- Boost-users - lists.boost.org

Serialization / Fast [non-portable] serialization for binary data
by Dan Eaton 14 May '09

14 May '09

Hi, One more question for the list: I have my own 1, 2 and 3 dimensional array classes, and would like to serialize large such objects quickly. In other words, for( unsigned i=0; i<some_large_dynamic_number; i++ ) ar & ptr[i]; is not desirable (or perhaps modern compilers actually do a sufficiently good job at optimizing this case? opinions?) Is this a case where I'd be better to specialize my "serialize" method for both text_?archive (portable) and binary_?archive (fast), and then further split the binary case into save/load, which would in turn invoke save_binary and load_binary (as appropriate)? In this case (assuming the above is right, and out of curiosity) why is there no "serialize_binary" method instead of separate save/load_binary methods? I'm thinking that with this method, there would be no need to split "serialize" into "save"/"load" for this case... Thank you, Danny

2 1

multi-index-container with nested templates => problem getting iterator on 2nd index.
by JohnQ 14 May '09

14 May '09

Hi, I'm really stuck on this one, so I'm hoping someone can lend me a hand. Basically, I'm defining a templated multi-index-container to hold one piece of data along with two indexes to that data. Next, I'm wrapping that in another templated class that will handle concurrent access to that structure. My problem is that I'm having trouble figuring out how to define a type that will give me an interator on the 2nd index of the multi-index-container ("bimimapIterator2" below). I'm basing this on the Boost multi-index-container example here: http://www.boost.org/doc/libs/1_39_0/libs/multi_index/example/bimap.cpp Here's what I have... //******************************************************************** // multi-index-container //******************************************************************** struct index1{}; struct index2{}; template<typename Key1,typename Key2, typename Data> struct bimultimap { typedef boost::tuple<Key1,Key2,Data> value_type; typedef boost::multi_index_container< value_type, indexed_by< ordered_non_unique< tag<index1>,member<value_type,Key1,&value_type::first> >, ordered_non_unique< tag<index2>,member<value_type,Key2,&value_type::second> > > > bimimap; }; //******************************************************************** // concurrent access class. //******************************************************************** template<typename Key1, typename Key2, typename Data> class concurrent_bimultimap { private: typename bimultimap<Key1, Key2, Data>::bimimap the_bimultimap; mutable boost::mutex the_mutex; boost::condition_variable the_condition_variable; public: typedef typename bimultimap<Key1,Key2,Data>::bimimap bimimapType; typedef typename bimultimap<Key1,Key2,Data>::bimimap::iterator bimimapIterator; typedef typename bimultimap<Key1,Key2,Data>::bimimap::index<index2>::type::iterator bimimapIterator2; void insert(Key1 const& key1, Key2 const& key2, Data const& data) { boost::mutex::scoped_lock lock(the_mutex); the_bimultimap.insert( bimultimap<Key1,Key2,Data>::bimimap::value_type(key1,key2,data) ); lock.unlock(); // Note: because we unlock above and notify below, we can starve the "pop" thread for short durations in burst situations. That's ok so far. the_condition_variable.notify_one(); } bool findKey1(Key1& key1, Data& data) { boost::mutex::scoped_lock lock(the_mutex); bimimapIterator itFindIt = the_bimultimap.find(key1); if (itFindIt != the_bimultimap.end()) { data = itFindIt->third; return true; } else return false; } bool findKey2(Key2& key2, Data& data) { boost::mutex::scoped_lock lock(the_mutex); bimimapIterator2 itFindIt = the_bimultimap.get<index2>().find(key2); if (itFindIt != the_bimultimap.get<index2>().end()) { data = itFindIt->third; return true; } else return false; } ..... there's more but this gives the essence of the problem .... };

3 3

Bidirectional Graph and undirected graphs
by Daniel Russel 14 May '09

14 May '09

I'm trying to add the needed functions to make one of our graphs work as a boost graph and ran into a contradiction in the docs: The introduction to bidirectional graph says "For undirected graphs this [need for extra information] is not an issue, since the in_edges() and out_edges() functions are the same, they both return the edges incident to the vertex." however the function docs for in_edges say: "For both directed and undirected graphs, the target of an out-edge is required to be vertex v and the source is required to be a vertex that is adjacent to v." The latter implies that undirected graphs have to have two copies of each edge, one pointed each direction. Which is correct? The former, I hope :-) Thanks.

3 5

boost:statechart: How can I access the state machine from within a custom error handler?
by Jale Gulea 13 May '09

13 May '09

Hello, I have replaced the null_exception_translator (http://www.boost.org/doc/libs/1_34_1/libs/statechart/doc/reference.html#nul…) with one of my own brew and I cannot figure a way to get the state machine in the context of which the the handler gets called. The 'machine_' member is private in both the 'Action' and the 'ExceptionEventHandler'. The 'translator_' is also private - thus inaccessible in my derived state machine. Any ideas are appreciated. Kind regards /JG

3 2

[operators] equality_comparable
by Robert Jones 13 May '09

13 May '09

Given a template class like template < typename T > struct A { template < typename U > friend bool operator == ( const A &, const U & ); template < typename U > friend bool operator == ( const U &, const A & ); template < typename U > friend bool operator != ( const A &, const U & ); template < typename U > friend bool operator != ( const U &, const A & ); // ...other stuff... }; can I express it using equality_comparable? Thanks, Rob.

