February 2006 - Boost-users - lists.preview.boost.org

Using boost bind library to loop thru a list
by Meryl Silverburgh 16 Feb '06

16 Feb '06

I have a int vector, and I would like to loop thru the vector and call a function which takes an int (in this case, I want to print). void A::aFunction() { vector<int> aVector; aVector.push_back(1); aVector.push_back(2); aVector.push_back(3); aVector.push_back(4); // want to loop thru the list and call A::print(int i)? } void A::print(int i) { } how can I do that using boost bind library?

7 10

hardware bit set, clr, tst templates
by Perry Smith 16 Feb '06

16 Feb '06

Before I jump off and re-invent the wheel -- is there a boost library to help with manipulating hardware registers? What I'm hoping to come up with (somehow) is something like: class dev_registers { reg<uint32_t> reg0; bit<reg0, 0> b0; bit<reg0, 1> b1; reg<uint32_t> reg1; bit<reg1, 0> b3; bit<reg1, 1> b4; }; (The 0 and 1 are bit offsets. Integer masks like 0x0010 would work as well) The result would be a private members of type uint32_t called reg0, and reg1 (actually they can be called anything since they are private). And a set of public methods to set, clear, and test the bits. e.g. dev_registers sample; sample.b0.set(); sample.b3.clr(); It would be nice if the registers had a global clr like: sample.reg0.clr(); There are lots of alternative approaches like: sample.b0.assign(true); // set b0 to true (1) int val = sample.b0(); // test value of b0 The goal has a lot of constraints. The registers would need to be laid out properly without any extra overhead (assuming that I do not bump into normal C padding problems). It would also be nice if everything inline'd so the compiler could optimize things down to simple bit operations. Does boost have anything like that? I looked for "bit" and only found references to bitset. I also looked for "hardware" and "flag" and did not see anything. Thanks, Perry

1 0

[Graph] New user finding dijkstra's isn't returning shortest paths
by Greg Link 16 Feb '06

16 Feb '06

In my first use of the boost library, I'm using the BGL's dijkstra_shortest_paths algorithm to try and (obviously) find the shortest path between two points. I am then iterating along the shortest path to a given node to find the total path weight along that path. I compare this value to the path weight along a manhattan path, and find that the manhattan path is shorter. Rather than claim this is a problem with boost (yet), I'm at first believing it's my interpretation of the library. If I've piqued your interest, feel encouraged to read on. My graph is a representation of a 2-dimensional mesh, so I've got an NxN graph, where the vast majority of nodes have out-edges in all four directions (N,S,E,W), and in-edges in all four directions. I generate the graph with the following: typedef adjacency_list < listS, vecS, undirectedS, no_property, property < edge_weight_t, double > > graph_t; graph_t gGraph(num_nodes); then iterating over each node (i) and adding edges as follows: add_edge(i, easterlydest, easterlyweight, gGraph); finally, I call: dijkstra_shortest_paths(gGraph,sourcenum,predecessor_map(&p [0]).distance_map(&d[0])); to evaluate the shortest paths. (Note that much of this is a cut-and- paste from the Dijkstra_shortest_path example found at http:// www.boost.org/libs/graph/example/dijkstra-example.cpp ) Note also that my understanding of the last argument to dijkstra_shortest_paths is probably my weakest part of understanding the library. I then iterate along a given shortest path by using the following loop structure: while(node != sourcenum) // iterate from dest to source, summing as you go { int u = p[node]; int v = node; graph_traits < graph_t >::edge_descriptor e = edge(u,v,graph).first; double this_weight = get(edge_weight,graph,e); path_length += this_weight; node = p[node]; } This returns a value of 1309.16 as the total path weight. I then attempt to determine the manhattan path weights as well, to compare how much of an improvement the shortest path is to the manhattan paths. I attempt to traverse the manhattan paths using a similar approach, with something similar to the following (irrelevant lines omitted): while(node != destnode) // iterate from source to dest, iterating as you go { int nextnode = get_nextnode(xfirst,node); // gets index of next node in an x-first y-next manhattan path int u = node; int v = nextnode; graph_traits < graph_t >::edge_descriptor e = edge(u,v,graph).first; double this_weight = get(edge_weight,graph,e); xfpath_length += this_weight; node = nextnode; } When this runs, I get a path length of 1299.79, a value more than 0.1% smaller. Either I'm determining the manhattan and shortest path weights incorrectly, or it's not actually finding the shortest path. So I did a bit of debugging (as programmers are wont to do). Tracing the paths themselves, I eventually find that at one point, the shortest path (hereafter abbreviated DSP) chooses to follow an out- edge that doesn't have minimum weight. This is perfectly reasonable, as it's possible (and probable) that this choice made now, will have greater returns later, allowing an overall shorter path. However, it never again returns to even equality with the manhattan path - the manhattan path is shorter along the entire length, to the end. I've checked my corner cases, and the paths are only being summed along relevant edges (no off-by-one errors that I can see, at least). If anyone has any ideas on how I might best debug/change my code, or a means by which I can perform some other test, I'd love to hear it. If anyone needs any more information, the code is by no means confidential (though a bit messy) and I've got both .dot (graphviz) format and an svg representation of the graph available. My development platform is currently Mac OS X 10.4.4 PPC using Xcode, if it matters to anyone. Thanks, and I hope you don't mind the rather lengthy post! - Greg Link ---- 20 70 80 95 106

