February 2007 - Boost-users - lists.preview.boost.org

[program_options] unsigned variables silently cast
by Torrey Pines 28 Feb '07

28 Feb '07

Hi With the current implementation, unsigned variables accept negative values, that get silently cast. E.g. on my platform, if "--length" expects size_t, --length -32 assigns it a value 4294967264 without warnings or errors. This behavior surprised me, and I'm using a workaround (parsing an int, and checking its range manually). However, is this an intended feature of the library? Thanks

2 1

[program_options] disable partial matches
by Torrey Pines 28 Feb '07

28 Feb '07

Hi With the current implementation, the user does not have to supply the full option names, e.g. "--he" is as good as "--help", unless there is an ambiguity. Is there any way to disable this, i.e. allow only full matches? Thanks

2 1

[program_options] checking the presence of required arguments
by Torrey Pines 28 Feb '07

28 Feb '07

Hi My program has a bunch of required arguments. I grouped them into one options_description: options_description required("Required options"); required.add_options() /* .... */ ; // ... options_description desc("Allowed options"); desc.add(required).add(other); I'd like to check that all required arguments are in fact supplied without having to do if(vm.count("arg1") == 0) { cerr << "usage error: arg1 is required\n"; return 1; } for each of them. Any suggestions? Thanks

2 1

boost::any template argument deduction
by Digvijoy Chatterjee 28 Feb '07

28 Feb '07

Hi, I have a function template like : template <class A> A what_type ( const boost::any &operand ){ A a; a = boost::any_cast<a>(operand); return a; } The idea is to provide the client code an easy way extract the type hidden by the boost::any value, Now there is no way to make the compiler deduce the type to return from an expression like below: std::string s = "hello"; boost::any a = s; some_op( what_type(a) ); //calls some_op(string &); Instead now the client has to write code using type() which returns std::typeinfo & if ( a.type().name() == "Ss" ) some_op(what_type<string> a); else if ( a.type().name() == "d" ) some_op(what_type<double> a); Or use functions to test type and cast: if ( is_string (a) ) some_op( what_type<string>(a)); else ( is_int(a) ) some_op ( what_type<int>(a)); ….. Is there any elegant way to handle this problem ?? --Digz

1 0

[MPL] Problem with vector
by Daniel 28 Feb '07

28 Feb '07

Hi, I have few Adapter classes. Each of them provide typedef operations which specifies which operations that Adapter supports (vector_c, with 0 or 1 at appropriate positions). I want to assert that two Adapters support different operations. However, my code doesn't compile. Any ideas? Daniel template<unsigned int N1, unsigned int N2> struct Adapter { typedef vector_c<unsigned int, N1, N2> operations; }; template<class TAdapter1, class TAdapter2> struct Check { typedef typename transform< typename TAdapter1::operations, typename TAdapter2::operations, plus<_1, _2> >::type sum_of_operations; typedef typename max_element<sum_of_operations>::type max_of_operations; BOOST_MPL_ASSERT((less_equal<max_of_operations, int_<1> >)); }; int main() { Check< Adapter<0, 1>, Adapter<1, 0> >(); return 0; }

2 1

[signals] slot<Signature>, slot::track() and bind syntactic sugar
by Frank Mori Hess 27 Feb '07

27 Feb '07

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 I've added template parameters to the slot class in thread_safe_signals for specifying its signature, as slot<Signature> or slot1<R, T1>, etc. So now slots can be bound to each other. I've also dropped the automatic discovery of tracked objects, and added slot::track() functions. There is now some syntactic sugar for bind, if a slot constructor is called with more than one argument, then the arguments will automatically be passed through bind(). I tried using bind<R>(), since the slot knows what its result type should be, but gcc gave me ambiguous overload errors when bind<R> was given a function pointer. There is a little test program in the sandbox now too, at boost-sandbox/libs/thread_safe_signals/test/track_test.cpp. I encountered one problem with the bind syntactic sugar, described in the test program's comments. I found the error message unhelpful, therefore it must be a compiler bug :) - -- Frank -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.1 (GNU/Linux) iD8DBQFF5KRY5vihyNWuA4URAo2mAJ45BAJvKKl8rHPXGsgn7a2fjmDLVQCeLb0R ln07v8CCYxbwE7topMmhZQ4= =4kwy -----END PGP SIGNATURE-----

