- Boost-users - lists.boost.org

[dataflow] new frameworks and GUI editors + visually edit a BGL graph
by Stjepan Rajko 20 Jul '08

20 Jul '08

Hello, I uploaded some youtube videos of a new visual editor for the Dataflow library. There are three examples, each using a different underlying framework (the visual editor code is generic): visually editing a Dataflow.Signals network (data flows instantly): * http://www.youtube.com/watch?v=7a4HJ4zVPXc visually editing a Dataflow.Managed network (data flows on request): * http://www.youtube.com/watch?v=KwrnSjhJyf4 visually editing a BGL graph (this simple example lets you construct a directed graph and query the number of strongly connected components): * http://www.youtube.com/watch?v=gVLetAAnPrc The latter two examples use proof-of-concept implementations of alternate frameworks built on top of the generic framework that is part of the Dataflow library: Dataflow.Managed is unlike Dataflow.Signals in that doesn't couple data transfer with component invocation. In other words, if producer A is connected to consumer B, when A changes its value B will not be notified immediately. Instead, after any number of values in the network has changed, the network as a whole can be told to update (the network will then notify all components whose inputs have been modified, in the order of a topological sort of the components). Alternatively, components can be invoked individually, which gives them a chance to check their inputs / modify their outputs. The BGL support layer for the Dataflow library is an example of how something that is not strictly a dataflow thing (e.g., a graph) can still satisfy Dataflow concepts (there are Components (vertices) that can be connected), and be used with the framework (e.g., you get a GUI editor for free). If you're interested in playing with this: The new GUI uses GLV (http://mat.ucsb.edu/glv/) which is included with the dataflow library in the sandbox and integrated in the bjam build of the examples. As long as you have glut/freeglut available, it should build using gcc/darwin, MSVC8+, gcc/minGW, and if you are better than me at knowing which X libraries to link with and know how to manipulate a Jamfile, gcc/linux (any GCC version 4.x should work). If you run into any problems, please let me know. The latest code is currently available from the sandbox only, instructions/docs at: http://www.dancinghacker.com/code/dataflow/ Any feedback would be appreciated. Kind regards, Stjepan

1 0

regular expressions and binary files
by Brad Tilley 20 Jul '08

20 Jul '08

I'm using boost regular expressions with good results on text files (html, xml, txt, etc), however, I'm having issues on binary file formats (i.e. MS Office files, etc). I've goggled around, but not found any hints. Just wondering if someone had experience doing REs on binary files. I've tried different open methods ios::binary, etc... still no luck. Thanks

2 1

[lambda] Using lambda expressions with STL function adaptors
by Robert Jones 20 Jul '08

20 Jul '08

This little piece of code, #include <algorithm> #include <functional> template <typename InputIterator, typename Predicate> inline bool if_all( InputIterator first, InputIterator last, Predicate predicate ) { return std :: find_if( first, last, std :: not1( predicate ) ) == last; } #include <vector> #include "boost/lambda/lambda.hpp" int main( ) { using namespace boost :: lambda; std :: vector<int> v; if_all( v.begin( ), v.end( ), _1 == 3 ); return 0; } Gives a compiler error, which I think tells me that my lambda expression, _1 == 3 does not provide the argument_type typedef. I thought lambda expressions were careful to to do this. What am I doing wrong? Thanks, Rob. -- ACCU - Professionalism in programming - http://www.accu.org

2 4

[filesystem] linker error with trunk
by Christian Henning 20 Jul '08

20 Jul '08

Hi there, when including boost\filesystem.hpp I get the following linker error with VC8. error LNK2019: unresolved external symbol "class boost::system::error_category const & __cdecl boost::system::get_generic_category(void)" (?get_generic_category@system@boost@@YAABVerror_category@12@XZ) referenced in function "void __cdecl boost::system::`dynamic initializer for 'generic_category''(void)" (??__Egeneric_category@system@boost@@YAXXZ) Since I'm using autolinking I don't think it's due to a missing library. Anyone? Thanks, Christian

1 0

checking that two types are equal
by e r 20 Jul '08

20 Jul '08

Hi all, what is the command to check that two types, say Type0 and Type1 are equal? I'm looking for something like BOOST_STATIC_ASSERT(...) Thanks.

2 2

Looking for perl interface to boost::serialization...
by Sangeeta Singh 20 Jul '08

20 Jul '08

Hi, I was wondering if a perl interface to the boost serialization library is available some where like the boost::graph cpan perl module. Thanks, Sangeeta

2 1

implicit astar_search weirdness, bug in mutable_queue?
by Christopher Flesher 20 Jul '08

20 Jul '08

Hello, I am trying to use this boost::graph implicit a-star search using boost 1.35.0 but have been having some difficulty. There seems to be a bug in the mutable_queue class where it will try to write to an index that is not valid: index_array[ tmp ] = c.size()-1; It seems someone else had posted an e-mail about this problem about four months ago. I checked CVS for any recent updates but this code hasn't been changed in the past 3 years. Anyways I got around this problem temporarily by adding some code to push indices onto the index_array until the code doesn't crash... not sure what the side effects of this would be though. Anyways after getting that working I find that my cost heuristic is never getting called! My implicit search basically goes off in the wrong direction to infinity. Anyways I've attached my test case. I've tried 2-3 different code samples that try to do this and none of them have worked so far! Chris

