday, February 01, 2010 12:09 PM Subject: Re: [Boost-users] generate structs with all combinations of members from a given struct My previous runtime solution was wrong. I just fixed it now. #include <iostream> #include <list> typedef std::list< std::list<size_t> > main_list_t; void combinations(main_list_t& m, const size_t* arr, size_t arrsize) { if (arrsize==0) return; const std::list<size_t> empty_list; std::list<size_t> l( m.empty()? empty_list: m.back() ); for (size_t s=0; s<arrsize; ++s) { l.push_back( arr[s] ); m.push_back(l); l.pop_back(); combinations(m, arr+s+1, arrsize-s-1); } } int main() { const size_t arr[] = { 1, 2, 3, 4, 5 }; const size_t arrsize = sizeof(arr)/sizeof(size_t); main_list_t mm; combinations(mm, arr, arrsize); std::cout<< "size="<< mm.size() <<std::endl; for (main_list_t::const_iterator i = mm.begin(); i!=mm.end(); ++i) { const std::list<size_t>& l = *i; for (std::list<size_t>::const_iterator ii = l.begin(); ii!=l.end(); ++ii) { std::cout<< *ii << '\t'; } std::cout<<std::endl; } }

3 4