2 2

Re: [Boost-users] Non-boost c++ novice help
by Matt Davies 16 Feb '06

16 Feb '06

I don't think it is the appropriate place. You need to make it static as well to be in the class: static const uint64_t mask = 0x12345678 Cheers, Matt Davies Rockstar Leeds -----Original Message----- From: boost-users-bounces(a)lists.boost.org [mailto:boost-users-bounces@lists.boost.org] On Behalf Of Perry Smith Sent: 16 February 2006 15:43 To: boost-users(a)lists.boost.org Subject: [Boost-users] Non-boost c++ novice help If this is not an appropriate list for this question, please let me know. The books tell me to prefer const to #defines -- but where do you put them to have the right scope and not take up space in the object? What I'd like to have is: class register { void set_mask() { reg |= mask; } private: uint64_t reg; const uint64_t mask = 0x12345678; }; But the compiler will not let me do this and even if it did,the mask would take up physical space. I don't want mask (or reg) to be visable outside of the class and I want the compiler to use mask as a compile time constant that does not consume space after the compile. I vaguely know about namespace but if I put this inside a namespace, then I have to do: namespace hardware { const mask = 0x12345678; class register { void set_mask() { reg |= mask; } private: uint64_t reg; }; }; // .. define a register hardware::register doddy; and "mask" is not really private -- just obscured. I can still do: int i = hardware::mask; This just feels like I've taken a wrong turn somewhere. How would this be "properly" done? Thanks, Perry _______________________________________________ Boost-users mailing list Boost-users(a)lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users

2 1

[TR1] Where is the TR1 library?
by Scott Meyers 16 Feb '06

16 Feb '06

Google found me a 12 October 2005 posting by Beman Dawes to this newsgroup stating that John Maddock's TR1 library has been accepted by Boost. But if I search for TR1 at http://boost.org/libs/libraries.htm, I don't find the library. Searching more generally yields lots of hits at www.boost.org/regression-logs/ (e.g., http://www.boost.org/regression-logs/cs-win32_metacomm/doc/html/boost_tr1.h…) and this suggests that just because a library has been accepted by Boost doesn't mean it will show up in the list at http://boost.org/libs/libraries.htm. Can somebody please explain to me what are probably the basics of what it means to be "in" Boost, how that interacts with the list of Boost libraries at boost.org, and what the significance of the regression-logs directory is? If this is explained somewhere at Boost, just point me to the page. Thanks, Scott

3 6

boost::test: expected error leads to failure - why?
by Markus Werle 16 Feb '06

16 Feb '06

Hi! When using the boost::test unit-test framework calls like boost::unit_test::test_suite* test = BOOST_TEST_SUITE( "Testing Something" ); test->add(BOOST_TEST_CASE(&test_function_with_expected_error), 1); lead to an error in return code when used together with VC7 as a post-build step Of course I could use runtime argument "--result_code=no", but then how can I make a difference between expected and unexpected errors? It makes sense to me to have expected errors in a testsuite, and especially to declare a test failure when an expected error does not appear. Could you please explain how boost::test supports this without having to write "failure-tests" myself? Markus

2 5

Non-boost c++ novice help
by Perry Smith 16 Feb '06

16 Feb '06

If this is not an appropriate list for this question, please let me know. The books tell me to prefer const to #defines -- but where do you put them to have the right scope and not take up space in the object? What I'd like to have is: class register { void set_mask() { reg |= mask; } private: uint64_t reg; const uint64_t mask = 0x12345678; }; But the compiler will not let me do this and even if it did,the mask would take up physical space. I don't want mask (or reg) to be visable outside of the class and I want the compiler to use mask as a compile time constant that does not consume space after the compile. I vaguely know about namespace but if I put this inside a namespace, then I have to do: namespace hardware { const mask = 0x12345678; class register { void set_mask() { reg |= mask; } private: uint64_t reg; }; }; // .. define a register hardware::register doddy; and "mask" is not really private -- just obscured. I can still do: int i = hardware::mask; This just feels like I've taken a wrong turn somewhere. How would this be "properly" done? Thanks, Perry

