May 2006 - Boost - lists.preview.boost.org

[serialization] std::vector<>
by Richard Crossley 22 May '06

22 May '06

Hi, Using CVS HEAD with VC8 fails when writing out an empty std::vector as T* get_data(STD::vector<T,Allocator>& v) attempts to access element 0. Looks to me as if the optimized versions of save() and load() for std::vector in vector.hpp should be checking the vector size first. Regards, Richard.

2 1

[serialization] MSVC8 runtime check error
by John Reid 22 May '06

22 May '06

Hi, I'm getting a runtime check error from MSVC8 in the following code checked out from cvs_head: serialization/collections_load_imp.hpp line 116 ... template<class Archive, class Container, class InputFunction, class R> inline void load_collection(Archive & ar, Container &s) { s.clear(); // retrieve number of elements unsigned int count; unsigned int item_version; ar >> BOOST_SERIALIZATION_NVP(count); if(3 < ar.get_library_version()){ ar >> BOOST_SERIALIZATION_NVP(item_version); } R rx; rx(s, count); InputFunction ifunc; while(count-- > 0){ ifunc(ar, s, item_version); } } ... The debugger breaks every time ifunc() is called which is frustrating to say the least. If load_collection() is wrapped with #pragma runtime_checks( "u", off ) ... #pragma runtime_checks( "u", restore ) it doesn't help - perhaps I've rebuilt it incorrectly, perhaps the pragma needs to be around where the code is instantiated. I managed to get around the problem by assigning 0 to item_version. Has anyone specifically chosen to turn runtime checks on for debug builds with MSVC8 or is this just the default? Thanks, John.

1 0

[serialization] patch for Borland
by David Abrahams 22 May '06

22 May '06

I just checked this one in. Wouldn't compile on Borland without it, due to duplicate definitions of the same function template. BTW, the comment in that file reads: // if function is partiall specialized - which is wrong, because there's no partial specialization of functions. It should be fixed to say what you really mean. -- Dave Abrahams Boost Consulting www.boost-consulting.com

1 0

timedout async operation problem in boost::asio
by 李慧霸 22 May '06

22 May '06

I'm trying to use boost::asio, and I find that there isn't a built-in timeout parameter in any async operation. I think it's a very important function when writing real life programs, and I have done some walk around work, how can I try to contribute?

3 2

Web access to the CVS
by Joe Gottman 22 May '06

22 May '06

For about a week now, every time I try to use the web link to the CVS repository (http://cvs.sourceforge.net/cgi-bin/viewcvs.cgi/boost/boost/) I get a "This page cannot be displayed" error. What gives? Joe Gottman

4 4

[serialization] concept requirements
by David Abrahams 22 May '06

22 May '06

Hi Robert, Once again I find myself confounded by the true concept requirements for archives. Once upon a time, the truth-as-we-knew-it was that the following (copied from the library docs) was part of the input archive requirements. All input archives should be derived from the following template: template<class Archive> detail::common_iarchive; But, as I just discovered: polymorphic_iarchive doesn't meet those requirements. Since you have struggled with how to write the concept requirements for archives, I have a brilliant suggestion: create concept checking classes and archetypes using Boost.ConceptCheck, and use those to formulate and validate your concepts. Then you can simply copy the concept checking classes into your documentation and add a description of semantic constraints. You can't go wrong if you do that; the compiler will check your work for you, and make sure your code is consistent with your documentation. [I have recently upgraded the ConceptCheck library to allow some nicer but as-yet-undocumented syntax. When (if?) you decide to do this, I'll give you the quick skinny] -- Dave Abrahams Boost Consulting www.boost-consulting.com

2 4

[serialization] array problems and patch
by David Abrahams 21 May '06

21 May '06

In my work on eliminating header ordering issues, I discovered the need for the enclosed patch. This is not a subtle issue; it seems to me that the trunk is very broken without it. -- Dave Abrahams Boost Consulting www.boost-consulting.com

2 2

[bjam/test] how to pass command arguments to test-case
by Thorsten Ottosen 21 May '06

21 May '06

Hi, The test docs state that "If you got "memory access violation" message (or any other message indication fatal or system error) when you run you test, to get more information of error location add --catch_system_errors=no to the test run command line:" (see http://www.boost.org/libs/test/doc/usage/msvc_net.html) How do I do this via bjam? Thanks -Thorsten

