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

building boost with python script
by Michał Nowotka 22 Oct '09

22 Oct '09

Hello, last time I didn't placed in my message an appropriate heading so I repeat it: I want to build boost library with python scrpt(s) (maybe SCons shuld be used). Or maybe someone has already done so - I don't want to reinvent the wheel. -- Regards Michał Nowotka

3 4

asio stability and scalability
by Eric Twietmeyer 21 Oct '09

21 Oct '09

Hello, The project I am involved with required porting a Windows only networking application that was using the WSAAsynSelect mechanism to GNU/Linux, while remaining Windows compatible. We chose Boost asio as the framework to use. After spending time with asio and understanding it, the port was relatively straight forward. It basically works, both on Windows and on GNU/Linux. However, on both systems we are now running into stack corruptions. We have spent quite a few weeks now trying to track this down, mostly on Windows, as we have access to very many tools on that platform. We have verified that all underlying Win32 (and Winsock) function calls are valid (all parameters check), that our heap is not corrupt (all memory allocations and deallocations are clean), and we have eliminated all threading issues by restructuring the code to be runnable on a single io_service::run() thread (and Intel's Thread Checker found no issues either). We have done many other verifications as well, everything comes out clean. At this point we have to strongly consider the possibility that it is asio itself that is causing this error, though we are quite at a loss to explain how asio could result in this behavior. So we are wondering, are there known large production systems out there that use asio? Are there high volume web servers for instance written with asio? I came across this comment: http://dobbscodetalk.com/index.php?option=com_content&task=view&id=1700&Ite… which, if true, would also imply that asio is perhaps not appropriate for high throughput applications, at least on GNU/Linux systems. My basic question is: Do people have success stories using asio in production applications? Have people used other frameworks, for instance ACE, and can comment on how they compare? Have others ran into random crashes and given up on using asio? Thanks for any feedback, -Eric

7 10

Question about Kolmogorov Max Flow function.
by Dev Ganesh 21 Oct '09

21 Oct '09

Hello, I am using the Kolmogorov Max Flow function from the BOOST Graph library to develop some software. I compared it with the Edmond Karps Max Flow function using some test cases. Each function gives the correct flow value, but the Kolomogorov function does not seem to give the correct set of vertices corresponding to the minimum cut. I would like to know if my input to the function is wrong or if I'm interpreting the vertex coloring wrongly. The Edmond Karps function seems to work fine, but the Kolomogorov function is faster and if possible, I would like to use it instead of the Edmond Karps function. Could someone please help me with this? Thanks, Dev Ganesh.

1 0

[date_time] binary serialization / integer representation
by Rutger ter Borg 21 Oct '09

21 Oct '09

Hello, I would like to serialize date_time types to a binary archive, and/or convert them to, say, a integer types. I see the counted time system does use a 64 bit integer for ptime, but doesn't provide access to it. Is there something available for this use, or should I create a conversion function? Thanks in advance, Cheers, Rutger

2 4

ASIO: sending QString (UTF-16)
by Rangel [LISTAS] 21 Oct '09

21 Oct '09

Hello, I am learning ASIO, and I need to send a QString (from QT) thru async_write. The "buffer" function does not supports QString, so I wrote my own, is it correct? "utf16()" returns a pointer to a ushort (16 bit values). namespace boost { namespace asio { inline const_buffers_1 buffer(const QString &data) { return const_buffers_1(const_buffer(data.utf16(), data.size()*sizeof(ushort))); } } }; As an example, I am trying to send HTTP header and data, using a simple HTTP server sample. As the header needs to be ASCII, I am sending a vector, with a std::string and a QString, is the way it is done right? (The "message_" had a Japanese string for effective test that was removed to put on this email) ba::ip::tcp::socket socket_; std::vector<ba::const_buffer> lbuf; static std::string header_; static QString message_; std::string connection::header_="HTTP/1.0 200 OK\r\nContent-Type: text/html; charset=utf-16\r\n\r\n"; QString connection::message_=QString::fromUtf16(L"<html><head><title>test</title>" L"</head><body><h1>Test</h1><p>This is a test!</p>" L"</body></html>"); void handle_read(const boost::system::error_code& error, size_t bytes_transferred) { lbuf.clear(); lbuf.push_back(ba::buffer(header_)); lbuf.push_back(ba::buffer(message_)); ba::async_write(socket_, lbuf, boost::bind(&connection::handle_write, shared_from_this(), ba::placeholders::error, ba::placeholders::bytes_transferred)); }