3 2

Re: [Boost-users] Non-boost c++ novice help
by RIVASSEAU Jean Noel 16 Feb '06

16 Feb '06

I admit I did not really take the time to read your whole post, but I think you just want to do: static const uint64_t mask = 0x12345678; // note the "static" Also, this is not a Boost related question!! I think you should ask this kind of questions on C++ forums or IRC channels. Regards Jean-Noël -----Message d'origine----- De : boost-users-bounces(a)lists.boost.org [mailto:boost-users-bounces@lists.boost.org] De la part de Perry Smith Envoyé : jeudi 16 février 2006 16:43 À : boost-users(a)lists.boost.org Objet : [Boost-users] Non-boost c++ novice help If this is not an appropriate list for this question, please let me know. The books tell me to prefer const to #defines -- but where do you put them to have the right scope and not take up space in the object? What I'd like to have is: class register { void set_mask() { reg |= mask; } private: uint64_t reg; const uint64_t mask = 0x12345678; }; But the compiler will not let me do this and even if it did,the mask would take up physical space. I don't want mask (or reg) to be visable outside of the class and I want the compiler to use mask as a compile time constant that does not consume space after the compile. I vaguely know about namespace but if I put this inside a namespace, then I have to do: namespace hardware { const mask = 0x12345678; class register { void set_mask() { reg |= mask; } private: uint64_t reg; }; }; // .. define a register hardware::register doddy; and "mask" is not really private -- just obscured. I can still do: int i = hardware::mask; This just feels like I've taken a wrong turn somewhere. How would this be "properly" done? Thanks, Perry _______________________________________________ Boost-users mailing list Boost-users(a)lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users

1 0

Re: [Boost-users] [serialization] Really strange error that crashes the application
by Scott 16 Feb '06

16 Feb '06

> Is any of this implemented in a DLL ? > > the program is haning up because its trying to register a class > that is in the process of being destroyed. Try setting > a break poin on the assertion on line 71 of exended_type_info.cpp > and try to figure out which class is being register. > > Set break point on the destructor of that class. > > This might lead one to the point whereby a destructor > is being called while someting is being serialized You are correct that the class causing the problem is implemented in another DLL and imported to my server .exe. I'm not sure how to go about placing a breakpoint in the extended_type_info.cpp file. Boost was built on another computer and I only have access to the header files and lib/dlls. I did place a breakpoint in the destructor of the class being serialized (in addition to cout messages since the server is currently a console app). The destructor is NOT getting called during the archive process. In addition, I changed the signature of the method to take only a const value of the class (instead of reference) and the error still happens. Here's the problem code: template <typename T, typename Handler> void asyncWrite(const T t, Handler handler) { // Serialize the data first so we know how large it is. std::ostringstream archiveStream; boost::archive::text_oarchive archive(archiveStream); archive << t; // error occurs here, t is not referenced anywhere after this I also get this message in the console: Assertion failed: lookup(eti) == m_self->m_map.end(), file C:\Documents and Sett ings\Scott.SMOORE\Desktop\boost_1_33_1\libs\serialization\build\../src/exten ded_ type_info.cpp, line 71 This error is quite frustrating. As I mentioned, the above code works flawlessly if I bypass reading in a serialized file earlier in the program. Thanks again for any help, Scott

2 1

Boost::MultiIndex dtor problem
by Martin Eigel 16 Feb '06

16 Feb '06

Hi list, I am encountering a problem with MultiIndex which can be resolved by disabling SAFE_MODE. As I actually do not get the problem in the code attached below I'd like to ask for an explanation. The marked line of code (i.e. the second iterator assignment) seems to be responsible for blocking MultiIndex's dtor in safe_mode.hpp: void detach_all_iterators() { for(safe_iterator_base* it=header.next;it;it=it->next)it->cont=0; } Why is this and how can it correctly be avoided? Cheers, Martin (boost 1.33.1, gcc 3.4.5) #define BOOST_MULTI_INDEX_ENABLE_INVARIANT_CHECKING #define BOOST_MULTI_INDEX_ENABLE_SAFE_MODE #include <boost/multi_index_container.hpp> #include <boost/multi_index/member.hpp> #include <boost/multi_index/ordered_index.hpp> using boost::multi_index_container; using namespace boost::multi_index; struct T1{}; struct T2{}; typedef int T; typedef std::pair<T,T> value_type; typedef multi_index_container< value_type, indexed_by< ordered_unique< tag<T1>, member<value_type,T,&value_type::first> >, ordered_unique< tag<T2>, member<value_type,T,&value_type::second> > > > miT; int main(){ miT::const_iterator it; miT map; it = map.get<T1>().end(); map.clear(); it = map.get<T1>().end(); // *** this hangs dtor! return 0; } __________________________________________________ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com

2 2