December 2005 - Boost-users

Re: [Boost-users] [Multi-index] Sorted range selection
by JOAQUIN LOPEZ MU?Z 29 Dec '05

29 Dec '05

Hello Tamas, ----- Mensaje original ----- De: Tamas Szepes <tamas.szepes(a)ieee.org> Fecha: Miércoles, Diciembre 28, 2005 9:11 pm Asunto: [Boost-users] [Multi-index] Sorted range selection > Hi All, > > I'm learning to use the multi-index library by rewriting an > existing STL > based code. > I'm storing rectangles (X, Y, Width, Height, ...) in a set ordered > by a > composite key on X, Y. > > What is the best way to select rectangles where X has a certain > value and > sort the result by Width? Can I use a sorted indice on Width, or I > must sort > the range? I also thought to create a composite key on X and > Width, but > sorting order is run time determined and I have a dosen other > properties. Your analysis is correct. I'm afraid you have to externally sort the interval returned by range(), pseudocode follows: std::vector<boost::reference_wrapper<const Rectangle> > rng; std::pair< YourCompositeKeyIndex::iterator, YourCompositeKeyIndex::iterator> p= yourcompositekeyindex.range(...); while(p.first!=p.second)rng.push_back(boost::cref(*p.first++)); std::sort(rng.begin(),rng.end(),WidthCompare()); Having a composite key on (X,Width) is also an option, but as you say the number of combinations might be too large, and manipulating those indices with run-time decisions is not trivial. So, my suggestion is to follow the scheme proposed on the pseudocode above. > I'm realy looking for the possibility to mimic SQL's ORDER BY clause. Well, Boost.MultiIndex does not aim to provide complete functionality for relational stuff, as stated explicitly in the rationale at http://boost.org/libs/multi_index/doc/tutorial.html. The idea is to keep Boost.MultiIndex application-agnostic and more focused on the STL conceptual framework. As I see it, this kind of funcionality could be better served by a relational-specific lib built on top of Boost.MultiIndex. There are currently two libs approaching this goal, Arkadiy Vertleyb's RTL and Calum Grant's RML (look for those on recent discussions on the Boost list.) Unfortunately, neither of these libs actually uses internally Boost.MultiIndex, but this is another story :) > > Thanks for any help in advance, I'm sorry I can't be more helpful. Thanks for using Boost.MultiIndex and please come back if you have further problems or want to comment on proposals for enhancements of the library. Joaquín M López Muñoz Telefónica, Investigación y Desarrollo

1 0

BGL question: non local exit by throwing an exception
by Andreas Fabri 29 Dec '05

29 Dec '05

Hi, I am wondering, if it is good style to throw an exception in the middle of a BGL algorithm, for example during BFS when I have found what I want. Or are there nicer ways to achive the same goal? Thanks in advance, andreas

2 1

[BGL] Mixing bundled properties and property lists
by Nicola Vitacolonna 29 Dec '05

29 Dec '05

Hi, consider the following (excerpt from a) class wrapper around an adjacency_list: using namespace boost; template <typename VertexProperty = no_property> class MyGraphWrapper { private: typedef adjacency_list<vecS, listS, bidirectionalS, property<vertex_index_t, int, VertexProperty> > GraphImpl; // ... }; I want to store some property (in this case, a vertex index) inside the graph implementation, but at the same time I want to allow the client to specify further properties via the template parameter. May the client use the bundled property mechanism, e.g. struct Label { std::string lab; }; MyGraphWrapper<Label> g; My test compiled and it seems to run fine. Is it really possible/safe to mix bundle properties with property lists in this way? Or, is there a way to avoid property lists? Nicola

3 2

Re: [Boost-users] [boost] bug in signals library? - segmentation fault (onlylinked) -> exit(11) called
by Oliver.Kowalke＠infineon.com 29 Dec '05

