April 2009 - Boost - lists.boost.org

[move] New version with move-aware containers library
by Ion Gaztañaga 03 May '09

03 May '09

Hi to all, Sorry for the delay, but finally I've found some time to uploaded a new version of Boost.Move: Sandbox: https://svn.boost.org/trac/boost/browser/sandbox/move ZIP file with docs: http://www.drivehq.com/web/igaztanaga/move.zip Online docs: http://www.drivehq.com/web/igaztanaga/libs/move/ Changes: -> Changed folder from boost/move_semantics to boost/move. -> I've extracted all the containers from Boost.Interprocess, put them in boost/container and they only depend on official Boost SVN code (I haven't tested it, but it should be also nice with Boost 1.38). There is no documentation for these containers. -> Fixed a bug with EDG-based compilers. -> Note: I haven't applied yet the patch for non-sfinae compilers. With this changes, I think move emulation is quite portable (at least MSVC, EDG and GCC). Now that containers are independent from Boost.Interprocess, I would like to ask about the next step. If the move library is added, where should we put those containers? I would like to just drop common (vector, list, ...) Boost.Interprocess containers and just maintain these move-aware ones. If move emulation is considered mature enough, we should also plan when could we port all custom move-emulations present in several boost libraries to the "official" move library. Best, Ion

8 20

[gil] Google analytics script tags
by Daniel James 02 May '09

02 May '09

Hi, If you've missed it, there was a recent thread about using google analytics in offline copies of boost and the consensus was firmly against them. The gil documentation includes the google analytics javascript but doesn't seem to use it. So I assumed that their inclusion was a mistake and went ahead and removed them on trunk. I hope that wasn't presumptuous. But I did this on the generated documentation as I couldn't work out how it is generated. It'd useful to have some brief instructions and any necessary files. Ideally we should be able to generate them with boost build but a manual doxygen call is fine. thanks, Daniel

3 4

Intentions of Boost.TR1
by David Abrahams 02 May '09

02 May '09

I just saw this post on comp.lang.c++.moderated. I wonder if the Boost.TR1 documentation shouldn't clarify its intended use or degree of conformity. I don't think TR1 conformity is something we have been paying much attention to, is it? -- Dave Abrahams BoostPro Computing http://www.boostpro.com

5 4

German Boost book updated
by Boris Schaeling 02 May '09

02 May '09

For those who know German: I've added five more chapters to my Boost book at http://www.highscore.de/cpp/boost/. There are now altogether 16 chapters covering 29 Boost libraries. The book does not try to replace the documentation but is written in a tutorial style to give developers a kick start. It explains what the numerous libraries are good for and how they are used without going too much into detail. Any feedback is, as always, appreciated! Boris

2 2

Boostcon Boost++0x Session
by Michael Caisse 02 May '09

02 May '09

Dave - I am sorting out which sessions to attend and wish there was more of me. For the Boost++0x sessions reading a tutorial is a prerequisite. Is the tutorial on the boostcon site and I just failed to find it? Thank you- michael -- ---------------------------------- Michael Caisse Object Modeling Designs www.objectmodelingdesigns.com

2 3

how to build boost version 1_34_1 with gcc in non-standard location?
by Oliver Kullmann 01 May '09

01 May '09

Hello, I'm wrestling here for a long time with the following problem, and I would be very glad if somebody could help me here (seems to be straight-forward, but I can't find anything anywhere on the subject): I have to build some C++ code which doesn't compile with the system gcc version 4.3.1. So I have installed gcc version 4.1.2 locally. Now that C++ code uses Boost version 1_34_1, and doesn't compile with higher versions. Alright, so I install 1_34_1; very unfortunately, linking is necessary, and here is the problem: Obviously I have to build Boost 1_34_1 with the locally build gcc 4.1.2, and now how can I tell the Boost-build where to find gcc??? What I found is to use PACKAGEPATH/bjam/bjam --toolset=gcc-4.1.2 --toolset-root=PATH_TO_GCC_DIR --prefix=INSTALLATION_DIR --build-dir=BUILD_DIR "-sGCC_ROOT_DIRECTORY=PATH_TO_GCC_DIR" install --without-python However, as I had to found out after some time, this doesn't have any effect --- Boost just builds with the system compiler, and ignores all these specifications of PATH_TO_GCC_DIR. Now what is the right way to do it?? I would be really glad for help. Oliver