1 0

[multi_array] more on assignability
by Dr Johann A. Briffa 20 Jul '08

20 Jul '08

I realize this topic has been discussed before (I've seen at least a thread in 2006 and another in 2007), but I believe there is more to add. At issue is the decision that multi_array objects cannot be assigned [with the operator=() that is] unless the target is already the same size as the source. Further, there is no documented way to easily resize an existing array to the same size and range as some other array. Let me start with the first issue: others have already said before that making multi_array assignable would be a good thing. Ron disagrees on this point, and from what I gather his principal reason is that multi_array should have the same semantics as multi_array_ref etc. [please do correct me if I'm wrong here, I've only had a day to sort this out]. I am unsure why multi_array_ref etc cannot also have such an interface, but then I haven't used that class enough so my knowledge is very limited there. As far as my needs are concerned, a multi_array that cannot be assigned is next to useless. Allow me to elaborate: I am using multi_array as a replacement to my own (old) 2/3D containers, as recently I've had need for higher dimensions. I have classes containing array members, where the size of the arrays is dynamic through the object lifetime. So my class default constructor uses the array default constructor for its member objects, creating empty arrays. Eventually as my classes get used, the array members are updated accordingly, and get to have some non-zero size. Now the problem is that my classes need to be assignable, and the only way that can be achieved is if the array members are assignable; the alternative is to have to write a specific assignment operator for any of my classes that use arrays, resizing each member of the target before copying. Clearly that would involve a lot of extra effort. The second (admittedly minor) problem is that the only way to resize the target is unclean and the semantics are bound to the dimensionality: target.resize(boost::extents[source.shape()[0]][source.shape()[1]]); etc. I would rather have something like: target.resize(source.extents()); Ron had mentioned in 2007 that the lack of such an interface was an oversight, but there is still no solution. I am bringing up these issues again because: 1) I am not convinced there is any valid reason not to have multi_array assignable. I do not buy the performance issue, as this would only kick in if (a) assignment is within a performance-sensitive path and (b) the target is not already of the same size. Further, at the very least there could be a derived class (somewhat like what I document below) that allows such assignment. 2) It would be nice to have the clean resize interface as part of the multi_array class. As an interim solution for myself, I derived a new class from multi_array that allows assignment and does the resize automagically. I am copying the code below for public benefit. Notes: i) anyone is free to (re)use this code under the same license as multi_array; in particular, Ron, please do feel free to adapt and reuse within a future multi_array as you see fit. ii) while the interface provided is sufficient for the way I use the class, it certainly needs to be adapted to be completely generic. Johann #ifndef __multi_array_h #define __multi_array_h #include "config.h" #include <boost/multi_array.hpp> namespace boost { /*! \brief Assignable version of Boost MultiArray. \author Johann Briffa \par Version Control: - $Revision: 1488 $ - $Date: 2008-07-11 10:35:16 +0100 (Fri, 11 Jul 2008) $ - $Author: jabriffa $ */ template<typename T, std::size_t NumDims> class assignable_multi_array : public multi_array<T,NumDims> { public: /*! \name Constructors / Destructors */ explicit assignable_multi_array() : multi_array<T,NumDims>() {}; explicit assignable_multi_array(const detail::multi_array::extent_gen<NumDims>& ranges) : multi_array<T,NumDims>(ranges) {}; explicit assignable_multi_array(const assignable_multi_array& x) : multi_array<T,NumDims>(dynamic_cast< const multi_array<T,NumDims>& >(x)) {}; // @} /*! \name Assignment */ assignable_multi_array& operator=(const assignable_multi_array& x) { if(!std::equal(this->shape(), this->shape()+this->num_dimensions(), x.shape())) { resize(x.extents()); for(std::size_t i=0; i<NumDims; i++) libbase::trace << "Output Extent " << i << " = " << this->shape()[i] << "\n"; } dynamic_cast< multi_array<T,NumDims>& >(*this) = dynamic_cast< const multi_array<T,NumDims>& >(x); return *this; } // @} /*! \name Informative functions */ /*! \brief Get array extents description Returns an object describing the array extents, in a format suitable for use with resize(). */ detail::multi_array::extent_gen<NumDims> extents() const { typedef typename multi_array<T,NumDims>::extent_range extent_range; detail::multi_array::extent_gen<NumDims> extents_list; for(std::size_t i=0; i<NumDims; i++) { extents_list.ranges_[i] = extent_range(this->index_bases()[i], this->index_bases()[i]+this->shape()[i]); libbase::trace << "Input Extent " << i << " = " << this->shape()[i] << "\n"; } return extents_list; } // @} }; }; // end namespace #endif

4 9

Testing a boost::signal
by Simon Pickles 19 Jul '08

19 Jul '08

Hi, What is the correct syntax to test whether a signal has any slots connected? boost::signal<void(void)> sig; if (sig) DoSomething(); This doesn't compile. Thanks Simon

3 2

Specifying starting vertex in depth_first_search
by Hiren Patel 19 Jul '08

19 Jul '08

I'm trying to use depth_first_search() and I want to specify the start node. However, I can't seem to specify the color map correctly. I'm encountering the same problem as described in one of the archive posts at ( http://groups.google.ca/group/boost-list/browse_thread/thread/679106e227e0e… ). Any small examples available? Regards, ~ Hiren

4 4