July 2008 - Boost - lists.boost.org

Interest in properties library?
by Sergey Sadovnikov 11 Jul '08

11 Jul '08

Hello All, Would there be any interest in another C++ properties implementation? I know that boost sandbox already contains properties library. But I can offer another implementation of such library with some advantages like this: - Simple properties set initialization. Only one call from the constructor: class SomeClass { public: //... SomeClass() : properties__(this) {/*...*/} //... }; - Many kins of properties. Now implemented: Simple properties similar to datamembers. Properties bindable to getter/setter methods of container class. Properties bindable to arbitrary methods. Index properties with up to three levels of indirection. - Properties could be "abstract" and need implementation in derived classes - In-place initialization of simple properties. For example: PROPERTY_I(int, ContainerSize, 10) - Generic runtime access to the properties via names: SomeClassWithProps c; int val; c.properties__.invoke_get("ContainerSize", var(val) = boost::lambda::_1); c.properties__.invoke_set("ContainerSize", 20); - Runtime enumeration of properties: // Property visitor function template<typename P> void PropertyEnumerator(const std::string& propertyName, P SomeClassWithProps::* propInfo) { //... ] // Properties enumeration: SomeClassWithProps::properties__::enum_s(PropertyEnumerator); - Simple serialization of properties: template<typename Ar> void serialize(Ar& ar, const unsigned int ver) { ar & properties__; } - Quite simple definition in class. A piece of code from real project: // Example of abstract properties definition in the base class: class IPlayerCraft : public IAirObject { public: IPlayerCraft() : properties__(this) {;} virtual ~IPlayerCraft(){} // Craft status enum ECRAFTSTATUS { csREADY =0, csFLYING, csREPAIRING, csREFUELING, csREARMING, csTRANSFERING, csMAXCOUNT }; PROPERTIES(IPlayerCraft, // Craft name ABSTRACT_RW_PROPERTY(std::wstring, Name) // Static craft info ABSTRACT_RO_PROPERTY(SCraftInfo const*, Info) // Craft home base ABSTRACT_RO_PROPERTY(IBase const*, HomeBase) // Is craft on the base? ABSTRACT_RO_PROPERTY(bool, IsOnTheBase) // Craft status ABSTRACT_RO_PROPERTY(ECRAFTSTATUS, Status) // Current craft speed ABSTRACT_RO_PROPERTY(TPARAM, Speed) // Craft weapon ABSTRACT_INDEX_PROPERTY(IItem*, CraftWeapon, 1, (CraftSlotCoord)) // Craft weapon ammo ABSTRACT_RO_INDEX_PROPERTY(TPARAM, CraftWeaponAmmo, 1, (CraftSlotCoord)) ); //... }; // Implementation class class PlayerCraft : public IPlayerCraft { public: PlayerCraft(SCraftInfo const*, Player* player); ~PlayerCraft(); PROPERTIES(PlayerCraft, IMPLEMENT_PROPERTY(IPlayerCraft, PROPERTY(std::wstring, Name)) IMPLEMENT_PROPERTY(IPlayerCraft, PROPERTY(SCraftInfo const*, Info)) IMPLEMENT_PROPERTY(IPlayerCraft, RO_PROPERTY(IBase const*, HomeBase, GetHomeBase)) IMPLEMENT_PROPERTY(IPlayerCraft, PROPERTY(bool, IsOnTheBase)) IMPLEMENT_PROPERTY(IPlayerCraft, PROPERTY_I(ECRAFTSTATUS, Status, csTRANSFERING)) IMPLEMENT_PROPERTY(IPlayerCraft, PROPERTY_I(TPARAM, Speed, 0)) IMPLEMENT_PROPERTY(IPlayerCraft, INDEX_PROPERTY(IItem*, CraftWeapon, 1, (CraftSlotCoord), GetCraftWeapon, SetCraftWeapon)) IMPLEMENT_PROPERTY(IPlayerCraft, RO_INDEX_PROPERTY(TPARAM, CraftWeaponAmmo, 1, (CraftSlotCoord), GetCraftWeaponAmmo)) ); // ... IBase const* GetHomeBase() const {return m_HomeBase;} private: // ... IItem* GetCraftWeapon(int slotId) const; void SetCraftWeapon(int slotId, IItem* weapon); TPARAM GetCraftWeaponAmmo(int slotId) const; }; As you can see, properties definition based on boost::preprocessor. It allows to implement quite complex background (with many special abilities) for quite simple front-end. It is assumed what access to the properties in many cases is same as access to the datamember: PlayerCraft craft; std::string craft_name = craft.Name; //... IItem* weapon = craft.CraftWeapon[1]; And there is special access form needs in some cases: craft.Name(); or craft.Name.get(); If there is an interest in such properties implementation I could start prepare code for formal library review. -- Best Regards, Sergey mailto:flex_ferrum@artberg.ru

