November 2008 - Boost-users

Boost.Asio 1.37 - can't compile with VC++ 2003 (version 1.35 is fine)
by lipik 29 Jan '09

29 Jan '09

Hi, I just upgraded my Boost distribution, and have immediately run into a compilation problem: both Intel and VC++ compilers complain that: - 'QueueUserAPC' : is not a member of 'operator``global namespace''' - 'QueueUserAPC': identifier not found, even with argument-dependent lookup 1.35 compiles fine. I checked and this API is available in Winbase.h. Can't figure out why it is not being found - there don't seem to be be any C++ version directives hiding it inside winbase. I am on XP, and _WIN32_WINNT is set to 0x0501, so this is not the culprit either. Pre-compiled headers are turned off, and I have defined WIN32_LEAN_AND_MEAN (to avoid winsock conflict). Any help greatly appreciated. Thanks Amit

2 1

Updating my local boost.
by Robert Jones 06 Jan '09

06 Jan '09

I guess this question is more appropriate to the Subversion maillist, but I'm not subscribed to that, so it might take upto 24 hours to appear there. I'm hoping someone here will know the answer! My organisation maintains its boost libraries under its own Subversion repository. This strikes me as a bit odd, but no reason really why not AFAICS. However, the version in our repository is 1.33.1, and I'd like to upgrade to latest 1.36 release. Our network prevents me accessing the Boost subversion repository, so I have to download the tarball and unpack. Then what do I do? I basically want to overwrite our existing boost code with the newer code, but maintain our local edit history so we can see what's changed locally between our old boost code and new. Hence there are lots of additions, changes and deletions, all of which I want to be able see (that's see the differences between our old code and new code, not all the individual boost commits). How can I do this? Thanks - Rob.

5 5

[BGL] Bundled properties and property maps
by Alex Brooks 05 Jan '09

05 Jan '09

Hi, I've used BGL before for representing graphs, but this is my first time with an algorithm. I've been stuck for many hours trying to define a property map with bundled properties, I'd really appreciate some help. I'm using gcc4.3, boost 1.34. An example program and the compiler error are: ------------------------------------------------------------- #include <boost/graph/graph_traits.hpp> #include <boost/graph/adjacency_list.hpp> #include <boost/graph/astar_search.hpp> using namespace std; class Point { public: double p_[2]; }; struct EdgeProp { double cost; }; typedef boost::adjacency_list<boost::vecS, boost::vecS, boost::undirectedS, Point*, EdgeProp> Graph; typedef boost::graph_traits<Graph>::vertex_descriptor VertexRef; class AStarHeuristic : public boost::astar_heuristic<Graph, double> { public: double operator()( VertexRef vertex ) const {return 0.0;} }; // exception for termination struct FoundGoal {}; // // A* Visitor that terminates when we reach the goal // class AStarVisitor : public boost::default_astar_visitor { public: AStarVisitor( const VertexRef &goal ) : goal_(goal) {} template <class Graph> void examine_vertex( VertexRef vertex, Graph &g ) { if( vertex == goal_ ) throw FoundGoal(); } private: VertexRef goal_; }; bool searchThroughGraph( const Graph &graph, VertexRef start, VertexRef goal, std::vector<VertexRef> &shortestPath ) { AStarHeuristic aStarHeuristic; AStarVisitor aStarVisitor( goal ); typedef boost::property_map<Graph, double EdgeProp::*>::type EdgeWeightMap; EdgeWeightMap edgeWeightMap = get(&EdgeProp::cost, graph); std::vector<VertexRef> predecessors(num_vertices(graph)); try { astar_search( graph, start, aStarHeuristic, boost::weight_map(edgeWeightMap). predecessor_map(&predecessors[0]). visitor(aStarVisitor) ); } catch ( const FoundGoal& ) { shortestPath.clear(); for ( VertexRef vertex = goal;; vertex = predecessors[vertex] ) { shortestPath.push_back( vertex ); if ( predecessors[vertex]==vertex ) break; } std::reverse( shortestPath.begin(), shortestPath.end() ); return true; } // If we get to here, we couldn't reach the goal return false; } int main() { Graph graph; std::vector<VertexRef> path; searchThroughGraph(graph,0,1,path); return 0; } ------------------------------------------------------------- /home/a.brooks/shh/src/libs/prm/test/stupid.cpp: In function ‘bool searchThroughGraph(const Graph&, VertexRef, VertexRef, std::vector<unsigned int, std::allocator<unsigned int> >&)’: /home/a.brooks/shh/src/libs/prm/test/stupid.cpp:64: error: conversion from ‘boost::bundle_property_map<const boost::adjacency_list<boost::vecS, boost::vecS, boost::undirectedS, Point*, EdgeProp, boost::no_property, boost::listS>, boost::detail::edge_desc_impl<boost::undirected_tag, unsigned int>, EdgeProp, const double>’ to non-scalar type ‘searchThroughGraph(const Graph&, VertexRef, VertexRef, std::vector<unsigned int, std::allocator<unsigned int> >&)::EdgeWeightMap’ requested ------------------------------------------------------------- Does anyone know what might be wrong? More generally, I'm a bit mystefied by the "<Graph, double EdgeProp::*>::type" syntax -- can anyone give me any pointers about what this syntax means, or where I can look it up? (googling for colon-colon-star doesn't help much). I also tried pattern-matching from the documentation page about bundled properties (the weight_map(get(&Highway::miles, map)) example) but couldn't get this to compile either. Is this example correct? Thanks very much in advance, Alex

