October 2009 - Boost-users - lists.boost.org

Re: [Boost-users] Boost.Asio -- combining sync read with async read
by Sameer Parekh 15 Oct '09

15 Oct '09

Igor-- So I'm not sure what you mean exactly by "the server might send a few lines in a bunch." The server will only send one line at a time, in that it will call write_async for every line that it sends-- it will never call write_async on more than one line. However, the client might be doing some processing so the server might send a few lines before the client "comes up for air" next. If that's the case then all three lines will end up in the same buffer when the read_async handler is called? I would imagine so but it would be nice if I could not worry about this, so I am looking for any excuse. =) I also realized upon looking at the chat example in more detail that I missed the whole queue thing that it does. I will aim to replicate that in my app now. In reading through, it occurs to me that I might be making the same mistake with write_async -- i.e. if I call write_async twice in quick succession, that could cause a problem unless I wait for the first write_async to complete before I call write_async a second time? Best, -Sameer On Oct 14, 2009, at 6:21 PM, boost-users-request(a)lists.boost.org wrote: > If this is the case in your application (i.e. the server might send > few > lines in a bunch), then I think yes, you have to inspect the sterambuf > manually. > Actually, this is the reason I never used read_until - in most cases > it > doesn't make your code simpler. The one case where it does is when > your > protocol has a text header and a body (like in HTTP).

3 2

Question on unordered_map
by Arun Raghavan 15 Oct '09

15 Oct '09

Hi, Do the underlying hashtable implementations of unordered_map/multimap store a size field (i.e. do they support a constant-time size function)? Same query about the load factor and bucket sizes. thanks. -b

2 1

[units][ublas] how would i use typedef in units?
by David Carricajo 15 Oct '09

15 Oct '09

Hi, Using units and ublas for new code is great, but how about using both in legacy code? I’ve got some legacy numerical code using primitive types, i plan to use units for some reasons, i.e. to check performance using both primitive types and boost::units types, and to check my code at compile time which is its whole purpose after all. I spent a day and my code it’s already broken, specially because of the initializations and ublas, some self-explanatory code follows: ----- calculations.hpp ----------- //typedefs to get library-free code #include <complex> #include <iostream> #if USING_BOOST_UNITS #include <boost/typeof/std/complex.hpp> #include <boost/units/systems/si/power.hpp> //... using namespace boost::units; using namespace boost::units::si; using namespace std; typedef std::complex<double> complex_type; typedef quantity<electric_potential,complex_type> tPotencial; typedef quantity<current,complex_type> tIntensidad; typedef quantity<resistance,complex_type> tResistancia; //... #elif typedef double tPotencial; typedef double tIntensidad; typedef double tResistancia; //... #endif ------------------------------------------ ----- some_legacy_calculations.cpp ------ #define USING_BOOST_UNITS tPotencial tension = complex_type(12.5,0.0)*volts; //ok tPotencial tension2 = complex_type(12.5,0.0); //auch! some_file << tesion + tension2 << std::endl; some_file.close(); //get Real part from 'tension' and pretend it's a current quantity<current, double> phony_tension = tension.value().real * amperes;//compiler doesn't complain, but how do I prevent this to compile? #undef USING_BOOST_UNITS tPotencial tension = complex_type(12.5,0.0)*volts; //auch! tPotencial tension2 = complex_type(12.5,0.0); //ok some_file.open("/*some_file*/"); some_file >> tesion + tension2 >> std::endl; //auch! #define USING_BOOST_UNITS //use ublas ublas::matrix<tAdmitancia, ublas::column_major> admi(2, 2); ublas::matrix<tPotencia, ublas::column_major> ind(2, 1); ind(0, 0) = complex_type (1, 0)*volts; ind(1, 0) = complex_type (-1, 0)*volts; admi(0, 0) = complex_type (0, -2)*siemens; admi(0, 1) = complex_type (0, 2)*siemens; admi(1, 0) = complex_type (0, 2)*siemens; admi(1, 1) = complex_type (0, -2)*siemens; int er = lapack::gesv(admi, ind);//auch! ------------------------------------------ For each of those problems i found some workaround but I'd like to discuss whether it's ok or not. Thanks for your time.

4 4

[serialization] Short form of empty tags in XML archive
by Sergey Filippov 15 Oct '09

15 Oct '09

Hi, I'm evaluating possibility of using the serialization library (1.40.0) to save/load an application settings to/from XML that must be a part of another XML. I faced with some minor issues that were successfully solved but then arisen the major one: the serialization library does not support <tag/> syntax. Why I need this? After the whole XML is generated it is transformed by some XML processor to internal data structures and back to text. During the transformation empty tags <str></str> (for example, produced when serializing an empty std::string object) are replaced with the short form: <str/>. From the XML perspective, these forms are equivalent so XML processors allowed to do so. Since I cannot change the XML processor, I started investigating the serialization library. Specifically, how XML is processed. I found out that parsing XML is responsibility of xml_grammar class. It does not recognize <tag/> as a start tag so an exception is generated. It is easy to write a new class with needed functionality but xml_grammar is not designed to be used polymorphically... So, after all, I did not find an easy way to write my own archive. After that, I tried to modify the library. It was straightforward, and it works for me. Attached are original (*.bak) files and modified. Can anyone suggest a way to achieve such a functionality without modifying the library? Don't you think, this functionality is quite general to be included in the library? Sergey