2 1

Recursive iterator/range adaptors
by Mathias Gaunard 11 Jul '08

11 Jul '08

The traversal library that has been discussed allows tree-like traversal of structures using a functional approach. I suggest an alternative approach: a range-based one. It certainly has limitations: ranges are monomorphic. So you can only do it if is_same< typename range_value<Range>::type, typename range_value< typename range_value<Range> >::type >. However, it would be very practical to use, since you could simply use a foreach loop to perform depth-first search and stop iteration whenever you like. Also, it is lazy and you can combine it easily with other range adaptors (filtering, transforming...) or algorithms. For performance, one would need to provide a way to get the father range of a given range. Maybe that's the occasion to create a new concept. Otherwise iterators will need to have their own stack. Did anyone ever work on that? Does it look like a good idea?

2 3

[pool] segfault
by Phil Bouchard 11 Jul '08

11 Jul '08

I was testing my own "shifted_ptr_test2.cpp" on RHEL 4.5 using Gcc 3.4.6 and the following object declared in "sh_owned_base_nt.hpp" caused a segmentation violation inside its destructor. It doesn't seem to happen on Windows, just Linux: pool owned_base::pool_; // derives from boost::pool<> Thanks, -Phil

3 7

[graph] generating CSR graph from undirected
by Mojmir Svoboda 11 Jul '08

11 Jul '08

hello, i suspect there is something wrong with assign routine in compressed_sparse_row_graph.hpp. when i am constructing directed csr from undirected one via // From any graph (slow and uses a lot of memory) // Requires IncidenceGraph, a vertex index map, and a vertex(n, g) function // Internal helper function template<typename Graph, typename VertexIndexMap> void assign(const Graph& g, const VertexIndexMap& vi, vertices_size_type numverts, edges_size_type numedges) { inherited_vertex_properties::resize(numverts); m_rowstart.resize(numverts + 1); m_column.resize(numedges); it overwrites non-allocated memory, because the number of edges is in fact doubled.. is it an imperfection or am i doing something stupid? have a nice day, mojmir

1 0

Boost DateTime package question
by Mohnish Kodnani 11 Jul '08

11 Jul '08

Hi All, I have a question regarding the boost DateTime package. In our application we have to support storing time from various time zone in UTC and we need to know DST of various timezones. We are currently using standard C runtime libraries for this purpose on various platforms like windows , linux, aix, hpux. My question, Recently there has been a need to store dates prior to epoch in our system and this causes problems since none of the standard C runtime libraries support dates prior to EPOCH. We are also interested in DST settings prior to EPOCH, since in our experiments it looks like Java knows DST till 1942 or in its vicinity, however C only deals with dates after epoch. Our server is in C++ and client in Java, so we have issues here when client sends a date prior to 1942 and we have to store its UTC value, we cannot do this since C runtime mktime like functions dont work. Does Boost DateTime package provide this functionality, if not is that in the plans. >From what I read so far it looks like this package does not have this functionality, but has the functionality to create our own rules, however this is difficult considering all countries around the world, it would be good if we had something similar to C runtime. Please let me know. thanks Mohnish

1 0

[constrained_value] Formal review request
by Robert Kawulak 11 Jul '08

11 Jul '08

Dear Boosters, I'm glad to say the Boost Constrained Value library is finally ready for a formal review. To remind what this library is about, an excerpt from the motivation section of the documentation is attached at the bottom of this message. Jeff, you volunteered to be the review manager - is your proposition still up to date? If so, I'd be thankful if you could do the work. ;-) The code and documentation can be downloaded here: http://rk.go.pl/f/constrained_value.zip The documentation is also available online here: http://rk.go.pl/r/constrained_value The only thing that's missing is test programs - I will try to add them during/after the review. The changes made since previous version: - rewritten the documentation at some points and added tutorial, - no more value generators - direct types or MPL integral constants are used instead, - bounded types use bounds exclusion instead of inclusion (so bounds are included by default in all cases), - changed order and default values of some of template parameters, - removed throw_verbose_exception error policy, - made other minor fixes and improvements. As always, feedback is welcome. Best regards, Robert Kawulak FROM THE MOTIVATION SECTION OF THE DOCUMENTATION: The Boost Constrained Value library contains class templates useful for creating constrained objects. A simple example is an object representing an hour of a day, for which only integers from the range [0, 23] are valid values: bounded_int<int, 0, 23>::type hour; hour = 20; // OK hour = 26; // exception! Behavior in case of assignment of an invalid value can be customized. For instance, instead of throwing an exception as in the example above, the value may be adjusted to meet the constraint: wrapping_int<int, 0, 255>::type buffer_index; buffer_index = 257; // OK: adjusts the value to fit in the range by wrapping it assert( buffer_index == 1 ); The library doesn't focus only on bounded objects as in the examples above -- virtually any constraint can be imposed by using a predicate: // constraint (a predicate) struct is_odd { bool operator () (int i) const { return (i % 2) != 0; } }; // and the usage is as simple as: constrained<int, is_odd> odd_int = 1; odd_int += 2; // OK ++odd_int; // exception! The library has a policy-based design to allow for flexibility in defining constraints and behavior in case of assignment of invalid values. Policies may be configured at compile-time for maximum efficiency or may be changeable at runtime if such dynamic functionality is needed.

