- Boost-users - lists.boost.org

Iostreams - GZip problem
by Jabin Reinhold 02 Apr '07

02 Apr '07

Im trying to use the iostreams gzip compression on xml serializations produced using boost::serialization and Im having a bit of trouble with it. The variable to be serialized is savedClass. As shown below, I create the serialization, save to a file with gzip_compressor, then read it from the file with gzip_decompressor. My issue is that at the end of the code, test1.xml(~10k) and test2.xml(~12k) are not identical. Test2.xml has some extraneous characters added to the end (~2k worth). I assume these are leftovers from a memcopy somewhere, but Im not sure why. I have the same effect when I use zlib_compressor as well. What is wrong with my usage of the library(s)? //////////////////////////////////////////////////////////////// std::strstream sStream; if(true) { //Scope used to ensure that the archive is fully processed boost::archive::xml_oarchive oArchive(sStream); oArchive << BOOST_SERIALIZATION_NVP(savedClass); } std::ofstream oStreamXML1( "c:\\test1.xml", std::ios::out ); oStreamXML1 << sStream.str(); if(true) { //Scope used to ensure that the archive is fully processed std::ofstream file("c:\\test.gz", std::ios_base::out | std::ios_base::binary); boost::iostreams::filtering_streambuf<boost::iostreams::output> outStream; outStream.push(boost::iostreams::gzip_compressor()); outStream.push(file); boost::iostreams::copy(sStream, outStream); } std::strstream sStream2; if(true) { //Scope used to ensure that the archive is fully processed std::ifstream file2("c:\\test.gz", std::ios_base::in | std::ios_base::binary); boost::iostreams::filtering_streambuf<boost::iostreams::input> inStream; inStream.push(boost::iostreams::gzip_decompressor()); inStream.push(file2); boost::iostreams::copy(inStream, sStream2); } std::ofstream oStreamXML2( "c:\\test2.xml", std::ios::out ); oStreamXML2 << sStream2.str(); ////////////////////////////////////////////////////////////////

2 2

Intrusive library review results: library accepted
by "JOAQUIN LOPEZ MU?Z" 02 Apr '07

02 Apr '07

I'm pleased to announce that Ion Gaztañaga's Intrusive library has been accepted into Boost. Prior to its inclusion in Boost, the author is encouraged to address several issues raised during the review, a summary of which is given below. A total of 14 full reviews were received, all of them positive. Additionally, other members did not submit full reviews but expressed their support for the library. One Boost member raised some concerns about the usefulness of the library; I hope these concerns have been satisfactorily addressed by those review participants who explained their current and potential usage scenarios for Boost.Intrusive. This has shown that more movitational material need to be provided in the documentation introduction and in the form of full-fledged examples. Thank you to Ion Gaztañaga for all the hard work involved in creating a Boost library, to Olaf Krzikalla, who wrote the first versions of this piece of work, and to all the participants in the review. The list of issues to be addressed follows. Feel free to comment if you detect some error or omission from my part. List of issues to be addressed prior to inclusion of the Intrusive library into Boost: * Documentation - Provide a motivating introduction about usefulness of Boost.Intrusive. - Add some hello world motivating examples. - Class names linked to their section in the reference (except in code due to Quickbook limitations). - Document iterator and const_iterator as part of the container interfaces. - Add overview of iterator validity rules. - Document usage before concepts. - Improve concepts part. - Provide container concepts à la STL. - Add performance section. - Add quick guide to docs. - Make the tests more comprehensive. - Add examples section. - State clearly the purpose of the lib in the introduction. - Add more examples using the different hooks. - State in docs the possibility that tags be incomplete types. - State in docs the fact that Boost.Intrusive does not support shared_ptr's. - Document the fact that member hooks rely on non-standard implementation techniques. * Naming - Change Destroyer to a more suitable name (e.g. Sink or Disposer), rename member functions accordingly. - Rename current() as iterator_to(). - Rename [s]list_algorithms as circular_[s]list_algorithms. - Drop the 'i' prefix from the names of containers. - Rename linked() as is_linked(). - Rename assign_and_destroy() as destroy_and_assign(). * Interface improvements - Study ways to simplify the usage syntax of hooks. - Add a default value to Tag. - Member hooks shouldn't be parameterized on the host class type. - Make hooks assignment operator a no-op. - bucket_size() y bucket() should be const. - Provide cbegin(), cend() etc. - Add *_fwd.hpp headers. - Fuse autolink and regular/safe hooks into one class whose second parameter is a category class. * Implementation improvements - Iterators to be implemented with iterator_facade. - Use Boost.Concept for concept checking if compilation times are not severely affected. - Extend bucket primes table to 64 bits. - Use template trick to avoid direct use std::size_t in template parms of containers, thus pulling the std namespace in via ADL. - Use Watanabe's version of swap_nodes() for ilist. - Use insert_after() in clone_from(). - Reimplement clone_from() to not delete partially constructed data on exception. - Investigate bug with member hooks and virtual inheritance. * Extensions and issues to be considered in the future - Order-statistics tree. - Study other libraries for ideas: DOL, PTL, jjLib. - Study feasibility of implementing a "destroy multihooked element when completely unlinked" mode. - Investigate use of two-way pointers with Boost.Intrusive. - Consider a shift() member function for lists and slists. - Additional slist-like container with a pointer-to-last member to make swapping constant time. The Review Manager, Joaquín M López Muñoz Telefónica, Investigación y Desarrollo

