October 2012 - Boost-users - lists.boost.org

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