2 1

NumPy integration with Boost.Python
by Tanja Zaric 21 Oct '09

21 Oct '09

Hello, I am new to Boost. Python. I am trying to find out how and maybe to integrate C++/Boost Python with Numpy? I have tried to find any documentation and tutorials about it on boost.org and elsewhere, but it is scarse. Can someone give any answer? Thanks __________ Information from ESET NOD32 Antivirus, version of virus signature database 4528 (20091021) __________ The message was checked by ESET NOD32 Antivirus. http://www.eset.com

1 0

issues appending strings
by Sean Farrow 21 Oct '09

21 Oct '09

Hi: I have a std::wstring defined like so: std::wstring OutputString; I then use the following code to try to append to it: OutputString.append(boost::gregorian::to_iso_extended_string(d)); I receive the following errors: d:\work\access for everyone\ghg\tpproductszw\tpproductszw.cpp(63) : error C2664: 'std::basic_string<_Elem,_Traits,_Ax> &std::basic_string<_Elem,_Traits,_Ax>::append(const std::basic_string<_Elem,_Traits,_Ax> &)' : cannot convert parameter 1 from 'std::string' to 'const std::basic_string<_Elem,_Traits,_Ax> &' with [ _Elem=wchar_t, _Traits=std::char_traits, _Ax=std::allocator ] Reason: cannot convert from 'std::string' to 'const std::basic_string<_Elem,_Traits,_Ax>' with [ _Elem=wchar_t, _Traits=std::char_traits, _Ax=std::allocator ] No user-defined-conversion operator available that can perform this conversion, or the operator cannot be called Using visual c++ 2008 on a windows xp sp3 box. What's happening here, and more importantly what am I doing wrong/or not doing as the case may be? Cheers Sean.

2 2

[Boost.Asio] Is there a way to find out first address of first NIC in the local host using Asio?
by Javier Estrada 21 Oct '09

21 Oct '09

2 1

[boost.datetime] outputting to a widestring
by Sean Farrow 21 Oct '09

21 Oct '09

Hi: Is thee a version of the toiso_extended_string for wide character output? I have the following line: std::wstring test =boost::to_iso_extended_wstring(d); wich I assumed would work and fails. Any ideas? Cheers Sean.

2 1

date time format
by Sean Farrow 21 Oct '09

21 Oct '09

Hi: Within a string I have a date in the format: 22 July 2009 How can I input this in to the boost::datetime library as I need to transform it in to another format. As far as I can tell, the constructor of boost::gregorian::date doesn't takethis format as one of it's string parameters. Cheers Sean.

2 4

Re: [Boost-users] [pool] Singleton implementation
by Paul Byrne 21 Oct '09

21 Oct '09

> > Message: 8 > Date: Tue, 20 Oct 2009 06:43:40 -0800 > From: "Robert Ramey" <ramey(a)rrsd.com> > To: boost-users(a)lists.boost.org > Subject: Re: [Boost-users] [pool] Singleton implementation > Message-ID: <hbkeik$jah$1(a)ger.gmane.org> > > You might want to look at the singleton which is part of the serialization > library. > > Robert Ramey > Thanks Robert, Ray et al for the useful pointers.

1 0

Boost.Python: wrapping OpenSceneGraph smart pointers, again
by Randolph Fritz 20 Oct '09

20 Oct '09

What is wrong with the following code? I suspect I'm doing something silly, and I suspect the answer is obvious. #include <osg/Node> #include <osg/ref_ptr> #include <boost/python.hpp> #include <boost/python/pointee.hpp> using namespace osg; typedef ref_ptr<Node> NodePtr; using namespace boost::python; BOOST_PYTHON_MODULE(NodePtr) { class_<NodePtr, pointee<NodePtr> >("NodePtr") ; } -- Randolph Fritz design machine group, architecture department, university of washington rfritz(a)u.washington.edu -or- rfritz333(a)gmail.com