2 1

Typetraits for decomposing member function pointers?
by Tom Bachmann 02 Apr '07

02 Apr '07

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hi, is there a way to get more information about a member function pointer, using type traits, than is_member_function_pointer<T>::value? I'm looking for information about e.g. the constness and function_traits-like information. - -- - -ness- -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.6 (GNU/Linux) iD8DBQFGBs++vD/ijq9JWhsRApJNAJ9nbbZnRMnTMBi0l09yqBaCIiYGmgCfTjRO I12viJTyF3sUp0sICPR/nDg= =ym+h -----END PGP SIGNATURE-----

2 1

[mpl] find_all type vectors in a type matrix that match a predicate.
by Christian Henning 02 Apr '07

02 Apr '07

Hi there, is it possible to find all type vectors in a type matrix ( 2D type vector ) that match a predicate? In the end I would like to have them in another type matrix. Can somebody point out how I would put together such a metafunction? I'm not sure if the mpl lib supports my feature request. Consider the following code: #include "boost/mpl/vector.hpp" #include <boost/mpl/find_if.hpp> #include <boost/mpl/at.hpp> using namespace std; using namespace boost; namespace fields { struct id {}; struct name {}; struct address {}; struct birthday {}; } namespace tags { struct primary {}; struct foreign {}; struct no_key {}; } typedef mpl::vector<fields::id , tags::primary> id_field_t; typedef mpl::vector<fields::name , tags::foreign> name_field_t; typedef mpl::vector<fields::address , tags::foreign> address_field_t; typedef mpl::vector<fields::birthday, tags::no_key> birthday_field_t; typedef mpl::vector< id_field_t , name_field_t , address_field_t , birthday_field_t > table; // aggregate all foreign keys in the following vector typedef mpl::vector0<> foreign_keys; /// ??? metafunction for aggregating all foreign keys ??? int main() { return 0; } BTW, I'm not particular interested in a certain order of the aggregated type vectors. Reagrds, Christian

3 7

newbie's questions about multi_index
by Guillaume Lazzara 02 Apr '07

02 Apr '07

Hi, I'm currently using multi_index and I have few questions: - I have a multi_index with composed keys, sometimes I need to get a range of items, iterate over them and modify their key. How can I do this properly? sample code: EquivSet::iterator tmp; std::pair<EquivSet::iterator, EquivSet::iterator> es = contains.equal_range(*it); for (EquivSet::iterator it_equiv = es.first; it_equiv != es.second;) if (!X[it_equiv->id]) { tmp = it_equiv++; contains.modify_key(tmp, update_equiv(nbEns)); } else it_equiv++; In this example, the second iterator in the pair, normaly used as end iterator, can be changed and, sometimes, the loop iterates beyond the latter. - If I can choose between a std::set and a multi_index used as a set, which one should I use? Thank you very much for your answers! -- Guillaume Lazzara Epita CSI 2008