4 19

Idea: Chained comparison
by Michiel Helvensteijn 31 Dec '08

31 Dec '08

I just came up with this handy class. I was wondering if it might have a place in Boost somewhere. Perhaps it already exists, and I didn't find it? It creates a syntax for chained comparison expressions. Ideally you'd want to be able to do the following in C++: if (0 <= i < N <= INT_MAX) f(); But this doesn't work as expected, because it would actually be evaluated like this: if (((0 <= i) < N) <= INT_MAX) f(); You would actually need to use the following code: if (0 <= i && i < N && N <= INT_MAX) f(); The alternative I propose is this: if (c(0) <= i < N <= INT_MAX) f(); It has the advantage of being shorter, of being closer to the usual mathematical notation, and of evaluating i and N only once. This is useful if they have side-effects. It requires the following class and function (which I'm sure could be improved). What do you think? --------------------------------------------------------------------- template <class T> class Comparison { public: Comparison(T value, bool truth = true) : _last(value), _truth(truth) {} template<class U> Comparison operator==(U val) { return Comparison(val, _truth && _last == val); } template<class U> Comparison operator!=(U val) { return Comparison(val, _truth && _last != val); } template<class U> Comparison operator<(U val) { return Comparison(val, _truth && _last < val); } template<class U> Comparison operator>(U val) { return Comparison(val, _truth && _last > val); } template<class U> Comparison operator<=(U val) { return Comparison(val, _truth && _last <= val); } template<class U> Comparison operator>=(U val) { return Comparison(val, _truth && _last >= val); } operator bool() { return _truth; } private: T _last; bool _truth; }; template <class T> Comparison<T> c(T val) { return Comparison<T>(val); } --------------------------------------------------------------------- -- Michiel Helvensteijn

2 1

boost test + MFC
by Andrew Wan 19 Dec '08

19 Dec '08

Does anyone know how to use boost test library with MFC programs? All the examples & documentation only covers CONSOLE programs. Would be good if there was an MFC example.

2 3

tweaking the review process (was: signals2 review results)
by Stjepan Rajko 16 Dec '08