2 4

error setting up boost on visual c++ 6.0
by pedro chaparro 20 Oct '09

20 Oct '09

hi, i wanna use boost libraries over visual c++, i have followed all steps in the getting start guide for windows, but even so i get error, what could be the cause? thanks Compiling... example.cpp c:\documents and settings\juan antonio\escritorio\boost_1_40_0\boost_1_40_0\boost\aligned_storage.hpp(65) : error C2989: 'aligned_storage_imp<0,alignment_>' : template class has already been defined as a non-template class c:\documents and settings\juan antonio\escritorio\boost_1_40_0\boost_1_40_0\boost\aligned_storage.hpp(65) : error C2988: unrecognizable template declaration/definition c:\documents and settings\juan antonio\escritorio\boost_1_40_0\boost_1_40_0\boost\type_traits\floating_point_promotion.hpp(78) : fatal error C1506: unrecoverable block scoping error Error executing cl.exe. example.obj - 3 error(s), 0 warning(s) -- Ing Pedro Alonso Chaparro Valero Ciudad Politecnica de la Innovación iTEAM - Mobile Communications Group Polytechnic University of Valencia C\ Camino de Vera S/N, Edificio 8G 46022 Valencia, Spain

4 3

[pool] Singleton implementation
by Paul Byrne 20 Oct '09

20 Oct '09

Hi all, I am interested to find a sound C++ implementation of a Singleton, and have attempted to find one within the Boost libraries. It seems that "pool/detail/singleton.hpp" would fit the bill for me, but (as I understand it) this class/header is not intended for direct use by library users. Have I overlooked a good alternative within the Boost suite? Thanks, Paul

6 7

building boost with python script
by Michał Nowotka 20 Oct '09

20 Oct '09

Hello I want to build boost library with python scrpt(s) (maybe SCons shuld be used). Or maybe someone has already done so - I don't want to reinvent the wheel. -- Regards Michał Nowotka

1 0

[Boost.Python] Using Inheritance *and* Virtual Functions in a Python Binding
by Dan Bailey 20 Oct '09

20 Oct '09

Hi, I can't seem to get inheritance and virtual functions to work together in a simple python binding: #include <boost/python.hpp> using namespace boost::python; struct Base { virtual ~Base() { } void nonVirtualFunc() { } virtual void virtualFunc() { } }; struct Derived : public Base { ~Derived() { } void virtualFunc() { } }; struct BaseWrap: public Base, public wrapper<Base> { void virtualFunc() { if (override virtualFunc = this->get_override("virtualFunc")) virtualFunc(); else Base::virtualFunc(); } void default_virtualFunc() { this->Base::virtualFunc(); } }; BOOST_PYTHON_MODULE(test) { class_<Base, boost::noncopyable>("__Base") .def("nonVirtualFunc", &Base::nonVirtualFunc) ; class_<BaseWrap, boost::noncopyable>("Base") .def("virtualFunc", &Base::virtualFunc, &BaseWrap::default_virtualFunc) ; class_<Derived, bases<Base> >("Derived") .def("virtualFunc", &Derived::virtualFunc) ; } Now in my python interpreter: >>> import test >>> dir(test.__Base) [<built-ins>, 'nonVirtualFunc'] >>> dir(test.Base) [<built-ins>, 'virtualFunc'] >>> dir(test.Derived) [<built-ins>, 'virtualFunc'] Without using BaseWrap to do the virtual functions: >>> dir(test.Derived) [<built-ins>, 'nonVirtualFunc', 'virtualFunc'] It seems that by using the boost python wrapper, I now lose any inherited functions, and the nonVirtualFunc does not get defined in the subclasses. I may be going about this in the wrong way, but is there a way of using inheritance and virtual functions together? Thanks, Dan

