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

Re: [Boost-users] [asio] Custom asynchronous service
by Marat Abrarov 04 Oct '12

04 Oct '12

Hi Acer, > I would like to adapt my class to asio for offering > async_do_something(param, callback) interfaces. http://sourceforge.net/projects/asio-samples/ especially (SVN): https://asio-samples.svn.sourceforge.net/svnroot/asio-samples/trunk/include… https://asio-samples.svn.sourceforge.net/svnroot/asio-samples/trunk/include… or https://asio-samples.svn.sourceforge.net/svnroot/asio-samples/trunk/include… https://asio-samples.svn.sourceforge.net/svnroot/asio-samples/trunk/include… > After reading examples presented in asio/examples/services/ and some asio > source code. > > I found the example does not do invoke callback stuff (e.g. > asio/examples/service/logger_service.cpp). > > In case of asio source code, there are classes such as > detail::wait_handler<> that wrap user callbacks for invoked later. It's not a wrapper for deferred invocation. Here's my implementation of handler parking (Asio custom memory allocation and custom invocation strategy are supported): https://asio-samples.svn.sourceforge.net/svnroot/asio-samples/trunk/include… Regards, Marat Abrarov.

1 0

[BGL] vecS for vertex container and vertex index adjustments
by Sergey Mitsyn 03 Oct '12

03 Oct '12

Hi all, I know that while using vecS for vertices one has no vertex iterator stability upon removing vertices. But I have a question if there exists something like vertex _index_ stability. The documentation says that indices for vertices are automatically adjusted so that they are still in contiguous range, but it is not specified how it is done. Is it okay to assume that if a vertex with an index I is removed, then vertices with indices < I have their indices unchanged, while vertices with indices > I have their indices decreased by one? The actual problem is removing all vertices meeting some condition. Is the following code ok? // traverse all vertices and remove those for which predicate // condition_to_remove_vertex returns true. // graph is adjacency_list<.., vecS, ...> vertex_iterator vi = vertices(graph).first, vi_end = vertices(graph).second; while (vi != vi_end) { if (condition_to_remove_vertex(vi, graph)) { vertices_size_type end_index = vertex_index[*vi_end] , vi_index = vertex_index[*vi]; remove_vertex_f(*vi, graph); // vi and vi_end are invalid --end_index; vi = vertices(graph).first + vi_index; vi_end = vertices(graph).first + end_index; // now vi and vi_end are valid again } else ++vi; } This way seems to be more effective than to start iteration again every time a vertex is removed. ------------ Sergey Mitsyn.

3 14

C++ (boost) library for serial port communication
by Alexander Dietz 03 Oct '12

03 Oct '12

Hi all, I wonder if there is a boost-library that enables easy and robust communication to a serial port, just as e.g. the python library pyserial does. Maybe there is a boost-library which can be used for those purpose instead of implementing my own C-style classes and functions? Thanks a lot, Alex

5 7

A couple of odeint questions
by Tiago Gehring 02 Oct '12

02 Oct '12

Hello, I've been using the odeint library for some time now - and I really like it, thanks for the great work! I know that the odeint library is currently being reviewed but I'm actually writing here to ask a couple of questions about my usage scenario, hope that I'm in the right place. My usage scenario involves neural networks where I have a set of n equations for the N neurons themselves and another m equations for all the M synapses (connections) where usually M >> N and the two sets of equations depends on each other. What I'm trying to do is to put all these N + M equations into one large system and to integrate them at once while also keeping the state vectors separated. This is because of the different dimensionality of the two sets of equations and because my intention is to later use OpenCL to do the calculations (which as I understand works better if you use something like a vexcl::multivector<double, M> + another vexcl::multivector<double N> and I don't see how to pack all of this in one large state vector); So my first question is: can I somehow "combine" two sets of state vectors into one thing or integrate two systems in parallel (synchronously)? I tried the former (one combined state object) and even defined my own combined algebra (a std::pair of std::vectors - later vexcl::multivectors) but then stumbled into problems with my value type (which I defined as a std::array<double, 2>). It seems that I have to define all possible mathematical and comparison operations for my pairs (arrays). And my second question: my problem involves integrating the system until one of the neurons fires - at which point it's state variable (membrane potential) jumps to a fixed value, i.e., I have a sharp discontinuity in the state variable. My way of dealing with this right now was to implement a custom integration function (based on the adaptive integration one) which has a built in function that checks if the threshold was reached at every integration step. When this happens I break the integration, reset the state of the neuron in question to the fixed value and start integrating again. I wonder if it's possible to solve this in an easier way... Is it possible for example to change the state on the fly and can the system handle discontinuities? (i.e. can I integrate without stopping each time a neuron fires). Thanks a lot, Tiago

2 2

lisp like operetor collection on well kown C++ data structure
by Niitsuma Hirotaka 02 Oct '12