16 Dec '08

On Thu, Nov 20, 2008 at 11:40 PM, vicente.botet <vicente.botet(a)wanadoo.fr> wrote: > > I would like to make some suggestion to improve the review management: > Thank you for starting this discussion. > * I had notice only at the end of the the review that the review tokes place on two mailing lists (devel ans user). Maybe this is usaual on Boost reviews but I was not aware. It will be more transaparent if all the reviews were posted to the same mailing list, maybe a specific one should be created. Yes, this is a bit of an issue. I mentioned that both lists are used in my post that opened the review (in my notes to first-time reviewers) to try to give people a heads-up (but... I often get long-winded and I'm sure it's easy to miss parts of my posts :-)). Having a dedicated, or at least recommended mailing list (either dev or user) might be a good thing. The only problem I see regarding a completely separate mailing list is that it requires active effort to sign-up for and follow. I think that for many people (even those that are genuinely interested in the library), the fact that the review is happening at time X-Y very easily slips off the radar. Having the e-mails constantly appear in the common lists also serves to remind everyone of the ongoing review, and I suspect it leads to many valuable impromptu comments and even full reviews from people that perhaps weren't originally planning on participating in the review, but end up following the review because it happens on the list (I recall being in this situation myself several times). > > * From the 5 committing reviewers making this review possible only 3 made a review and 2 of them late. I'm wondering if this new rule must be preserved as the review can be accepted without the commiting reviewers review. Trying to use committed reviewers was a new thing for this review (it was proposed by others in http://tinyurl.com/48tdjs) Now that it has been tried out, it would be good to reflect on how things went. For the reviews that were late, the reviewers were very diligently contacting me about their changing time constraints and offered times that would work for them, and I found those times acceptable. Out of the two committed reviewers that did not submit, one also diligently worked with me on setting up an alternative timeline, but in the end contacted me saying that a full review would not be possible after all, and offered a brief report of looking into the library and a brief rationale for acceptance (as the points matched those of other reviewers, I didn't forward it to the list). I was not able to contact the final reviewer after a certain point. I think the major benefit of the committed reviewers is the high probability that they actually will submit a review. Things always happen, and boost contributions are all volunteer time, but I found that overall the committed reviewers *very responsibly* succeeded in keeping their commitment as well as they could. Were they not committed reviewers, I suspect many of them would have considered giving up. In addition, the extensions that the committed reviewers negotiated gave other reviewers a chance to contribute. So, I feel that committed reviews are a good way of reasonably making sure that a certain number of reviews will be submitted. I'm not sure that the library acceptance / rejection should be influenced by the number of committed reviews (reviews from others are a perfectly good substitute). While we're on this topic... The one thing that the number of committed reviewers should definitely influence is whether the review happens in the first place. We were fortunate to get 5 people sign up, which was the only threshold proposed, so we followed through with the review. I was wondering what to do if the threshold wasn't crossed, and I think what I would have proposed is some sort of a holding pattern, where the review schedule page is updated to say something like "signals2 - reviewers needed, contact review manager to sign up as a reviewer". Once the threshold is crossed, the reviewers, manager and author can schedule a time. Another thing... many of the committed reviewers reported that they were first time reviewers (and the reviews they submitted were impressively detailed and very valuable). Personally, I was thrilled by this, since getting new participation in boost is critical to sustaining its quality . Perhaps some combination of allowing reviewers to commit beforehand, and explicitly encouraging reviews focused on the user perspective (another suggestion from the http://tinyurl.com/48tdjs thread), helped in getting first-time reviewers (perhaps the reviewers can comment on this?). > > * There were some reviews that came into this list by the intermediation of the review manager with a delay between the posting of the reviwer and forward from the RM. One negative review posted the 4th and reveived in this list the 11th other positive posted the 2nd and received in this list 3rd. I think that the review manager should not encourage the reviewers to send the reviews to himself. This will avoid this kind of delays. So I purpose that only the reviews sent to this single mailing list must be taken in account. Yes, this particular delay is entirely my fault. I can't explain why I didn't notice this email until so late (which is the point where I contacted the poster notifying him I would like to forward the mail, and forwarded it the following morning). Allowing reviews to be sent to the review manager is straight from http://www.boost.org/community/reviews.html (Introduction paragraph). There are two valid reasons for this, IMO: * the reviewer is only subscribed to one of {boost, boost-user, boost-announce} lists, and would like the reviewer to be forwarded to both boost and boost-user lists (if there was a dedicated review list, like you suggest, that was also open to all posters, then this reason would go away) * the reviewer would like to remain anonymous to the list (granted, this could also be accomplished by sending from an anonymous e-mail). If this stays as it is, the RM should be more diligent about monitoring her personal mail, which I apparently wasn't. > > * Even if the review was over the 10th there were 2 accepting reviews comming from the commiting reviewers just some hours before the review result annoncement 19th. I think that the review manager must state clearly when the review is over and do not accept any review after. This do not means that the RMcannot change this date, but annnce it clearly. I tried to keep my timeline as transparent as possible. In my review-closing email, I asked people to let me know if they were considering writing a review. Those that contacted me were informed of the timeline of pending reviews. When the last promised review was submitted, I set a hard deadline. Given the uncertainty in people's schedules, I decided to approach this in a flexible way. That was just my personal preference - I'm sure setting a hard deadline earlier would have been a fine choice as well. > > I hope this will help to improve future reviews, > Thanks again for starting this discussion and for your suggestions - I also hope good improvements will come out of it. Stjepan