29 Dec '05

Hi, Strange - the segmentation fault disappears if I link with the debug-version of libboost_signals (-lboost_signals-gcc-d-1_33_1)! It seams to me that some assert statements contain required code (removed in release version if signals)?! >> I get an segmentation fault if I try to link against >> libboost_signals-gcc-1_33_1 or libboost_signals-gcc-mt-1_33_1! >> The application consist only of an empty main function and the binary is >> linked against the signals library. >> I've appended an output (tusc.out) of tusc (trace unix system calls) >> which indicates, that exit(11) is called and so a core dump created. > >Can you run the program under gdb debugger, and when hittin segfault >produce >backtrace with the "backtrace" command. That will be more usefull, in any >case, then syscall trace. I'm not familiar with gdb! What I did (hopefully correct): [src]$ /opt/langtools/bin/gdb32 test HP gdb 5.2 for PA-RISC 1.1 or 2.0 (narrow), HP-UX 11.00 and target hppa1.1-hp-hpux11.00. Copyright 1986 - 2001 Free Software Foundation, Inc. Hewlett-Packard Wildebeest 5.2 (based on GDB) is covered by the GNU General Public License. Type "show copying" to see the conditions to change it and/or distribute copies. Type "show warranty" for warranty/support. .. (gdb) run Starting program: /eda-home/kowalke/Projects/test/debug/src/test Program received signal SIGSEGV, Segmentation fault si_code: 0 - SEGV_UNKNOWN - Unknown Error. 0xc0017cd8 in <unknown_procedure> () (gdb) backtrace #0 0xc0017cd8 in <unknown_procedure> () warning: Attempting to unwind past bad PC 0xc0017cd8 #1 0xc0027400 in <unknown_procedure> () #2 0xc0027400 in <unknown_procedure> () Error accessing memory address 0x3e7: Bad address.

2 1

boost: debug step into extremely slow at VC8
by David Xiao 29 Dec '05

29 Dec '05

Hi Guys: In my computer, if compile with VC8 with boost library, the debug stepping speed is extremely slow! I have to wait 3-5 second for one step! With VC7.1 everything is ok. Who has the same problem with me ? Rgds, David Xiao

2 1

[Multi-index] Sorted range selection
by Tamas Szepes 28 Dec '05

28 Dec '05

Hi All, I'm learning to use the multi-index library by rewriting an existing STL based code. I'm storing rectangles (X, Y, Width, Height, ...) in a set ordered by a composite key on X, Y. What is the best way to select rectangles where X has a certain value and sort the result by Width? Can I use a sorted indice on Width, or I must sort the range? I also thought to create a composite key on X and Width, but sorting order is run time determined and I have a dosen other properties. I'm realy looking for the possibility to mimic SQL's ORDER BY clause. Thanks for any help in advance, Tamas

1 0

cannot open file 'libboost_date_time-vc71-sgd-1_33_1.lib'
by Subrahmanyam (Arun) 28 Dec '05

28 Dec '05

Hi, I am new to boost libraries. I have viewed some samples in http://www.boost.org/doc/html/date_time/examples.html#date_time.examples.da… in which i tried to build "Days Alive" example. The application reports: boost1 fatal error LNK1104: cannot open file 'libboost_date_time-vc71-sgd-1_33_1.lib' I have searched this in my list of libraries/files installed, but failed to find the same. Can any one please help how to overcome this error. Thanks, Arun.

2 1

Re: [Boost-users] [serialization]Mal-formed XML output
by Merrill Cornish 28 Dec '05

28 Dec '05

Robert, I changed the archive output from XML to text. The example I sent you prevously ran to completion without error. However, the text format doesn't have enough syntax detail to make a syntax error obvious. The text archive it generated was 22 serialization::archive 3 0 1 1 1 0 0 0 0 0 1 -1 4 abcd 0 0 0 0 8 20051227 0 0 0 18 41 40 0 I had also written a serialization test for another class. Its serialize() function looks like this private: friend class boost::serialization::access; template<class Archive> void serialize(Archive& ar, const unsigned int version) { ar & BOOST_SERIALIZATION_NVP(mNode); ar & BOOST_SERIALIZATION_NVP(mInputIndex); ar & BOOST_SERIALIZATION_NVP(mIsInbound); }//serialize() where, again, mNode is a shared_ptr, mInputIndex is size_t, and mIsInbound is a bool. The code to generate the archive and to then read is back is as I showed you before. The archive generated was 22 serialization::archive 3 0 1 1 1 0 0 0 1 Is that enough of an archive to record the serialize() function above? The serialization of the mNode data member should have included the value 1142 in it, but I see nothing that would represent that. Merrill

