February 2006 - Boost-users

[regex] - bk_parens and bk_braces missing from 1.33.x
by Keith MacDonald 12 Feb '06

12 Feb '06

Resending due to no response since 4th Feb: It states in history.html for regex 1.33.0 that "some of the more esoteric options have now been removed, so there is the possibility that existing code may fail to compile: however equivalent functionality should still be available". I suppose bk_parens and bk_braces could be considered esoteric, but I cannot find any equivalent functionality. Is that intentional? Thanks, Keith MacDonald

4 7

[ptr_container] ptr_map predicates
by Bill Buklis 12 Feb '06

12 Feb '06

How would I accomplish the following using ptr_map instead of std::map? typedef std::map< std::string, int* > stlmap; bool predicate( stlmap::value_type& mapitem ) { // return true if key value has at least 4 characters return( mapitem.first.size() > 3 ); } void sample( stlmap& themap ) { std::find_if( themap.begin(), themap.end(), predicate ); } The predicate for std::find_if requires *iterator for the parameter. However, with ptr_map *iterator returns a reference to the element value instead of the key/value pair. Normally this is fine (in general code I like this feature), but in this case I need access to the key in the predicate. How do I get at it? ptr_map version of above: typedef boost::ptr_map< std::string, int > bpmap; bool predicate( ??? ) { ??? } void sample( bpmap& themap ) { std::find_if( ??? ); }

3 12

boost::xtime quesion
by Tim Blechmann 12 Feb '06

12 Feb '06

hi all, maybe it's just a stupid newbie question, but i'm a bit confused by boosts timed locks ... they're using a time, that's based on the timeval struct on posix systems ... what's the difference between this and ptime from the date_time library? ptime can be used with a clock with microsecond granularity, too ... in any case, how do i convert times from ptime to xtime? thanks in advance ... tim -- mailto:TimBlechmann@gmx.de ICQ: 96771783 http://www.mokabar.tk Every word is like an unnecessary stain on silence and nothingness Samuel Beckett

1 0

[serialization] I am unable to link the serialization demo.cpp
by Michael Shapiro 12 Feb '06

12 Feb '06

I would like to use the boost serialization library to do checkpointing for a simulation which runs for one or more weeks. However, I have not been able to get past the "hello world" stage. I have tried to do this on both a linux machine running gcc 3.4.5 and Mac OS X using gcc 3.3. In both cases, my attempts to compile the demo.cpp program fail with long lists of undefined symbols. I will give the specifics for the Mac because this is where it will ultimately be used and because the instalation seems more standard there. The demo.cpp I am trying to compile is the one available on http://www.boost.org/libs/serialization/doc/tutorial.html The boost include files are installed in /usr/local/boost/gcc/ The libraries are installed in /usr/local/lib/ For example, here I find libboost_serialization.a and libboost_serialization.dylib. I am attempting to compile as follows: gcc \ -L/usr/local/lib/ \ -I/usr/local/include/ \ -lboost_serialization \ demo.cpp I have tried various variations on this including adding all the .a files to the list of libraries to link, and adding stdc++. I get error messages of which the following are typical ld: Undefined symbols: __ZNKSs5c_strEv __ZNKSt9basic_iosIcSt11char_traitsIcEE4failEv __ZNSaIcEC1Ev __ZNSaIcED1Ev followed by more of the same. I have no doubt there is something very simple I am failing to do, very likely something so simple no one knowledgeable would think to put it in the documentation. So I will probably feel very dumb when someone educates me about this, but I will also feel very relieved. Thanks in advance for your help.

2 1

[bjam] Boost Jam 3.1.12 release.
by Rene Rivera 12 Feb '06

12 Feb '06

Like the subject says, Boost Jam 3.1.12 is available on the SourceForge release system: <https://sourceforge.net/project/showfiles.php?group_id=7586&package_id=7294…>. This should be the bjam used for the next Boost release. Enjoy. -- -- Grafik - Don't Assume Anything -- Redshift Software, Inc. - http://redshift-software.com -- rrivera/acm.org - grafik/redshift-software.com -- 102708583/icq - grafikrobot/aim - grafikrobot/yahoo

1 0

Creating a copy of a subset of a graph
by Adam Teasdale Hartshorne 12 Feb '06

12 Feb '06

Hi All, I want to form a new graph B from an existing graph A, copying out only edges(and their all their vertices) given that an edge contains at least one of the vertices in a given list (i.e. I am going to end up with a graph with more than just the vertices in the list and their adjacency information. With some extra vertices, where an edge exists between them and vertices in the defined list.) I know I should be using filter_graph and copy_graph, but not 100% sure how. Any help/code snippets would be much appreciated, I am new to boost graph library so please make any explanations as simple as possible, Adam