2 1

[thread] mixing boost::mutex from different boost versions
by klaus triendl 13 May '09

13 May '09

Hi, are boost::mutexes from different boost versions ABI compatible, provided that the modules (boost thread library, my own) are compiled with the same compiler? My question arises because I'm creating a windows DLL with class methods taking a boost::mutex& from let's say boost 1.38, but an application (the EXE) might use e.g. boost 1.39. Klaus Triendl

2 3

[asio] using ASIO with pcap_get_selectable_fd?
by Dustin Spicuzza 13 May '09

13 May '09

Hey, We're wanting to use Boost ASIO for doing socket operations and other asynchronous operations, but we also are using libpcap on multiple links. Currently, we're just getting fd's from pcap_get_selectable_fd() and using a select() loop to dispatch the loops when it's ready. Ideally, we want to use ASIO to do the select() part now instead of what we have, so theres less code to maintain. It seems like that if you add a descriptor to an io_service, you *have* to read or write some bytes to/from it, as opposed to just waiting for it to be ready (as libpcap wants). Is there a way to add an fd to an ASIO io_service and *not* read from it, but only call a routine when its ready to read? Thanks! Dustin -- Innovation is just a problem away

2 1

[fusion][concept] sequence_base
by Pieter Op de Beeck 13 May '09

13 May '09

Hi all. I'm having difficulty understanding why the code listed below is not compiling (at least not on my box, gcc-4.3.2 / boost 1.38). (the code has been stripped and simplified for demonstration). Context: I have an interface and a model providing an implementation of this interface. I use a concept to make sure that the model provides a member function "foo" of a specific signature (whether doind it this way is useful or not, is not the point). In the example below, the member function "foo" should accept one parameter. This parameter should be of type "boost::fusion::sequence_base<SequenceT> const&". Except, when compiling the compiler informs me: invalid use of incomplete type ‘struct boost::fusion::extension::begin_impl<boost::fusion::non_fusion_tag>::apply<const boost::fusion::sequence_base<boost::fusion::list<int, int, int, boost::fusion::void_, boost::fusion::void_, boost::fusion::void_, boost::fusion::void_, boost::fusion::void_, boost::fusion::void_, boost::fusion::void_> > >’ Without the concept check in the interface class this compiles fine. Appreciating all your comments on this. Regards, Pieter. ======================main.cpp================================= #include <boost/concept/usage.hpp> #include <boost/concept/requires.hpp> #include <boost/fusion/support/sequence_base.hpp> #include <boost/fusion/container/generation.hpp> // simplified concept that requires from the model a member function 'foo' taking in a fusion:sequence_base template<class Model, typename SequenceT> struct concept { template <void (Model::*) (boost::fusion::sequence_base<SequenceT> const&) const> void model_requires_foo(); BOOST_CONCEPT_USAGE(concept) { model_requires_foo<&Model::foo>(); } }; // an interface which gives 'foo' taking in sequences template<class Model> class interface { public: template<typename SequenceT> BOOST_CONCEPT_REQUIRES( ((concept<Model, SequenceT>)), (void)) foo(SequenceT const& c) const { model_.foo(c); } private: Model model_; }; // a model that can handle fusion::sequence_base class model { public: template<typename SequenceT> void foo(boost::fusion::sequence_base<SequenceT> const& b) const { SequenceT const& sequence = b.derived(); } }; int main() { interface<model> bar; bar.foo(boost::fusion::make_list(1,2,3)); }; ==================end main.cpp==============================

2 1

Boost Unit Tests on iPhone
by Manfred Schwind 13 May '09

13 May '09

Hi, I successfully built the boost libraries for the iPhone and for the iPhone Simulator. Now I have done some work on cross plattform C++ code and I am testing this code with Boost Unit Tests. It's running fine on Mac and Windows. It's also running fine on the iPhone Simulator. But unfortunately it's not working on the iPhone Device. It's not even getting into my test code, it just "hangs" somewhere before when collecting the test cases, if I can see this right in the debugger. I experienced that if I just have one BOOST_AUTO_TEST_CASE it also works fine on the iPhone Device. The test case executes, it succeeds and the app quits. But as soon as I have two or more BOOST_AUTO_TEST_CASEs, none of these test cases gets executed and the iPhone just hangs forever and loops around in some boost unit test initialization code using "spinlocks" and "yields" ... Is this a known issue? Anyone successfully using Boost Unit Tests on iPhone? Any ideas what I could try? Thanks, Mani -- http://mani.de - friendly software iVolume - listen to music freehand LittleSecrets - the encrypted notepad Sahara - sand in your pocket Watchdog - baffle the curious

3 5

[fusion] mutable transform
by David Sankel 13 May '09

13 May '09

Hello, I'm using boost fusion to do some interesting robotics work. This has come up on the list before, but I'd like to point out my scenario where a mutable transform function would be useful. I'm writing a generic interface generator for various datatypes. For a given instance of the type, the interface generator return a widget that can modify the instance. I'd like my implementation for map types to look like this: //... for_each( transform( mapInstance , toWidget() ) , bind( Layout::addWidget, layout, _1 ) ); //... It has a very nice separation of concerns: 1) the "transform toWidget" function, and 2) the "for_each addWidget" which acts much like a fold. However, my toWidget function requires a non-const reference so this doesn't work without a hacky cast to remove the const. Hackery aside, the only way I know how do to this is using a for_each who's function argument explicity does both the conversion to a widget (with toWidget()) and the folding to the layout. Those two concepts seem separate to me. Maybe there's a better way or maybe I'm not understanding the drawbacks of exposing a mutable transpose function. Thanks, David -- David Sankel Sankel Software www.sankelsoftware.com

1 1