1 0

poor boost::mpi performance
by Jonas Juselius 20 Oct '09

20 Oct '09

Hi! I'm using the boost::mpi library for a HPC project. I really like the interface, but I'm currently getting very poor performance from the library. I started out by serializing my objects (which are full of pointers and allocated memory, and what not), but that didn't perform at all, so I went for a more brute force approach instead. No luck there either. Essentially what I want to do in a typical case, is to send a set of indexes (4 integers) followed by an array of doubles. The arrays sizes are fixed at startup, and have a size between 10-60 kB each. There are usually many of these arrays, and the total amount of data to be communicated at the end of a calculation is of the order of 1GB. Here is my current implementation: 1) Pack the indexes into an array of 4 integers and send (or broadcast) to the receiver(s). The receiver figures out where to store the next packet based on the indexes (this takes next to no time). 2) Send the array to the receivers using: double *data = coefs->data(); world.send(who, tag, data, nCoefs); where coefs is a pointer to an Eigen2 vector, and data is a pointer to a contiguous array of doubles. I'm running the code on a big HPC cluster with individual nodes with 8 cores and 16 GB memory, all connected with Infiniband. Using this setup I achieve a maximum transfer rate of 66 MB/s doing all to one communication, which is approx. 10 times less than what I'm supposed to get. I will not even mention how long a broadcast takes, but suffice to say that it takes 20-25 times longer than doing the calculation. I get the same poor performance regardless of whether I'm communicating only over 127.0.0.1 or over the net. Since out environment is homogeneous, I have compiled the both the mpi library and my program defining the BOOST_MPI_HOMOGENEOUS macro. I will try to batch more packages into larger units, but earlier experiences (with basic MPI) has shown that with 65 kB arrays, transfer rates of 1 GB/s are possible over our Infiniband switch. Asynchronous transfer is an option, but that complicates the load balancing algorithms to a point where I really don't want to go unless under gun point. Any suggestions? Best regards, -jonas- -- ________________________________________________________________________ Dr. Jonas Jusélius Centre for Theoretical and E-mail : jonas.juselius(a)uit.no Computational Chemistry Telephone : +47 77644079 Department of Chemistry Fax : +47 77644765 University of Tromsø Mobile ph.: +47 47419869 N-9037 Tromsø, NORWAY http://jonas.iki.fi _______________________________________________________________________ [ PGP key : keyserver or http://jonas.iki.fi/pubkey.asc ] [ Fingerprint: 2516 A57A 3012 7962 287D B66E C1A9 157F 0A59 7A66 ]

2 2

invalid argument?
by pedro chaparro 20 Oct '09

20 Oct '09

hi, i working on the creation of a co-simulation framework , for that i'm using asio libraries and serialization to send exchange information between simulators, i'm getting a error that say me " invalid argument" but just that, i dont know about what argument is refered =(. some know which could be the cause(s)?thanks -- Ing Pedro Alonso Chaparro Valero Ciudad Politecnica de la Innovación iTEAM - Mobile Communications Group Polytechnic University of Valencia C\ Camino de Vera S/N, Edificio 8G 46022 Valencia, Spain

3 6

[Boost.Python] Binding overloaded static functions
by Dan Bailey 20 Oct '09

20 Oct '09

Hi, I'm sure this is really trivial, but I'm not sure how to handle overloaded static functions with Boost.Python, and I can't figure it out by going through the documentation. Here's the simple example I'm trying to compile: #include <boost/python.hpp> using namespace boost::python; class Test { public: Test() { } ~Test() { } static bool func(); static bool func(int a); }; BOOST_PYTHON_MODULE(test) { bool (Test::*func1)() = &Test::func); class_<Test>("Test") //.def("func", func1) ; } Error: src/main.cpp: In function 'void init_module_test()': src/main.cpp:18: error: no matches converting function 'func' to type 'bool (class Test::*)()' src/main.cpp:12: error: candidates are: static bool Test::func() src/main.cpp:13: error: static bool Test::func(int) How do I do this? Thanks, Dan

2 1

[Spirit] How to use the scanner against UTF-8 characters
by HT4N - 20 Oct '09

20 Oct '09

Is it possible to customize the scanner in Spirit to parse UTF-8 characters? I can't seem to find any notes in the documentation specific to handling UTF-8 encoding. Thanks HT _________________________________________________________________ Hotmail: Powerful Free email with security by Microsoft. http://clk.atdmt.com/GBL/go/171222986/direct/01/

2 1

random
by Robert Ramey 19 Oct '09

19 Oct '09

I'm interested in using the functionaliy from the random library. I arm wrestled with the documentation and couldn't get my ap to compile. of the first page of the documentation. I checked the test matrix and the library is showing no errors. So I just tried to compile random_demo.cpp mentioned in the forth line. I got a bunch of cryptic (to me) warnings. I don't think this should happen. They are just warnings so I suppose they are irrelevant to my own problems but still it's annoying. Robert Ramey Compiling... generate_test_data.c c:\BoostRelease\boost\random\detail\pass_through_engine.hpp(49) : warning C4512: 'boost::random::detail::pass_through_engine<UniformRandomNumberGenerator>' : assignment operator could not be generated with [ UniformRandomNumberGenerator=boost::random::linear_congruential<boost::int32_t,48271,0,2147483647,399268537> ] c:\BoostRelease\boost\random\variate_generator.hpp(92) : see reference to class template instantiation 'boost::random::detail::pass_through_engine<UniformRandomNumberGenerator>' being compiled with [ UniformRandomNumberGenerator=boost::random::linear_congruential<boost::int32_t,48271,0,2147483647,399268537> ] c:\Projects\vario\test_windspeed\generate_test_data.c(32) : see reference to class template instantiation 'boost::variate_generator<Engine,Distribution>' being compiled with [ Engine=base_generator_type &, Distribution=distribution_type ]