2 2

[Graph] transitive reduction
by boese-wolf＠math.uni-frankfurt.de 30 Apr '09

30 Apr '09

Hi! May I propose the following implementation for a transitive reduction to be added to the boost graph library? I'm a novice and the quality of the code I produced surely does not meet boosts standards, but maybe with some help, I'm able to improve it. So, please don't shoot me. #ifndef MY_OWN_TRANSITIVE_REDUCTION_HPP_INCLUDED #define MY_OWN_TRANSITIVE_REDUCTION_HPP_INCLUDED #include <vector> #include <algorithm> //std::find #include <boost/concept/requires.hpp> #include <boost/concept_check.hpp> #include <boost/graph/topological_sort.hpp> #include <boost/graph/graph_traits.hpp> #include <boost/property_map.hpp> //I actually didn't care which of those heades include some of the //others and whicht could be left out, for this reason, I simply added a //header when I used a function, described to be in this header by the //boost graph documentation //also I didn't got all of the concepts thin. Am I suppose to check //for all concepts, which are needed for functions I call? (As if I //wouldn't do that, the users would see the functions called by //complaining about missings concepts, which would be clearly an error //message revealing internal implementation and should therefore be avoided?) //the pseudocode which I followed implementing this algorithmn was taken //from the german book Algorithmische Graphentheorie by Volker Turau //it is proposed to be of O(n + nm_red ) where n is the number //of vertices and m_red is the number of edges in the transitive //reduction, but I think my implementation spoiled this up at some point //indicated below using namespace boost; template <typename Graph, typename GraphTR, typename G_to_TR_VertexMap, typename VertexIndexMap> BOOST_CONCEPT_REQUIRES( ((VertexListGraphConcept< Graph >)) ((IncidenceGraphConcept< Graph >)) ((MutableGraphConcept< GraphTR >)) ((ReadablePropertyMapConcept< VertexIndexMap, typename graph_traits<Graph>::vertex_descriptor >)) ((Integer< typename property_traits< VertexIndexMap >::value_type >)) ((LvaluePropertyMapConcept< G_to_TR_VertexMap, typename graph_traits<Graph>::vertex_descriptor >)), (void)) transitive_reduction( const Graph& g, GraphTR& tr, G_to_TR_VertexMap g_to_tr_map, const VertexIndexMap & g_index_map ) { typedef typename graph_traits<Graph>::vertex_descriptor Vertex; typedef typename graph_traits<Graph>::vertex_iterator VertexIterator; typedef typename std::vector<Vertex>::size_type size_type; std::vector<Vertex> topo_order; topological_sort( g, std::back_inserter( topo_order ) ); std::vector<size_type> topo_number_storage( num_vertices(g) ); iterator_property_map<size_type*, VertexIndexMap, size_type, size_type&> topo_number( &topo_number_storage[0], g_index_map ); { typename std::vector<Vertex>::reverse_iterator it = topo_order.rbegin(); size_type n = 0; for(; it != topo_order.rend(); ++it,++n ) { topo_number[ *it ] = n; } } std::vector< std::vector< bool > > edge_in_closure(num_vertices(g), std::vector<bool>( num_vertices(g), false)); { typename std::vector<Vertex>::reverse_iterator it = topo_order.rbegin(); for( ; it != topo_order.rend(); ++it ) { g_to_tr_map[*it] = add_vertex(tr); } } for( typename std::vector<Vertex>::iterator it = topo_order.begin(); it != topo_order.end(); ++it ) { size_type i = topo_number[ *it ]; edge_in_closure[i][i] = true; std::vector<Vertex> neighbors; //I have to collect the successors of *it and traverse them in //ascending topological order. I didn't know a better way, how to //do that. So what I'm doint is, collection the successors of *it here { typename Graph::out_edge_iterator oi,oi_end; for( tie(oi, oi_end) = out_edges( *it, g ); oi != oi_end; ++oi ) { neighbors.push_back( target( *oi, g ) ); } } { //and run through all vertices in topological order for( typename std::vector<Vertex>::reverse_iterator rit = topo_order.rbegin(); rit != topo_order.rend(); ++rit ) { //looking if they are successors of *it if( std::find( neighbors.begin(), neighbors.end(), *rit) != neighbors.end() ) { size_type j = topo_number[ *rit ]; if( not edge_in_closure[i][j] ) { for(size_type k = j; k < num_vertices(g); ++k) if( not edge_in_closure[i][k] ) //here we need edge_in_closure to be in topological order, edge_in_closure[i][k] = edge_in_closure[j][k]; //therefore we only access edge_in_closure only through //topo_number property_map add_edge(g_to_tr_map[*it], g_to_tr_map[*rit], tr); } //if ( not edge_in_ } //if (find ( } //for( typename vector<Vertex>::reverse_iterator } // { } //for( typename vector<Vertex>::iterator } //void transitive_reduction #endif //#ifndef MY_OWN_TRANSITIVE_REDUCTION_HPP_INCLUDED Best regards, Eric

2 3

[ANN] GCC 4.4.0 Released
by Eric Niebler 30 Apr '09

30 Apr '09

This might interest some here: http://gcc.gnu.org/ml/gcc/2009-04/msg00590.html It'd be nice to get some test coverage. Any takers? -- Eric Niebler BoostPro Computing http://www.boostpro.com

5 6

[multi-index] Changing hash_index_node.hpp to double linked list
by brad higgins 29 Apr '09

29 Apr '09

I am using boost multi-index, with a hashed-non-unique lookup. In my data, I am expecting many hash collisions. When I erase the multi- index, I find very poor performance. On investigation, I found that the hash collision data structure is a single linked list, and that removing an element involves traversing the entire list. Removing all of the elements involves traversing the list many times. Since my data has many collisions, I am traversing a very long list very many times. I know that the hash table is not designed for use with many collisions, but it is still faster in my testing on insert and serialization than the ordered_non_unique index specifier. I may end up using ordered_non_unique to avoid the worst-case erase() performance, and it may be good enough for the insert() performance. However, I would like to investigate making the hash collisions double- linked. I tried editing boost/multi_index/detail/hash_index_node.hpp, to make it a double-linked list, but I am seeing one of the included tests fail (test_safe_mode). I can't spot any bugs in my hash_index_node.hpp changes. What, apart from hash_index_node.hpp, needs to be updated, to support a double- linked list? boost/multi_index/detail/bucket_array.hpp? Others? Thanks, Brad

3 3

[encrypted strings]
by Sid Sacek 29 Apr '09

29 Apr '09

Does boost have any compile-time classes for string encryption? Is it even possible? When a hacker dumps an executable, they can see all of the strings the program might use, and some of those strings may contain sensitive information. Does boost have any classes that can encode the strings at compile-time? Ideally, the third string in the code below would never compile the "secret" string into the final binary. Regards, -Sid void foo( void ) { char const * s1 = "abc"; string s2 = "xyz"; encrypted_string s3( "secret" ); }

11 28