- Boost-users - lists.boost.org

Is this shared_ptr usage okay?
by Hickman, Greg 11 Nov '02

11 Nov '02

I'd like to prevent clients from directly creating and deleting instances of a type like 'Bill' below, so Bill's constructor and destructor have been made private to prevent ordinary client access. Since Bill has reference semantics, construction is handled by a static 'Bill::create()' factory method which returns a dynamically allocated Bill instance wrapped in a shared_ptr. Destruction is handled by a private Bill::Deallocator who also happens to be a Friend of Bill (FOB). My question is: Is this an acceptable use of the shared_ptr deallocator, or am I just asking for trouble? Thanks, Greg Hickman //-------------------------------------- #include <boost/shared_ptr.hpp> using namespace boost; class Bill; typedef shared_ptr<Bill> Bill_ptr; class Bill { private: Bill() {} ~Bill() {} struct Deallocator { void operator() (Bill* b) const { delete b; } }; friend struct Deallocator; public: static Bill_ptr create() { return Bill_ptr(new Bill, Deallocator()); } };

2 1

Re: uBLAS Matrix Multiplication
by Paul C. Leopardi 10 Nov '02

10 Nov '02

Hi, There are at least two strong arguments here: 1. Use * for element-by-element multiplication of matrices. This treats matrices as if they were just arrays, with an added "mult" function, but it has the advantage that it prevents or at least discourages users from writing poorly performing code. 2. Use * for matrix multiplication. This treats matrices as if they were elements of a linear algebra, with algebra operations "+" and "*". C++ allows such abstractions, but there may be a penalty in performance if the abstraction is abused. In GluCat - http://glucat.sf.net - I implemented Clifford algebras using "+" and "*". GluCat also has two other products, for which it uses, "&" and "^". I used "*" for the main product which defines the algebra. Underlying each Clifford algebra is an isomorphic matrix algebra. Best regards

1 0

Possible Error (with Fix?) in debug build of thread(?) library created with msvc/stlport.
by Mike Wilson 10 Nov '02

10 Nov '02

As a newbie I'd like to say a big thank to all those concerned for boost. I found that the following (MSVC6) program fails when in the (default) debug build, but works in release: #include <iostream> #include <boost/thread/thread.hpp> int main(int , char* ) { int NUM_CLIENTS = 3; boost::thread_group thrds; std::cout << NUM_CLIENTS; // prints 0 in _DEBUG build, 3 in release return 0; } It was linked using the libboost_thread.lib file created using the following bjam command: bjam "-sTOOLS=msvc-stlport" "-sSTLPORT_PATH=F:\TemplateStuff" "-sBUILD=debug <runtime-link>static <stlport-iostream>on" POSSIBLE FIX Compiling the above program with _STLP_DEBUG defined gives the correct answer. Is this the correct fix to the problem? (I've noted that there is no mention of _STLP_DEBUG in the distributed docs) Mike Wilson Thanks again to the boost community.

3 2

Q: Boost Python / weave / swig
by Claudius Schnörr 09 Nov '02

09 Nov '02

Hello, I would like to build a python-interface for a C++-template library I wrote. After looking around a bit, besides Boost-python I found swig and weave, which seem to have a similar purpose in mind!?! I would appreciate it if someone could provide some comparing hints/links/information about these various solutions (I'm new to all these packages). How do they differ in terms of - the learning curve, - their capabilities and - their foci ? Thank you in advance, Claudius

1 0

N-dimensional array besides multi_array
by Paul Cameron 09 Nov '02

09 Nov '02

I'm looking for an N-dimensional array that doesn't require the dimensionality be specified at compile time, like multi_array. Does anyone know of such a beast? It doesn't need a lot of functionality, just basic storage/retrieval and some slicing. Just thought I'd ask before I roll my own. I'm thinking of tweaking Giovanni Bavestrelli's implementation which is what multi_array is based on. Thanks, Paul

1 0

RE: [Boost-Users] Boost (shared_ptr) vs Visio 2002 reverse engineering
by Randy Bowen 09 Nov '02

09 Nov '02

I assume you meant, What's the matter with Visio, right? -----Original Message----- From: Marcio Del'Valle [mailto:mjdelvalle@terra.com.br] Sent: Friday, November 08, 2002 4:11 PM To: Boost-Users(a)yahoogroups.com Subject: [Boost-Users] Boost (shared_ptr) vs Visio 2002 reverse engineering Hi, Every time I try to reverse engineer a project where I've used boost::shared_ptr Visio 2002 locks my Visual C++. What's the matter with the shared_ptr library? Marcio Del'Valle [Non-text portions of this message have been removed] Yahoo! Groups Sponsor ADVERTISEMENT <http://rd.yahoo.com/M=238900.2578845.3984497.2273195/D=egroupweb/S=1705 006788:HM/A=1309024/R=0/*http://www.gotomypc.com/u/tr/yh/cpm/grp/300_you H1/g22lp?Target=mm/g22lp.tmpl> <http://us.adserver.yahoo.com/l?M=238900.2578845.3984497.2273195/D=egrou pmail/S=:HM/A=1309024/rand=607495686> Info: <http://www.boost.org> Wiki: <http://www.crystalclearsoftware.com/cgi-bin/boost_wiki/wiki.pl> Unsubscribe: <mailto:boost-users-unsubscribe@yahoogroups.com> Your use of Yahoo! Groups is subject to the Yahoo! Terms of Service <http://docs.yahoo.com/info/terms/> . [Non-text portions of this message have been removed]