2 6

Re: [boost] [multi_index]terminology/design concern
by JOAQUIN LOPEZ MU?Z 21 May '06

21 May '06

----- Mensaje original ----- De: Gennadiy Rozental <gennadiy.rozental(a)thomson.com> Fecha: Miércoles, Abril 12, 2006 8:41 pm Asunto: Re: [boost] [multi_index]terminology/design concern Joaquín wrote: > > From your description above, seems to me what you understand > > as an "order" is verily what the index concept (as used in > > Boost.MultiIndex) tries to model. Basically, an index is > > an access interface to a set of objects jointly owned > > by several other indices --in contrast with regular STL > > containers who own their elements in an exclusive fashion. > > > > [...] > > > "Order" would be a good candidate, except that it's generally > > tied to fixed-order containers (sets, maps), and > > Boost.MultiIndex also includes sequence-like constructs. > > I do not see a problem with term "Order". You could have > mulit_order_container. While STL present single order ones. As I said, I'm reluctant to name sequenced indices as "orders", since it is precisely the lack of any fixed order that defines them. Besides, the differences between index interfaces are huge, and referring to them collectively as orders tends to blur (IMHO) these differences, as if the interfaces were homogeneous, which they aren't. Anyway, names were already chosen, published, and pretty much cast in stone when the library was first released in 2004. Back then, this conversation could have had more practical implications, right now I can only concede/deny your points, but no name change is likely to occur. > > > > I'm afraid I'm not quite getting what you mean by "index" > > in this context. Certainly, the kind of operations you describe > > Index is something that provided direct (constant-time) access > somewhere > within selected order. > > > can be performed in Boost.MultiIndex by means of so-called > > "compatible keys" as described in http://tinyurl.com/rouet, > > I am not sure how this is related. > > > so that, for instance, given an ordered index sorted by > > (last name, first name) one can jump to the records > > beginning with some given last name. Is this what you're > > referring to? > > Essencially. The same way as book index allows to jump to proper > page based > on keyword. Could you show some code example? What I want is to > iterate > through all persons with last name "Smith". All operations should > have > contant complexity. So, if your multi-index container is defined like this: typedef multi_index_container< address, indexed_by< ordered_non_unique< // (last_name,first_name) index composite_key< address, member<address,std::string,&address::last_name>, member<address,std::string,&address::first_name> > > ... // other indices > > address_book_t; address_book_t b; then you can write the following: void list(const std::string& last_name) { std::pair<address_book_t::iterator,address_book_t::iterator> p= b.equal_range(last_name); std::copy( p.first,p.second, std::ostream_iterator<address>(std::cout)); } You require that all ops are constant complexity, which is not entirely fulfilled by my example, since equal_range is O(log n). On the other hand, I cannot see why your indexing concept must guarantee constant complexity and not only fast lookup. > > If so, the only thing I can say is I never > > thought of these operations as directly related to an > > index concept. If you could elaborate on your discussion > > maybe I'll be seeing your point more clearly. > > > > Thanks for discussing your concerns, I hope at least I've > > shed some light on them. > > IMO we need another level of indirection that would support Index > within an > Order concept. I do not see how existing design could handle this. > If the need for the kind of bookmarking (allow me to use this term so as to not mix the different meanings we're giving to the word "index") you describe is sufficiently general, such a facility can be built separately from B.MI itself (like a standalone view-like library) or integrated into B.MI as an index decorator: typedef multi_index_container< element, indexed_by< bookmarked< ordered_unique<...> > ... > > So, the bookmarked<...> construct can embed any preexistent index ("order" in your terminology) type and supplement it with bookmarking facilities, provided we had a precise definition of what bookmarking boils down to. This kind of stuff can be done within the current design, but then again I yet fail to see how what you describe is a general need not sufficiently covered by current facilities, like for instance compatible keys in ordered indices --clarification here most welcome. Joaquín M López Muñoz Telefónica, Investigación y Desarrollo

4 4

[serialization] portable binary archives invert no_header sense
by Kim Barrett 20 May '06

20 May '06

The example portable binary archives (both directions) contain the following lines in their constructor bodies: if(0 != (flags & boost::archive::no_header)){ this->archive_base_t::init(flags); // ... more stuff } The test on flags value should be "==" rather than "!=", i.e. if(0 == (flags & boost::archive::no_header)){

2 1