1 0

Already successfully running Boost on Linux, looking for advice from Windows Boost adopters
by Ramon F Herrera 15 Oct '09

15 Oct '09

Some weeks ago, Boost caught my attention, and I am now a more than satisfied user. The assistance received in this forum is one of the factors in my "two thumbs up" evaluation of Boost. Give yourselves a round of applause. :^) Since I can't help but getting into progressively harder stuff (i.e., getting into trouble), I am considering using Boost on the Windows side. My background is Unix development but I have been teaching myself Visual C++ lately. I mostly build and integrate stuff written by other programmers. I am trying to decide whether I should delve deeper into Boost on the Linux side (which is quickly becoming a walk in the park), or take the plunge into Windows. Are there gotchas or caveats? Can I just download Boost and select "Build solution"? My current platform is Visual Studio, Version 8 (2005). I don't intend to upgrade it until I have no choice. I have found that to be a safe policy -better safe than sorry-, specially when the software comes from Redmond. -Ramon

7 12

Boost.Serialization not building
by OvermindDL1 14 Oct '09

14 Oct '09

I attempted the build Boost.Serialization from trunk and got some errors: binary_iarchive.cpp .\boost/serialization/void_cast.hpp(255) : error C2572: 'boost::serialization::void_cast_register' : redefinition of default parameter : parameter 2 .\boost/serialization/void_cast_fwd.hpp(30) : see declaration of 'boost::serialization::void_cast_register' .\boost/serialization/void_cast.hpp(255) : error C2572: 'boost::serialization::void_cast_register' : redefinition of default parameter : parameter 1 .\boost/serialization/void_cast_fwd.hpp(30) : see declaration of 'boost::serialization::void_cast_register'

2 7

Anyone having Intellisense weirdness with 1.39 and VS 2008?
by Eric J. Holtman 14 Oct '09

14 Oct '09

First off, as a newbie to the list, is there a searchable archive anywhere? I found the archives, but other than scanning the thread by subject, I couldn't see a way to find out if this question had been answered.... so, here goes... I was using 1.35. I decided to upgrade to 1.39, since I was having one of those days where I started upgrading other stuff, and figured, "what the hell....." With 1.39, certain classes in certain projects seem to show up in Intellisense (that is, the Class Viewer) as if they were best 7 or 8 levels deep in a boost namespace. That is, under my project, there will be a boost {} entry. I expand that, there's another boost {} entry. I keep doing that for a while, and finally, my class shows up. That class isn't really in the boost namespace. And it's only the MSFT browser that is confused, everything compiles and runs normally. If I back out to 1.35, this goes away. Any clues?

