- Boost-users - lists.boost.org

Tricky lambda usage?
by Jason House 05 Mar '05

05 Mar '05

"In a nutshell" one can do the following: for_each(a.begin(), a.end(), std::cout << _1 << ' '); (where a is something like std::list<int>) Is there any trick to print out something like std::map<std::string, std::list<int> > ?

3 2

[Foreach]VC70 problem
by Martin 04 Mar '05

04 Mar '05

Just wanted to say that VC70 got problems with BOOST_FOREACH. Apparantly an internal compiler bug sometimes generates the wrong code and causes a "Run- Time check failure #2 - Stack around ... was corrupted". It has happened to me twice now in completly different contexts. In both cases there is no problem if I compile with VC71 or if I replace BOOST_FOREACH with a normal for loop. Unfortunatly I can't find any small example which reproduces the problem. In both cases I can see in the debugger that code uses the wrong stack offset for a local variable.

2 1

make_variant_over an mpl::vector: Compilation fails on gcc
by Johannes Dahlström 04 Mar '05

04 Mar '05

I ran into a bizarre problem trying to build a variant from an MPL vector. The compilation of make_variant_over seems to fail when given any vector which a) uses all its template arguments, and b) contains *only* user-defined types; ones with at least one built-in type work fine! The problem occurs at least with gcc 3.3.5 and gcc 3.4.4 20050203 prerelease, on Debian unstable/x86. The following code demonstrates the problem: //////////////////////////////////////////////////////////////////////// #include <boost/variant.hpp> #include <boost/mpl/vector.hpp> #include <cstdio> // a template for generating dummy types template <int i> struct d { }; typedef boost::mpl::vector < d<0>, d<1>, d<2>, d<3>, d<4>, d<5>, d<6>, d<7>, d<8>, d<9> > vector; boost::make_variant_over<vector>::type v; //////////////////////////////////////////////////////////////////////// The same thing happens with vector10 and others, but lists do seem to work correctly, providing a work-around. Swapping any built-in type for any of the d<>'s makes the code above to compile cleanly. I can post the complete compiler error output if you like, but there's rather much of it :p -Johannes

1 0

Any constructor evaluation in compile time?
by Yutaka Leon Suematsu 04 Mar '05

04 Mar '05

Hi everyone, I am new at the excellent Boost library. It is an incredible source of smart techniques and ticks. I would like to ask if there is a way to evaluate, in compile time, the existence of a specific constructor, except from the constructor default and copy constructor (I know the functions has_trivial_constructor and has_trivial_copy from the type_traits library). For example in the following code, I would like to evaluate the existence of both "A(const Config& config)" and " template <class Random> A(const Config& config, Random& random) " constructors. Class A { A(){.} A(const Config& config){.} template <class Random> A(const Config& config, Random random){.} . } Also, I would like to know why and how the has_trivial_constructor works. Thank you in advance for your kindly help. All the best, Yutaka Leon Suematsu.

6 9

uBlas matrix superclass?
by V. C. 04 Mar '05

04 Mar '05

I am considering using the uBlas library. After a couple of days of researching all available matrix packages for C/C++, I have come to the conclusion that it is the best one in terms of support, ease-of-use, performance(relatively speaking) and documentation. >From what I have been able to gather, I can't figure out whether there is any superclass to all the matrices.. I want to be able to have a list of matrices, some of which may be sparse and some dense... What type should that list be? Or is that impossible (at first glance it appears that inheritance is not really supported for efficiency reasons, hence the widespread templatization)... Thanks in advance for any help.

1 0

[Random] Documentation enhancement
by Richard Jennings 04 Mar '05

04 Mar '05

I would like to suggest an enhancement to the documentation for the random library. The documentation for the number generators advises not to construct generators frequently for various reasons. On some platforms there could also be an issue with memory use where the generators are created on the stack. I have used the suggested mt19937 generator that is about 5000 bytes in size on my system. If I create two of these on the stack then I get a crash. I followed the "very quick start" example that creates the generator on the stack and has a variate_generator like this: boost::mt19937 rng; boost::uniform_int<> six(1,6) boost::variate_generator<boost::mt19937, boost::uniform_int<> > die(rng, six); Thus, each instance of variate_generator has a copy of the generator too. The random_demo.cpp works just fine because it uses an alias for the generator: boost::variate_generator<base_generator_type&, boost::uniform_real<> > thus multiple copies of the variate_generator on the stack are fine. I guess what this comes down to is a warning about creating multiple instances on the stack of generators and/or variate_generators where the engine template parameter to variate_generator is a engine rather than a pointer or reference to an engine. Richard