1 0

Template metaprogramming for compile-time initialization of complex types?
by Arno Schödl 12 Feb '06

12 Feb '06

Hello, here is my problem: I have a complex data type, like a std::vector< std::string > and need some global constant values of this type. The type is also used non-constant throughout the program, so I do not want to use dumb types, but rather STL containers. In an ideal world, initialization of these constants would not happen at runtime, but at compile time, essentially generating a constant memory image of this type via meta-programming. The exact memory layout is compiler-dependent, of course, but the difficult, but portable part would be the programmer's interface, in particular for nested types and user-defined ones. Why do I want this? Start-up time is a big problem for larger apps, and CRT init code filling std::vectors with stuff causes lots of page faults slowing down the start. This is particularly annoying if the constant never happens to be used in that particular run. And of course it is inelegant to keep data around twice, once in dumb constant arrays, and once on the heap after the containers have been filled from these arrays. Why can't I use POD arrays for constant data instead of STL containers? Because the same data type is often used both const (for default values, to list possible choices in menus, etc.) and non-const (for user-defined values). Unifying the access would be awfully convenient. I have implemented this approach with the boost preprocessor library for MSVC 8 std::vector< integral type >, which works, but is nowhere general enough. Has anyone else made attempts in this direction? Arno

1 0

Boost, STLport, and Visual Studio 2005
by Robert S. Grimes 11 Feb '06

11 Feb '06

Hi all, Trying to get Boost 1.33.1 and STLport 5.0.1 to co-exist under Micro$oft Visual Studio 2005. I got them both to apparently compile and install okay, but when I attempt a Boost::Serialization test, I run into a problem: LINK : fatal error LNK1104: cannot open file 'libboost_serialization-vc80-mt-gdp-1_33_1.lib' >From the docs, I indeed find that the "-gdp-" library is non-existent; specifically, the 'p' for STLport is missing. How do I build the STLport runtimes? Thanks! -RSG

2 1

Minor spurious warning message -- Very Trivial
by Robert S. Grimes 10 Feb '06

10 Feb '06

Hi all, Using Boost 1.33.1 serialization library with Micro$oft Visual Studio 2005, I get two minor warnings while compiling the demo_xml.cpp program. They are in boost/serialization/collections_save_imp.hpp. Lines 36 and 37 in the function template save_collection are the culprit, resulting in warnings of the form: warning C4267: 'initializing' : conversion from 'size_t' to 'unsigned int', possible loss of data These are the problem lines: unsigned int count = s.size(); ar << make_nvp("count", const_cast<const unsigned int &>(count)); Changing the "unsigned int" to "size_t" fixes the warnings, like this: size_t count = s.size(); ar << make_nvp("count", const_cast<const size_t &>(count)); Trivial problem, I admit, but warnings are warnings... Cheers, -RSG

2 1

Boost thread library wrapper
by boostguy＠kapiladhanvi.com 10 Feb '06

10 Feb '06

Hi all This mail is in continuation to the earlier mail I had written asking for help in writting a thread wrapper for boost.threads. We already have an existing interface to which we would like our wrapper for boost.threads to adhere to. We have a runnable class which is inherited by the applications wanting to create threads and they override the run method. This is how our application currently creates threads. Thanks Peter for your earlier mail explaining on how to use boost::shared_ptr and boost::bind Your code for a Runnable interface > struct Runnable > { > virtual ~Runnable() {} > virtual void run() = 0; > }; We have a class Runnable with nearly the same code as above. To make a dummy implementation, I made the run methd a normal function. Nothing derives from it.. run is what I want to execute in a thread. > > void execute_in_thread( boost::shared_ptr<Runnable> pr ) > { > boost::thread th( boost::bind( &Runnable::run, pr ) ); > } I am creating a thread in MyThread class using the following code. --- MyThread.h ---- typedef boost::shared_ptr<Runnable> myPtr; class MyThread { public: MyThread( ); ~MyThread( ); private: myPtr p1; } --- MyThread.cpp ---- MyThread::MyThread( ) : p1( new Runnable() ) { cout << "MyThread::MyThread( ) constructor " << endl; boost::thread th( boost::bind( &Runnable::run, p1) ); } ---- Runnable.cpp --- // this is the only function in Runnable.h void Runnable::run() { cout << "Runnable::run() YUPPIE! "<< endl; } ------ end ----------- The code compiles, but nothing happens.. I am missing something here ?? Is there something else I need to do to start my thread?? Thanks Dhanvi

4 6