3 2

thread_group / removal of items
by Christian Ost 13 Dec '08

13 Dec '08

Hi, I have a question concerning the removal from entries of a thread_group. Consider the following code (i am working on a small threaded server): in TCPServer.h: boost::thread_group threads in TCPServer.cpp: int TCPServer::loop(int listen_fd) { for (;;) { [connect client] boost::thread* thrd = new boost::thread(boost::bind(&TCPServer::readMessages, this, rfd)); threads.add_thread(thrd); } } } int TCPServer::readMessages(int rfd) { boost::this_thread::at_thread_exit(boost::bind(&TCPServer::exitThread, this)); for(;;) { [read message] if ([buffer is empty]) { [close connection] return(0); } } return 1; } void TCPServer::exitThread() { cout << "thread exited" << endl; } Now I would like to remove the exited thread from the thread_group, however to do this I would need its thread* object - but the maximum of information I can get from boost::this_thread is a boost::thread::id and I haven't found a possibility to get a thread by its thread::id so far. How can/should I solve this issue? I would be quite happy about any advice... Thank you very much! Christian

3 2

[BGL] passing a graph from one function to another
by r89 10 Dec '08

10 Dec '08

Hi all, I have a graph defined as follows (the same as in the Dijkstra shortest path example code): typedef adjacency_list < listS, vecS, directedS, no_property, property < edge_weight_t, float > > graph_t; The following works fine: graph_t myGraph(array, edge_array + num_arcs, weights, num_nodes); However, now I want to do the above line inside a function (say, fun_A), and I want another function (say, fun_B) to have access to "myGraph". How should I do that? I wish I could do something like this: graph_t myGraph; fun_A(graph_t &myGraph) { myGraph(array, edge_array + num_arcs, weights, num_nodes); } fun_B(graph_t &myGraph) { // using myGraph to run some graph algorithm... } but it doesn't work. Any suggestions? Thanks! -- View this message in context: http://www.nabble.com/-BGL--passing-a-graph-from-one-function-to-another-tp… Sent from the Boost - Users mailing list archive at Nabble.com.

5 12

[Review] UUID library (mini-)review starts today, November 23rd
by Hartmut Kaiser 10 Dec '08

10 Dec '08

Hi all, The mini-review of Andy Tompkins UUID library starts today, November 23rd 2008, and will end on November 30th. I really hope to see your vote and your participation in the discussions on the Boost mailing lists! The library can be downloaded from the vault here (it's the file uuid_v13.zip in the topmost directory): http://tinyurl.com/6xx28b The initial review of the UUID library ended with a provisional acceptance (read here: http://article.gmane.org/gmane.comp.lib.boost.user/27774/) This mini review is meant to allow for a final decision after looking at the changed parts of the library. Here is a list of things fixed and changed: - fixed the licensing issues revealed by the initial review - fixed the security vulnerability revealed by the initial review - renamed to uuid, moved all classes, functions, etc. to namespace boost::uuids - implemented sha1 hash function (thus no license problem) - new class basic_uuid_generator to create random-based uuids. The random number generator is no longer hard coded. It can use any random number generator, default is boost::mt19937 - implemented a good seed function for random number generators - all functions are now reentrant, all classes are as thread safe as an int and the library is no longer dependent on Boost.Thread --------------------------------------------------- Please always state in your review, whether you think the library should be accepted as a Boost library! Additionally please consider giving feedback on the following general topics: - What is your evaluation of the design? - What is your evaluation of the implementation? - What is your evaluation of the documentation? - What is your evaluation of the potential usefulness of the library? - Did you try to use the library? With what compiler? Did you have any problems? - How much effort did you put into your evaluation? A glance? A quick reading? In-depth study? - Are you knowledgeable about the problem domain? Regards Hartmut Review Manager

14 39

[random]how to initialise a rng for use in the initialiser list?
by Moritz Beber 09 Dec '08

09 Dec '08

Hello again, sorry for the many questions to this list right now, I'm just beginning to use boost. This time I have some issues regarding the boost.random usage. I have used it just fine for many examples but there's one tricky bit that I was wondering whether anyone of you knows a solution for. I use a class which keeps the random number generator around as a member variable; this class also has a const matrix. Now in the constructor of the class I do something like: my_class::my_class(args) : rng(seed), matrix(initialise()), ... { ... } Of course, I want to use the rng to initialise the matrix and that's where things fail. I guess, I could use a different rng just to fill the matrix, or make it non-const and fill it in the constructor body. But does anyone know a neat trick to do this in the way proposed? Thank you for any answers, Moritz

4 5