4 5

[test][sun-5.7 -5.8][trunk] regressions for Boost.Test starting at revision 47170
by JOAQUIN M. LOPEZ MUÑOZ 10 Jul '08

10 Jul '08

Hello, Recent updates to the trunk version of Boost.Test have resulted in a massive number of regressions for Sun 57 and 5.8, as shown for instance in: http://tinyurl.com/5lm7ha http://tinyurl.com/66qcze The error message is "../boost/test/utils/iterator/input_iterator_facade.hpp", line 56: Error: Multiple declaration for input_iterator_facade. After examining the updates to Boost.Test in the last days I think I found the change that's causing the regressions: http://svn.boost.org/trac/boost/changeset/47170/trunk/boost/test The change in line 26 of boost/test/impl/unit_test_main.ipp causes the header <boost/test/utils/iterator/token_iterator.hpp> to be included where previously that was not the case. This in turns includes <boost/test/utils/iterator/input_iterator_facade.hpp>, which contains the following problematic code (for Sun 5.7-8) around line 25: #if defined(BOOST_NO_MEMBER_TEMPLATE_FRIENDS) || BOOST_WORKAROUND(__BORLANDC__, BOOST_TESTED_AT(0x551)) public: #else template <class I, class V, class R, class TC> friend class input_iterator_facade; #endif This perfectly correct friend declaration is not processed properly by Sun 5.7 and 5.8 hence the error. May I ask that the problem be fixed by applying either one of the following? 1. Add a workaround for Sun 5.7-8 at line 35 of boost/test/utils/iterator/input_iterator_facade.hpp 2. Add a !workaround for Sun 5.7-8 at line 26 of boost/test/impl/unit_test_main.ipp Thank you! Joaquín M López Muñoz Telefónica, Investigación y Desarrollo

1 0

[inspect] A few issues in non-ASCII checker
by Gennaro Prota 10 Jul '08

10 Jul '08

Hi, I had a summary look at the changes made to Boost.Inspect (since I last saw it) and noticed a few issues: * the "*A*" marker is used either for broken links, invalid urls etc. *and* non-ASCII chars * "non-ASCII" itself looks like a misleading name: see the string gPunct, defined in ascii_check.cpp * non portable (though widely portable :-)) code: if ( c >= 'a' && c <= 'z' ) return false; if ( c >= 'A' && c <= 'Z' ) return false; if ( c >= '0' && c <= '9' ) return false; If there aren't problems with standard library support I'd suggest replacing the above with if ( std::isalnum( static_cast< int >( c ), std::locale::classic() ) ) Similarly for the subsequent isspace-like tests * again in ascii_check.cpp there's some scaffolding to get a line number from a file position; this is needed in other inspector classes as well, and is done much differently. I think it is worth unifying this part and make it available in one place -- Genny

3 3

public beta test ?
by Ruediger Berlich 10 Jul '08

10 Jul '08

Hi there, I'd like to suggest a public beta test as part of the release procedure. >From what I can see, while the trunk version is available to everybody, there is no formal call for external testing. But this would increase visibility of Boost and would likely reveal a number of additional bugs. Best, Ruediger

2 2

a little imperfection in options_description.cpp example (program options)
by Guido Ziliotti 10 Jul '08

10 Jul '08

// A helper function to simplify the main part. template<class T> ostream& operator<<(ostream& os, const vector<T>& v) { copy(v.begin(), v.end(), ostream_iterator<T>(cout, " ")); /* this is line 21 */ return os; } SHOULD BE REPLACED WITH // A helper function to simplify the main part. template<class T> ostream& operator<<(ostream& os, const vector<T>& v) { copy(v.begin(), v.end(), ostream_iterator<T>(os, " ")); /* this is line 21: "os" replaces "cout" */ return os; }

2 1