1 0

[multi index] warning when used in copy constructor: should be explicitly initialized
by Christoph 27 Feb '07

27 Feb '07

Using gcc with option -W I get the following warning when using a Boost.MultiIndex container inside a copy constructor: /usr/local/include/boost-1_33_1/boost/multi_index_container.hpp: In copy constructor 'boost::multi_index::multi_index_container<Value, IndexSpecifierList, Allocator>::multi_index_container(const boost::multi_index::multi_index_container<Value, IndexSpecifierList, Allocator>&) [with Value = ListenMonitor::ListenZeile, IndexSpecifierList = boost::multi_index::indexed_by<boost::multi_index::ordered_unique<boost::multi_index::identity<ListenMonitor::ListenZeile>, mpl_::na, mpl_::na>, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na>, Allocator = std::allocator<ListenMonitor::ListenZeile>]': liste.hpp:139: instantiated from here /usr/local/include/boost-1_33_1/boost/multi_index_container.hpp:204: warning: base class 'struct boost::multi_index::detail::header_holder<boost::multi_index::detail::ordered_index_node<boost::multi_index::detail::index_node_base<ListenMonitor::ListenZeile> >, boost::multi_index::multi_index_container<ListenMonitor::ListenZeile, boost::multi_index::indexed_by<boost::multi_index::ordered_unique<boost::multi_index::identity<ListenMonitor::ListenZeile>, mpl_::na, mpl_::na>, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na, mpl_::na>, std::allocator<ListenMonitor::ListenZeile> > >' should be explicitly initialized in the copy constructor The 'ListenMonitori::Listenzeile' stuff is from me. But the issue does not depend on that. I would assume that copying a multi index is not forbidden and using Is there a way to get rid of that warning? I am trying hard to produce warning free code, and in this case I am not sure if the warning is just ugly or if it is a real problem in the code. [gcc --version: gcc (GCC) 4.1.2 20061115 (prerelease) (SUSE Linux) ] I have attached a trivial example that demonstrates the issue. I have modified the example/basic.cpp: I have added a trivial class XXX that contains employee_set. When copying objects of XXX the above warning is issued by gcc. Best regards Christoph

3 9

Possible bugs in date_time package
by Gregory J. Sharp 27 Feb '07

27 Feb '07

I'm trying to use the date_time/ptime module to parse strings. Platform: MacOS X 10.4.7 on Intel Compiler: gcc 4.0.1 I have three potential bugs, so far. Perhaps I'm unreasonable in hoping that the following strings should be acceptable to time_from_string. 1. time_from_string( "2006-12-06" ) != time_from_string( "2006-12-06 00:00:00" ) 2. time_from_string( "2006/12/06" ) throws an exception (probably related to 1) even though time_from_string( "2006,12,06" ) doesn't throw. 3. time_from_string( "2006-12-06 " ) throws an exception because it passes a null string to str_from_delimited_time_duration(), which tries to read the first character of the string, even though there isn't one. I think 1 & 2 are probably caused by parse_delimited_time() doing a split on ' ' and getting nothing to pass to parse_delimited_time_duration(). -- Gregory J. Sharp email: gregor(a)mail.lepp.cornell.edu Wilson Synchrotron Laboratory url: http://www.lepp.cornell.edu/ ~gregor Cornell University ph: +1 607 255 4882 Ithaca, NY 14853 fax: +1 607 255 8062

2 4

multi_index uses serialization?
by Javier Gonzalez 27 Feb '07

27 Feb '07

Hi, I have a simple question. Isn't there a way to create multi_indices without using the serialization library? I don't intend to stream them! thanks, Javier Gonzalez

5 6

State of the Endian library
by Mateusz Loskot 27 Feb '07

27 Feb '07

Hi, What's the state of the Endian library announced [1] by Beman Dawes? Is there any plan to include it to the Boost collection? May be it's already there, but I can not find it? [1] http://thread.gmane.org/gmane.comp.lib.boost.devel/143279 Cheers -- Mateusz Loskot http://mateusz.loskot.net

1 0