March 2006 - Boost-users - lists.boost.org

[Variant] Time complexity of apply_visitor
by Scott Meyers 25 May '06

25 May '06

If I invoke apply_visitor, what is the time complexity wrt the number of types that may be in the variant? The doc doesn't seem to address this. I looked at the source code, but all I can remember before my head exploded due to exposure to the MPL and the PPL was "switch" (which suggests constant time) and "unrolling" which suggests either linear or constant time, depending on what is being unrolled. Does anybody know what the time complexity of apply_visitor is? My real question is whether applying a visitor to a variant is more efficient than a cascading series of calls to get. I know that the following runs in linear time wrt the number of types in the variant, if (T1 *p = get<T1>(&myVariant)) ... else if (T2 *p = get<T2>(&myVariant)) ... else ... else if (Tn *p = get<Tn>(&myVariant)) ... but what about this? apply_visitor(myVisitor, myVariant); Thanks for any enlightenment. Scott

2 1

[Unit-Test] How to test private functions
by Leif Gruenwoldt 05 May '06

05 May '06

Is there a way to test the private functions of a class without moving them into public scope during test? What are some strategies for this? I was hoping to add "friend class boost::unit_test::test_suite;" to my class but that didn't work. -- Leif Gruenwoldt

8 11

Re: [Boost-users] [program_options] multitoken
by Sohail Somani 24 Apr '06

24 Apr '06

> -----Original Message----- > Edward Diener > Why not have the parameter as: > > "--first=11,12" > > instead of trying to create more than one parameter with the same > parameter name. > > Personally because of the type of solution I suggested, which is > probably the way most uses of such a multi-value value > parameter would > look like, with possibly a different separator than a comma, I do not > think a parameter library should have to cater to the same parameter > name more than once on the command line. What about parameters like -I for gcc? Utilities can add -I's to the command line. It would be an added headache to synchronize all these tools to eventually come up with a -I arg1,arg2,...

6 6

[program_options] additional parser
by Alexander Kondratyuk 24 Apr '06

24 Apr '06

Have a problem with using additional parser feature: I want to allow user to use double dashes for short options, so user will use "-h" or "--h". But using additional parser the code throws an error about syntax. It seems there is a bug in additional parser wrapper - in case then my option has no parameters defined and declared in description it adds an empty value to the vector of values when additional paser returns pair of strings. Perhaps there is another way to get what I want, please let me know how to do it? Here is an example that illustrates the problem: #define BOOST_WHATEVER_DYN_LINK #define BOOST_LIB_DIAGNOSTIC #include <boost/program_options/variables_map.hpp> #include <boost/program_options/options_description.hpp> #include <boost/program_options/parsers.hpp> #include <boost/program_options/detail/utf8_codecvt_facet.hpp> #include <boost/token_iterator.hpp> #include <boost/shared_ptr.hpp> using namespace boost::program_options; #include <string> #include <vector> using namespace std; pair<string, string> parserDashDashShort(const string& s) { string str = s.substr( 0, s.find('=') ); if( str.length()==3 && str[0]=='-' && str[1]=='-' && str[2]<='z' && str[2]>='a') { string val = s.substr(str.length()); return make_pair( str.substr(1), val ); } return make_pair(string(), string()); } int _tmain(int , _TCHAR* argv[]) { int style = command_line_style::allow_short| command_line_style::allow_dash_for_short | command_line_style::short_allow_adjacent; options_description desc; desc.add_options() ("help,h", "help message") ; char* args[]={"program.exe", "--h"}; int argc = sizeof( args ) / sizeof( args[0] ); parse_command_line<char>( argc, (char**)args, desc, style, parserDashDashShort ); return 0; }

2 1

link issues with BOOST_AUTO_TEST_CASE() facility
by Pierre-Jules Tremblay 12 Apr '06

12 Apr '06

I would like to use the BOOST_AUTO_TEST_CASE() facility in a static library. Obviously this library does not have a main() so it needs to be linked with a test executable which may run test cases found in one or more of those static libraries. Unfortunately, the tests do not make it past the linker. The executable links fine but runs none of the tests defined in the libraries. Could it be that BOOST_AUTO_TEST_CASE() works via some global which gets stripped when the static lib is created? Do you have a suggestion for a workaround? The idea is to keep some of the lowest-level unit test code in the same translation unit as the class implementation. Thanks, pj

2 3

Boost.Test - memory allocated during test case run prevents memory leak detection?
by Pierre-Jules Tremblay 12 Apr '06