2 2

[bitfield] Pre-review request
by vicente.botet 19 Oct '09

19 Oct '09

Hi all, I have adapted the Boost.Bitfield library from Emile Cormier with its permision and I have requested the Review Wizards to include Boost.Bitfield on the review schedule. The library is now quite stable. Even if there were not too much interest in the Boost ML, only 140 dowloads from the Boost Vault, the version on which it is based had more that 1500 downloads, so I suppose that the subject interest a lot of people. I really think Boost.Bitfield is the complement the Boost.Endian library to manage with all the endian issue, allowing to mange not only endian interger issues, but also endian bitfield issues. I hope that including it on the review schedule will generate more interest and help me to improve the library. I would appreciate it if people could participate to this Pre-review, reading the documentation, testing on their compiler, .... Every comment is wellcome. BTW, I'm looking for a review manager. Thanks for all, Vicente Boost.Bitfield Author(s): Vicente J. Botet Escribá, Emile Cormier Download: `Boost Vault: <http://www.boostpro.com/vault/index.php?action=downloadfile&filename=bitfie…> Boost Sandbox: https://svn.boost.org/svn/boost/sandbox/bitfield Documentation: https://svn.boost.org/svn/boost/sandbox/bitfield/libs/bitfield/doc/index.ht… Html doc also included the Vault package Categories: Portability, Integer Description: Portable bitfields traits. Boost.Bitfield consists of: * a generic bitfield traits class providing generic getter and setter methods. * a BOOST_BITFIELD_DCL macro making easier the definition of the bitfield traits and the bitfield getter and setter functions. Example declaration: struct Rgb565 { uint16_t storage; BOOST_BITFIELD_DCL(uint16_t, storage, unsigned char, red, 0, 4); BOOST_BITFIELD_DCL(uint16_t, storage, unsigned char, green, 5, 10); BOOST_BITFIELD_DCL(uint16_t, storage, unsigned char, blue, 11,15); }; Example usage: Rgb565 r; r.storage= 0xffff; // Write to a bitfield. Note that parenthesis are needed around the bitfield r.red() = 23; // Read from a bitfield unsigned char b = r.blue(); In a near future the library could add the following if the users consider it useful. * Add bitfields group * Used to easily manipulate groups of bitfields the same way as does C bitfields, but in a portable manner. Example declaration: struct Rgb565 { struct red {}; struct green {}; struct blue {}; typedef bitfields<mpl::vector< member<unsigned char, red, 5>, member<unsigned char, green, 6>, member<unsigned char, blue, 5> > > type; }; Example usage: Rgb565::type r = make_bitfields<Rgb565::type, 1,2,3>; // Write to a bitfield. r.get<Rgb565::red>() = 23; //or r.set<Rgb565::red>(23); // Read from a bitfield Rgb565::at<Rgb565::blue>::value_type b = r.get<Rgb565::blue>(); * Add pointer_plus_bits * Based on a.. The article of Joaquin Optimizing red-black tree color bits, b.. the implementation of Ion pointer_plus_bits from Boost.Intrusive , and c.. Clang's QualType smart pointer This class will allows to use the unsused bits of a pointer to reduce the size of the nodes containing pointers and bits and sometimes improving also the performances. I have not reached yet the interface I would like to have. For the moment we can do typedef pointer_plus_bits<int*,1,bool>::type pint_and_bool; int i=0; pint_and_bool v1; ASSERT_EQUALS(v1.pointer(),0); ASSERT_EQUALS(v1.small_int(),false); pint_and_bool v2(&i, true); ASSERT_EQUALS(v2.pointer(),&i); ASSERT_EQUALS(v2.small_int(),true); v1.pointer()=v2.pointer(); v1.small_int()=true; ASSERT_EQUALS(v1.pointer(),&i); ASSERT_EQUALS(v1.small_int(),true); typedef pointer_plus_bits< pointer_plus_bits<int*,1,bool>::type ,1, bool>::type pint_and_bool_bool pint_and_bool_bool v1; ASSERT_EQUALS(v1.small_int(),false); ASSERT_EQUALS(v1.pointer().get_pointer(),0); ASSERT_EQUALS(v1.get_pointer().get_pointer(),0); ASSERT_EQUALS(v1.get_pointer().small_int(),false);