2 2

[fusion] error in example code
by Christian Henning 02 Apr '07

02 Apr '07

Hi there, the example code of the list sequence doesn't compile. list<int, float> l(12, 5.5f); std::cout << at<0>(l) << std::endl; std::cout << at<1>(l) << std::endl; It should be: list<int, float> l(12, 5.5f); std::cout << at<mpl::int_<0> >(l) << std::endl; std::cout << at<mpl::int_<1> >(l) << std::endl; I'm using MSVC 7.1. Regards, Christian

2 1

Which is preferred? Static assertions with type traits or concept_check?
by Benjamin Lau 02 Apr '07

02 Apr '07

Hi all, I'm writing a template class and I have several requirements on the type parameters. A lot of it has to do with the inter-convertibility of the types. I have 2 options to choose from as to how I can check for these convertibility requirements: static assertions with the type traits selector is_convertible, or using concept_check to check if the assignment operation is defined for the types. Even though the former leads to shorter code, the latter works just as well as I already have concepts modeled for these types, so I could just dump the inter-convertibility checks inside as well without much hassle. In summary, I either choose this: BOOST_STATIC_ASSERT((boost::is_convertible<T1, T2>::value)); or this: template<class T1, class T2> class Concept { public: void constraints() { // insert other constraints a = b; b = a; } T1 a; T2 b; }; Thanks a lot, and have a good day Benjamin Lau

2 1

[mpl] erase_key confusion
by Scott Meyers 02 Apr '07

02 Apr '07

I expect the following code to compile cleanly. It does not. #include <boost/mpl/set.hpp> #include <boost/mpl/equal.hpp> #include <boost/mpl/erase_key.hpp> namespace mpl = boost::mpl; using mpl::_1; using mpl::_2; struct A {}; struct B {}; typedef mpl::set<A, B> set1; typedef mpl::set<B, A> set2; BOOST_MPL_ASSERT(( mpl::equal< mpl::erase_key<set1, A>::type, mpl::erase_key<set2, A>::type > )); I'm simply trying to assert that the sets {A,B} and {B,A} have the same elements after A has been removed from both. What am I doing wrong? Thanks, Scott

4 10

[threads] Passing Data and Thread Joining
by Sami Lakka 02 Apr '07

02 Apr '07

I have an method in my application which is called by a third-party app by every 1 second. In my method I have to do some intensive math operations as well as persist data. Because I have to do the math operations in the one second period I'm creating a thread that will handle the file operations (as there are no hurry for these). My question is how should I pass the data to my file saving thread? The data is in shared_array<double> so should the data be passed by value or by reference? My code looks like as follows: void workerMethod() { // do some mathematics shared_array<double> data = .... thread saveThread(bind(&DataFileSaver::saveInThread, dataFileSaver,data)); // Thread is not joined } I think that if I pass by reference the data variable can go out of scope and my thread can end up saving garbage. Am I correct? My second question is that in my saveInThread method I'm using scoped_lock to guard the file operations. When I run some unit tests for this method I'm experiencing crashes and memory leaks. This is probably because I'm not joining the started thread in the workerMethod but in my unit tests I use sleep() enough so that the threads should have performed the operations. What am I doing wrong?

1 0

[graph] adjacency_list with edges in mulitsetS does not seem to sort
by Carsten Scholtes 02 Apr '07

02 Apr '07

Hello! I'm trying to use boost::graph's adjacency_list<listS, multisetS,...>. The documentation ( boost_1_33_1/libs/graph/doc/adjacency_list.html ) states for the function edge_range(): > This function only works when the OutEdgeList for the adjacency_list > is a container that sorts the out edges according to target vertex, > and allows for parallel edges. The multisetS selector chooses > such a container. Attached you find a small program demonstrating a case where edges do not seem to get sorted accordingly. The last comment in the program contains a sample output. * The line below "All edges" shows the sequence of all edges. It does not seem to be sorted in the specified way. Probably as a consequnce of this, edge_range() does not work as expected: * The last line of the output is generated by using edge_range(). It should output edges from node 1 to node 0. There is only one such edge. Instead of that, edges (1,2) and (1,0) are returned. Am I using the interface in a wrong way or is there an error in the implementation of adjacency_list<>? Thank you in advance for any comment or advice. Carsten Scholtes

2 2