12 Apr '06

We use Boost.Test extensively as our unit test framework. Recently, we introduced some code to check for memory leaks. Each test case is expected to release all the memory it allocates (i.e. fixtures and what not). However, we observe that several of our test cases appear to leak memory when in fact they don't. For instance, a memory leak check when main() exits reveals no leaks, even though leaks were "found" in several test cases. Our observation is that the apparent leaks appear and disappear as we modify the test code. Scrutinizing the code has revealed that adding a BOOST_REQUIRE(true) here and there in the problematic test cases can make this problem go away. Is it at all possible that Boost.Test somehow allocates memory during a test case run, say for logging or other purpose, and doesn't release it before the test case ends? Here is the code we use to check for memory leaks after each test case is run: #define DECLARE_TEST_CASE(name) \ void test##name##__(); \ void test##name##() \ { \ mytestlib::TestCaseResetLeakCheck(); \ const uint32 uiMemUsageBeforeTest = MMGetAllocatedBytes(); \ test##name##__(); \ const uint32 uiMemUsageAfterTest = MMGetAllocatedBytes(); \ const uint32 uiMemUsageDiff = uiMemUsageBeforeTest > uiMemUsageAfterTest ? uiMemUsageBeforeTest - uiMemUsageAfterTest : uiMemUsageAfterTest - uiMemUsageBeforeTest; \ if (uiMemUsageDiff) \ { \ if (mytestlib::TestCaseLeakCheck()) \ { \ BOOST_ERROR("Test " #name " has memory leaks (" << uiMemUsageAfterTest << " != " << uiMemUsageBeforeTest << ", " << uiMemUsageDiff << " bytes leaked)"); \ } \ else \ { \ BOOST_WARN_MESSAGE(false,"Test " #name " has memory leaks (" << uiMemUsageAfterTest << " != " << uiMemUsageBeforeTest << ", " << uiMemUsageDiff << " bytes leaked) (marked as expected)"); \ } \ } \ } \ void test##name##__() DECLARE_TEST_CASE( SomeTest ) { // arbitrary test code goes here... }

2 2

[shared_ptr] commit/rolback
by Denis G. Priyomov 11 Apr '06

11 Apr '06

Wouldn't it be nice to extend shared_ptr to support transaction-like semantics? Or is it possible already? Sometimes it's needed to execute code on exit only until we haven't reached some point (of "no return"): { shared_ptr<void> cr_guard(static_cast<void*>(0), bind(rollback_fun, x, y)); [...some critical part which should call rollback_fun on failure only...] //don't call rollback_fun if we get here cr_guard.commit(); [...any code that follows...] } Thanks! Wish you well! -=-=-=-=-=-=-=-=-=- Denis G. Priyomov

4 3

[boost][bgl] kevin-bacon.cpp example
by Harsh Deshmane 10 Apr '06

10 Apr '06

Hello Doug, I am trying to create a custom visitor class, for bfs, which accumulates all predecessors (based on Doug Gregor's suggestion for finding all paths between 2 vertices) I compiled and ran this example from the examples dir. However, the output is not consistent with kevin_bacon.expected in the example dir. Output I get is listed below (top lines) $ a.out William Shatner has a Bacon number of 0 has a Bacon number of 1 Denise Richards has a Bacon number of 0 has a Bacon number of 0 Patrick Stewart has a Bacon number of 0 has a Bacon number of 0 Steve Martin has a Bacon number of 0 Gerard Depardieu has a Bacon number of 0 has a Bacon number of 0 Clint Howard has a Bacon number of 0 Sean Astin has a Bacon number of 0 ..... (basically, all values are zero) except the first. http://www.boost.org/libs/graph/example/kevin-bacon.cpp I am using boost version 1.33.0/x86_64/include/boost-1_33 gcc version 3.2.3 20030502 (Red Hat Linux 3.2.3-52) If possible, can someone please check whether the output you get is the correct one ? regards, -harsh

2 1

[BGL] bug in PLOD algorithm
by Johan Oudinet 10 Apr '06

10 Apr '06

Hi, I'm using Parrallel Boost Graph Library to generate some graphs and compute their pagerank. I found the PLOD algorithm and I'd like using it with bidirected graphs. But, the PLOD algorithm generates self_edges whereas I put allow_self_edges to false. I look at the code in plod_generator.hpp and found that the next(directed_tag) function never uses allow_self_edges boolean ! I think it's a bug... I suggest replace: current.second = x(*gen); with: do { current.second = x(*gen); } while (current.first == current.second && !allow_self_loops); Regards, -- Johan

2 1

[Graph] How to free memory allocated when constructing a graph?
by Olivier Delalleau 10 Apr '06

10 Apr '06

I am constructing a graph exactly the same way as in the spanning tree example here: http://www.boost.org/libs/graph/example/prim-example.cpp i.e. using a constructor of the kind Graph g(edges, edges + sizeof(edges) / sizeof(E), weights, num_nodes); where Graph is some kind of adjacency_list. In my application, edges is rather big (on the order of 1e7 elements). Problem: when this constructor is called, it allocates a huge amount of memory (I see this when I do a 'top' to monitor the memory usage while in gdb). And this memory is not freed when the graph is deleted (when it gets out of the scope, or, if I use a pointer instead, when I delete it explicitely). Is there a way to recover this memory? Any hint on how to make it use less memory? Thank you, Olivier

2 1