1 0

[interthreads] Pre-review request
by vicente.botet 19 Oct '09

19 Oct '09

Hi, I have requested the Review Wizards to include Boost.Interthreads library on the review schedule. Even if there were not too much interest for this library on this ML, there were more than 1000 dowloads from the Boost Vault. I'm open to reduce the scope of the library, if there are some parts that are not enough generic. I would appreciate it if people could participate to this Pre-review, reading the documentation, testing on their compiler, .... Every comment will help me a lot to improve the library. BTW, I'm looking for a review manager. Thanks in advance, Vicente InterThreads __________ :Author: Vicente J. Botet Escribá :Review Manager: ??? :Download: `Boost Vault: <http://www.boostpro.com/vault/index.php?action=downloadfile&filename=intert…> Boost Sandbox: https://svn.boost.org/svn/boost/sandbox/interthreads Documentation: https://svn.boost.org/svn/boost/sandbox/interthreads/libs/interthreads/doc/… Html doc also included the Vault package :Description: Boost.InterThreads extends Boost.Threads adding thread setup/cleanup decorator, thread specific shared pointer, thread keep alive mechanism and thread tuples. * thread decorator: thread_decorator allows to define setup/cleanup functions which will be called only once by thread: setup before the thread function and cleanup at thread exit. * thread specific shared pointer: this is an extension of the thread_specific_ptr providing access to this thread specific context from other threads. As it is shared the stored pointer is a shared_ptr instead of a raw one. * thread keep alive mechanism: this mechanism allows to detect threads that do not prove that they are alive by calling to the keep_alive_point regularly. When a thread is declared dead a user provided function is called, which by default will abort the program. * thread tuple: defines a thread groupe where the number of threads is know statically and the threads are created at construction time. * set_once: a synchonizer that allows to set a variable only once, notifying to the variable value to whatever is waiting for that. * thread_tuple_once: an extension of the boost::thread_tuple which allows to join the thread finishing the first, using for that the set_once synchronizer. * thread_group_once: an extension of the boost::thread_group which allows to join the thread finishing the first, using for that the set_once synchronizer. (thread_decorator and thread_specific_shared_ptr) are based on the original implementation of threadalert written by Roland Schwarz.

1 0

Re: [Boost-users] asio stability and scalability
by Marat Abrarov 19 Oct '09

19 Oct '09

At least at Windows asio is the best - better then ACE proactor because of strands and custom memory allocation support for them. You may look at http://sourceforge.net/projects/asio-samples/ for more complex echo server than the asio examples have. Stack corruption may be because of wrong lifetime of used objects... and it may be not seen and debug because of different time scale. Regards, Marat.

1 0

Iterators and subclasses
by Swerts, Benjamin 19 Oct '09

19 Oct '09

Hello all, I'm trying to create a base class that provides 2 iterators (begin/end). These will be used to iterate over data structures that will be provided by subclasses. These data structures are completely unrelated. Normally this can be solved using virtual functions but the problem is that iterators are normally returned by value so I cannot go that route. My concrete problem: struct Point { int x, y; }; class PointIterator : public boost::iterator_facade<PointIterator, const Point, boost::random_access_traversal_tag, Point>; class PointCollection { public: typedef PointIterator const_iterator; const_iterator begin(); const_iterator end(); }; >From the PointCollection base class I want to derive concrete implementations: class Collection1 : public PointCollection { public: Collection1(char* data, size_t size) : data_(data), size_(size) {}; private: char* data_; size_t size_; }; Collection1 would return a Point with x = data_[index] and y = index * 1000, for example. Other subclasses might contain 2 vectors or just generated values. In my first attempt I tried to subclass the iterator too and make it return the correct Point when the iterator is dereferenced. As the iterators are returned by value, however, the subclass was sliced right off. An iterator-less solution is easy: class PointCollection { public: virtual Point getPoint(size_t index) const = 0; virtual size_t size() const = 0; }; Easily subclassed but no iterator support. How would I solve this problem with iterators? Should I implement another iterator over them that doesn't know about the actual types of the iterators? Can indirect_iterator help me in any way? Thanks a lot in advance! Greets, Ben

4 4