4 6

Template class and any_image_view
by tatta 14 Oct '09

14 Oct '09

Hello, I'm in trouble using dynamic images into a template class. I have an abstract class Filter of this form: template <typename TView1, typename TView2> class Filter { public: virtual ~Filter(); virtual void setParameter(uint8 id, float16 value); virtual void apply(const any_image_view<TView1>& source, const any_image_view<TView2>& destination) = 0; protected: Filter(); }; which is subclassed by other concrete classes like this one: template <typename TView1, typename TView2> class BlurGaussSelective : public Filter<TView1,TView2> { public: BlurGaussSelective(); virtual ~BlurGaussSelective(); void setParameter(uint8 id, float16 value); void apply(const any_image_view<TView1>& source, const any_image_view<TView2>& destination); }; As you can see I need to use a template class because the "apply" method (where concrete filters implements own logic) is virtual. But when I instantiate a BlurGaussSelective object I don't know which template types pass to it. I tried this: typedef vector<rgb8_image_t,gray8_image_t> ImageType; typedef vector<rgb8_view_t,gray8_view_t> ViewType; any_image<ImageType> source; any_image<ImageType> dest; jpeg_read_image("input.jpg", source); jpeg_read_image("input.jpg", dest); BlurGaussSelective<ViewType,ViewType> filter; filter.apply(const_view(source),const_view(dest)); But the compiler gives me this error: ../src/RotoscoperApplication.cpp: In member function ‘void Rotoscoper::Application::rotoscope()’: ../src/RotoscoperApplication.cpp:73: error: no matching function for call to ‘Rotoscoper::BlurGaussSelective<boost::mpl::vector<boost::gil::image_view<boost::gil::memory_based_2d_locator<boost::gil::memory_based_step_iterator<boost::gil::rgb8_pixel_t*> > >, boost::gil::image_view<boost::gil::memory_based_2d_locator<boost::gil::memory_based_step_iterator<boost::gil::gray8_pixel_t*> > >, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na>, boost::mpl::vector<boost::gil::image_view<boost::gil::memory_based_2d_locator<boost::gil::memory_based_step_iterator<boost::gil::rgb8_pixel_t*> > >, boost::gil::image_view<boost::gil::memory_based_2d_locator<boost::gil::memory_based_step_iterator<boost::gil::gray8_pixel_t*> > >, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na> >::apply(boost::gil::any_image_view<boost::mpl::v_item<boost::gil::image_view<boost::gil::memory_based_2d_locator<boost::gil::memory_based_step_iterator<const boost::gil::gray8c_pixel_t*> > >, boost::mpl::v_item<boost::gil::image_view<boost::gil::memory_based_2d_locator<boost::gil::memory_based_step_iterator<const boost::gil::rgb8c_pixel_t*> > >, boost::mpl::vector0<mpl_::na>, 0>, 0> >, boost::gil::any_image_view<boost::mpl::v_item<boost::gil::image_view<boost::gil::memory_based_2d_locator<boost::gil::memory_based_step_iterator<const boost::gil::gray8c_pixel_t*> > >, boost::mpl::v_item<boost::gil::image_view<boost::gil::memory_based_2d_locator<boost::gil::memory_based_step_iterator<const boost::gil::rgb8c_pixel_t*> > >, boost::mpl::vector0<mpl_::na>, 0>, 0> >)’ ../src/RotoscoperBlurGaussSelective.hpp:119: note: candidates are: void Rotoscoper::BlurGaussSelective<TImage1, TImage2>::apply(const boost::gil::any_image_view<IVTypes>&, const boost::gil::any_image_view<Types2>&) [with TImage1 = boost::mpl::vector<boost::gil::image_view<boost::gil::memory_based_2d_locator<boost::gil::memory_based_step_iterator<boost::gil::rgb8_pixel_t*> > >, boost::gil::image_view<boost::gil::memory_based_2d_locator<boost::gil::memory_based_step_iterator<boost::gil::gray8_pixel_t*> > >, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na>, TImage2 = boost::mpl::vector<boost::gil::image_view<boost::gil::memory_based_2d_locator<boost::gil::memory_based_step_iterator<boost::gil::rgb8_pixel_t*> > >, boost::gil::image_view<boost::gil::memory_based_2d_locator<boost::gil::memory_based_step_iterator<boost::gil::gray8_pixel_t*> > >, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na>] make: *** [src/RotoscoperApplication.o] Error 1 Which template type I should use when I instantiate the BlurGaussSelectiveClass? Thanks in advance. Greetings, Raffaele -- View this message in context: http://www.nabble.com/Template-class-and-any_image_view-tp25891719p25891719… Sent from the Boost - Users mailing list archive at Nabble.com.

2 2

Re: [Boost-users] Boost.Asio -- combining sync read with async read
by Sameer Parekh 14 Oct '09

14 Oct '09

> Right. The most simple way to accomplish this is to use > asio::streambuf as a > buffer. When you process the input data, you call consume() function > only > for the amout you've really processed, and the rest of the data > remains > unused in the streambuf - until the next time you get something from > the > server (look at the asio examples with streambuf). Yes, I'm using the streambuf, and then constructing an istream from the streambuf, and then calling getline on the istream. (The chat client/server was the primary example that I used to build my app.) If I have 2.5 lines in the buffer then the first two calls to getline will give me the first two lines, but won't the third call to getline get me the half line, and consume that half line from the streambuf? Do I need to process the streambuf in a more manual fashion rather than calling getline on the istream? Also to make sure I understand it how it works with the streambuf, that I am understanding you correctly: if I was just processing one line in the async_handler, and the first time the server sends me 3 lines, then the handler is called and will process the first line sent. Then the server sends another 3 lines, the async handler is called and the client will process the 2nd line that it received, not the 4th line, right? Thanks for your help, -Sameer

2 1

[lambda] Initializing a std::list of unsigned short points
by Ryan McConnehey 14 Oct '09

14 Oct '09

In my project I'm trying to initialize a std::list of unsigned short pointers. I've created a couple of simple hand written loops but would like to convert them to use std::for_each. I've been unsuccessful in the past getting boost::lambda to work with std container of pointers. In the lambda documentation I know that new_ptr and delete_ptr are need in creating and destroying pointers in lambda functions. I've provided the hand written loop that does work and the "attempted" conversion to std::for_each. Any help on the correct implementation for the lambda values would be appreciated. Ryan typedef std::list<unsigned short *> ml; for (ml::iterator itor = m_list.begin(); itor != m_list.end(); ++itor, startValue++) { *itor = new unsigned short(startValue); } for (ml::iterator itor = m_list.begin(); itor != m_list.end(); ++itor) { delete *itor; } std::for_each(m_list.begin(), m_list.end(), *(bl::_1) = bl::new_ptr<unsigned short>()(bl::var(startValue)++) ); std::for_each(m_list.begin(), m_list.end(), boost::lambda::delete_ptr(*(boost::lambda::_1)));

3 3