1 0

Program options: terminating parsing.
by The Grumpiest Man You Know 04 Mar '05

04 Mar '05

As context I would like to say that read "The C++ Programming Language" first edition and thought, "That sounds interesting." and then didn't think of it again 'til about a a couple of months ago, when I read Accelerated C++ on the recommendation of a friend and then updated my original copy to the third edition. So I have NO useful C++ experience, please point out even the most obvious things, I really could be that dumb. I have started a non trivial project that I hope will cement some of the things I've read. And that's where boost comes in. I had some success with using boost_filesystem, so I think I am able to extract basic usage out of the documentation but I've failed to find what I'm looking for in program_options. I would really like to parse a line that has the the form: word [options]... nother option-like-non-option-word... so for instance -F after "word" indicates a legitimate value but -F after "nother" shouldn't set or change the value and indeed might not even have a "value" I've been considering producing a class defining operator() to be used as an additional parser that would recognize the second bare word and make all options thereafter into some list for a dummy option, like the file names in the example. However there is mention of allowing multi word options being "not enabled by default" and I thought that the style argument might be the key, but what I've managed to work out from reading a tarball I downloaded of 1_32_0 seems to use style for changing how single words are interpreted. So before I go crazy trying to implement something that sounds easy to say but beyond my ken, is there a method I'm overlooking for terminating the parsing of the command line before the words are all consumed? Thanks for your time, I'll try to be more brief in future. :) -- Blue Skies

3 6

Problems using boost::filesystem
by steve ahlgren 03 Mar '05

03 Mar '05

Hello : I have a question about boost::filesystem::path. My frustration level is growing quickly because I'm getting exceptions when constructing paths such as these : boostfs::path searchPath( "d:/temp", boostfs::native ); boostfs::path searchPath( "d:/temp" ); boostfs::path searchPath( "d:\\temp", boostfs::native ); boostfs::path searchPath( "d:\\temp" ); Looking at the boost test files [at least some of] these appear to be valid. The key seems to be in the method windows_name(...) in path_posix_windows.cpp. This method searches through the array invalid_chars[] which contains "<>:\"/\\|" ... hence the reason why "d:" causes a failure? I've rebuilt the source from the bjam files, tried compiling them into my own code, compiled as a dll etc. to no avail ... Could someone shed some light on this problem before I go even crazier? Thank you kindly in advance; Steve a.

2 1

Lambda question (deferring evaluation)
by Johan Nilsson 03 Mar '05

03 Mar '05

Hi, would it be possible to implement something the following lines using lambda, without having all values referring to the same shared object: ... #include <boost/shared_ptr.hpp> #include <boost/lambda/lambda.hpp> #include <algorithm> #include "Foo.h" ... using namespace boost; using namespace boost::lambda; std::vector<shared_ptr<Foo> > foos(10); std::for_each(foos.begin(), foos.end(), _1 = shared_ptr<Foo>(new Foo)); Thanks // Johan

4 4

BOOST_CHECK_EQUAL and operator<<
by Patrick Mézard 03 Mar '05

03 Mar '05

Hello, I am currently upgrading from boost-1.30.2 to boost-1.32.0 with MSVC 7.1 and I am running into the following issue: BOOST_CHECK_EQUAL requires the elements to be compared to provide an operator<< taking an std::ostream& as first parameter. I always had names lookups problems with that when the operator was provided by the test and not along with the element. Until now, I had the operator defined in "boost" namespace (that's ugly I know) in the test file. It breaks with 1.32.0 because the problematic function (print_log_value) moved deeped into test tools namespaces. Then I try to revert to a normal definition but it does not work either. For instance : //******************************* #include <boost/test/unit_test.hpp> using boost::unit_test::test_suite; namespace testns { struct Test { Test(std::string const& s_) : s(s_) {} bool operator==(Test const& t) const { return s==t.s; } std::string s; }; namespace { std::ostream& operator<<(std::ostream& o, Test const& t) { return o<<t.s; } void free_test_function() { Test t1("test1"), t2("test2"); BOOST_CHECK_EQUAL(t1, t2); } } //namespace } //namespace testns test_suite* init_unit_test_suite( int, char* [] ) { test_suite* test= BOOST_TEST_SUITE( "Unit test example 1" ); test->add( BOOST_TEST_CASE( &testns::free_test_function )); return test; } //*********************** Does not compile and gives me a: C2679: binary 'operator' : no operator defined which takes a right-hand operand of type 'const testns::Test' (or there is no acceptable conversion), the failure being issued by "void boost::test_tools::tt_detail::print_log_value<T>::operator ()(std::ostream &,const T &)" However, if I define operator<< as a public member of testns::Test it compiles. Same thing if I move it outside the anonymous namespace, with or without static linkage. Note these workaround do not work or cannot be applied if the type is declared by a typedef. In this case, I have to define again the operator in namespace boost.test_tools.tt_detail. Is it to be expected ? Is is an ADL issue ? Or something related to MSVC 7.1 ? What is the proper way to do that ? Patrick Mézard

4 9