02 Oct '12

I wrote scm2cpp.hpp https://github.com/niitsuma/scm2cpp.hpp/blob/master/scm2cpp.hpp scm2cpp.hpp is lisp like operetor collection on well kown C++ data structure lisp function car cdr cons list-ref set-car! .. etc can use for std::vector std::list boost::ptr_list boost::fusion::list and std::pair is regard as cons example usage https://github.com/niitsuma/scm2cpp.hpp/blob/master/usage.cpp https://github.com/niitsuma/scm2cpp.hpp/blob/master/list-test.cpp Also eq? eqv? equal? quote symbol functions are implemented. plz see usege https://github.com/niitsuma/scm2cpp.hpp/blob/master/equal-test.cpp eq? function implimeted as direct addrest comparison template<typename T> bool is_eq(T & x, T & y) {return (&x)==(&y);} thus cons(T, std::list<T>) becomes boost::ptr_list<T> All lisp like operator keep same address for same object.

1 0

[Build] configure files copied into stage directory
by Alexander Khomyak 01 Oct '12

01 Oct '12

Hi. I build library binaries from source using build system. And now the system copies dll and lib-files into stage directory. But I also need to copy pdb-files. How to configure the build system to do so? Regards, Alexander Khomyak

1 0

[ICL] Specific interval_map
by A G 01 Oct '12

01 Oct '12

Hello, I am trying to see whether I can use interval_map for my project. I was initially amazed with interval_map and thought this would be it, but then I have some doubts: I need a interval tree like interval_map, but the aggregated values have the following equivalent property: equal to the length of the interval it is aggregated with. An example: [0:3] -> 3 [4:8] -> 4 Then if I add the interval [2:5] with aggregated value 3 (step 1) to this container, this should end up with: [0:1] -> 1 [2:3] -> 1 [4:5] -> 1 [6:8] -> 2 which I would later need to change into (step 2) : [0:1] -> 1 [2:5] -> 3 ([2:3] and [4:5] are merged because what matter is the last inserted interval: [2:3]). [6:8] -> 2 I am not sure I can do step 1 in just overloading the operator+= over my aggregated values. Could you please confirm ? If so, is there any part of the library I can reuse to do this task (like a function overloading, but which one) or do I have to rewrite an ad-hoc insert-like method myself ? Thank you for your hints. BR, AG.

2 2

Video About Open Source Credit Rating Tool Using Boost
by Marc Joffe 30 Sep '12

30 Sep '12

I am looking for users and developers interested in collaborating on an open source sovereign rating model project. If you don't agree with your country's credit rating, here is a chance to do something about it! A video describing this project is available at http://www.youtube.com/watch?v=71d3FXCqKtY and a journal article explaining the rationale can be found at http://econjwatch.org/articles/rating-government-bonds-can-we-raise-our-grad e. The tool is available at http://www.publicsectorcredit.org/pscf.html. Source is posted at http://www.github.com/joffemd/pscf. The tool uses a Monte Carlo simulation which includes calls to Boost's normal and Cauchy random number generators. There are many opportunities for improvement upon the initial version. Please contact me if interested. Regards, Marc Joffe Public Sector Credit Solutions marc(a)publicsectorcredit.org

2 1

Re: [Boost-users] [Review] ODEInt review starts today, September 19, ends September 28
by Mario Mulansky 29 Sep '12

29 Sep '12

Hi Paul, thanks for your comments, I have just a quick note. > ... > Modern expression templates are de rigeur. Works with Boost.Units which > is a big plus. Also meshes > with Fusion and Arbitrary Precision types so will presumably work with > the recently accept > Multiprecision library. Yes, odeint works out-of-box with the Multiprecision library. I have checked this in the course of the multiprecision review process [1]. Best, Mario [1] http://lists.boost.org/Archives/boost/2012/06/194473.php

2 1

[Serialization] GCC Symbol Visibility
by drivehappy 29 Sep '12

29 Sep '12

I'm attempting to build the serialization library outside of the standard build mechanism for Boost. For the most part this works great, with the exception that the symbol visibility under GCC is not defined correctly - our compiler settings set visibility to "hidden". Snippets from boost/serialization/config.hpp: #define BOOST_SERIALIZATION_DECL(T) __declspec(dllexport) T ... #define BOOST_SERIALIZATION_DECL(T) __declspec(dllimport) T I believe there has been previous discussion on changing these to BOOST_SYMBOL_EXPORT and BOOST_SYMBOL_IMPORT respectively since the __declspec is MSVC specific (http://boost.2283326.n4.nabble.com/Questions-regarding-new-visibility-stuff…) Additionally a patch was suggested (https://svn.boost.org/trac/boost/ticket/2114) that I believe at the time covered all of the boost libraries with the same changes. It doesn't appear to have gotten into Boost.Serialization (at least as of 1.51). If it would save time I could provide a patch specifically for Serialization. Thanks.

2 1