2 1

multiple definitions of symbol with Boost.Test
by Ross Boylan 27 Dec '05

27 Dec '05

When I try to link a test suite I get this: ld: multiple definitions of symbol boost::test_tools::(anonymous namespace)::check_is_close mspathCEntry_test.o definition of boost::test_tools::(anonymous namespace)::check_is_close in section (__DATA,__data) ModelBuilder_test.o definition of boost::test_tools::(anonymous namespace)::check_is_close in section (__DATA,__data) ld: multiple definitions of symbol boost::test_tools::(anonymous namespace)::check_is_small mspathCEntry_test.o definition of boost::test_tools::(anonymous namespace)::check_is_small in section (__DATA,__data) ModelBuilder_test.o definition of boost::test_tools::(anonymous namespace)::check_is_small in section (__DATA,__data) LinearProduct_test.o definition of boost::test_tools::(anonymous namespace)::check_is_close in section (__DATA,__data) LinearProduct_test.o definition of boost::test_tools::(anonymous namespace)::check_is_small in section (__DATA,__data) ..... This is with boost 1.33 on Darwin/OS-X. The same code was building OK with the 1.31 libraries (and 1.32 on linux, I think). I'm using libtool. Here's the makefile output before those errors (I've deleted some of the files on the command line to save space): ../libtool --mode=link --tag=CXX g++ -o test1 -g AbstractTimeStepsGenerator.o Coefficients.o -lboost_unit_test_framework -L/usr/local/lib/boost-1_33_1 g++ -o test1 -g AbstractTimeStepsGenerator.o Coefficients.o Covariates.o -Wl,-bind_\ at_load -lboost_unit_test_framework -L/usr/local/lib/boost-1_33_1 There are several copies of various flavors of the library lying around, but I don't think any are on the default library paths. So I think I'm only getting the one in /usr/local/lib/boost-1_33_1. It looks as if each invocation of, e.g., BOOST_CHECK_CLOSE, is generating a defiition for the symbols on which I'm getting duplicates. But why would it do that? Any suggestions, or ideas what might be going wrong? Thanks. Ross Boylan Here are some possibly relevant details. Example code: -------------------------------------------------- #include <cstdlib> #include <fstream> #include <iostream> #include <iterator> #include <vector> #include <boost/test/auto_unit_test.hpp> #include <boost/test/floating_point_comparison.hpp> using namespace boost::unit_test_framework; #include "basic.h" #include "Data.h" #include "Model.h" #include "mspath.h" #include "main_test.h" using namespace mspath; using namespace std; static Data * readInputDataFile() { //.... } BOOST_AUTO_UNIT_TEST(CEntry){ if (gOptions.input1() == 0) { BOOST_ERROR("Skipping mspathCEntry test. You need to specify an input file."); return; }; cout.imbue(locale("")); Data* pdata = readInputDataFile(); // ..... BOOST_CHECK_CLOSE(results[0], 2835.43092, tol); // .... } ------------------------------ Maybe I'm messing things up by rolling my own outer level code for the auto-unit tests: ------------------------------------- #include "main_test.h" #include <boost/test/auto_unit_test.hpp> #include "basic.h" // g for global // It's ugly, but it works. mspath::TestOptions mspath::gOptions; /* The following function is just a slight modification to the usual product of #define BOOST_AUTO_TEST_MAIN so that I can capture command line arguments. See the top of the file for the meaning of those arguments. */ #if defined( BOOST_1_31) // code for 1.31 and 1.32 snipped for brevity #elif defined(BOOST_1_33) boost::unit_test::test_suite* init_unit_test_suite( int argc, char* argv[] ) { mspath::gOptions.init(argc, argv); boost::unit_test::auto_unit_test_suite_t* master_test_suite = boost::unit_test::auto_unit_test_suite(); boost::unit_test::auto_unit_test_suite_t* master_test_suite = boost::unit_test::auto_unit_test_suite(); boost::unit_test::const_string new_name = boost::unit_test::const_string( "BOOST_AUTO_TEST_MAIN" ); if( !new_name.is_empty() ) boost::unit_test::assign_op( master_test_suite->p_name.value, new_name, 0 ); master_test_suite->argc = argc; master_test_suite->argv = argv; return master_test_suite; } #else #error I do not know how to setup for this version of boost BOOST_VERSION #endif // etc --------------------------------------------------------------

2 2

Re: [Boost-users] [serialization]Mal-formed XML output
by Merrill Cornish 27 Dec '05

27 Dec '05

Robert, The immediate object being serialized was a shared_ptr and one of the data members being serialized was also a shared_ptr, although it doesn't appear that the process got to the point of serializing that data member. Removing the data member from serialization did nothing. The same faulty output was created. Removing the shared_ptr data member AND replacing the testObj_out shared_ptr with an ordinary pointer produced a different mal-formed XML output: <?xml version="1.0" encoding="UTF-8" standalone="yes" ?> <!DOCTYPE boost_serialization> <boost_serialization signature="serialization::archive" version="3"> <testObj_out class_id="0" tracking_level="1" version="0" object_id="_0"</boost_serialization> By the way, in your reply, you mentioned the missing </px element. However, the original error (and this one too) is more primitive: the initial <px element (or, in this case, the <testObj_out element) have no closing >. Therefore, the serialization process never got past outputting the FULL introductory element. I doubt that the processing ever got to the point of serializing any of the data members. Merrill

1 0