2 1

Boost (shared_ptr) vs Visio 2002 reverse engineering
by Marcio Del'Valle 09 Nov '02

09 Nov '02

Hi, Every time I try to reverse engineer a project where Ive used boost::shared_ptr Visio 2002 locks my Visual C++. Whats the matter with the shared_ptr library? Marcio DelValle [Non-text portions of this message have been removed]

1 0

RE: [Boost-Users] Boost (shared_ptr) vs Visio 2002 reverse engineering
by Randy Bowen 09 Nov '02

09 Nov '02

Actually, we've played with the Visio reverse engineering stuff a bit, but it appears to choke on any moderately complex use of templates in C++. I would assume that this is the issue you're encountering with the use of the Boost libraries. Randy Bowen Software Architect Stamps.com -----Original Message----- From: Marcio Del'Valle [mailto:mjdelvalle@terra.com.br] Sent: Friday, November 08, 2002 5:03 PM To: Boost-Users(a)yahoogroups.com Subject: RE: [Boost-Users] Boost (shared_ptr) vs Visio 2002 reverse engineering Yep!! =) Sorry about that. But, after all, it's M$! Marcio Del'Valle -----Original Message----- From: Randy Bowen [mailto:rbowen@stamps.com] Sent: Friday, November 08, 2002 8:41 PM To: Boost-Users(a)yahoogroups.com Subject: RE: [Boost-Users] Boost (shared_ptr) vs Visio 2002 reverse engineering I assume you meant, What's the matter with Visio, right? -----Original Message----- From: Marcio Del'Valle [mailto:mjdelvalle@terra.com.br] Sent: Friday, November 08, 2002 4:11 PM To: Boost-Users(a)yahoogroups.com Subject: [Boost-Users] Boost (shared_ptr) vs Visio 2002 reverse engineering Hi, Every time I try to reverse engineer a project where I've used boost::shared_ptr Visio 2002 locks my Visual C++. What's the matter with the shared_ptr library? Marcio Del'Valle [Non-text portions of this message have been removed] Yahoo! Groups Sponsor ADVERTISEMENT < http://rd.yahoo.com/M=238900.2578845.3984497.2273195/D=egroupweb/S=1705 006788:HM/A=1309024/R=0/* http://www.gotomypc.com/u/tr/yh/cpm/grp/300_you H1/g22lp?Target=mm/g22lp.tmpl> < http://us.adserver.yahoo.com/l?M=238900.2578845.3984497.2273195/D=egrou pmail/S=:HM/A=1309024/rand=607495686> Info: < http://www.boost.org> Wiki: < http://www.crystalclearsoftware.com/cgi-bin/boost_wiki/wiki.pl> Unsubscribe: <mailto:boost-users-unsubscribe@yahoogroups.com> Your use of Yahoo! Groups is subject to the Yahoo! Terms of Service < http://docs.yahoo.com/info/terms/> . [Non-text portions of this message have been removed] Yahoo! Groups Sponsor ADVERTISEMENT <http://rd.yahoo.com/M=237459.2482214.3917349.2146399/D=egroupweb/S=1705 0067 88:HM/A=1267611/R=0/*http://ad.doubleclick.net/jump/N2524.Yahoo/B1071650 ;sz= 300x250;ord=1036798890310715?> Info: < http://www.boost.org> Wiki: < http://www.crystalclearsoftware.com/cgi-bin/boost_wiki/wiki.pl> Unsubscribe: <mailto:boost-users-unsubscribe@yahoogroups.com> Your use of Yahoo! Groups is subject to the Yahoo! Terms of Service <http://docs.yahoo.com/info/terms/> . _____ Esta mensagem foi verificada pelo E-mail Protegido Terra <http://www.emailprotegido.terra.com.br/> . Scan engine: VirusScan / Atualizado em 06/11/2002 / Versco: 1.3.13 Proteja o seu e-mail Terra: http://www.emailprotegido.terra.com.br/ [Non-text portions of this message have been removed] Info: <http://www.boost.org> Wiki: <http://www.crystalclearsoftware.com/cgi-bin/boost_wiki/wiki.pl> Unsubscribe: <mailto:boost-users-unsubscribe@yahoogroups.com> Your use of Yahoo! Groups is subject to the Yahoo! Terms of Service.

1 0

[BGL] Problem getting a property map from a reverse graph
by kweeheong 08 Nov '02

08 Nov '02

When I create a reverse_graph from a graph, is it possible to get the property map as well? I have reduced the code to the following snippet: typedef adjacency_list<vecS, vecS, bidirectionalS, property<vertex_name_t,string> > GraphType; typedef boost::property_map<GraphType,vertex_name_t>::type VertexNameMapType; typedef reverse_graph<GraphType> ReverseGraphType; typedef boost::property_map<ReverseGraphType,vertex_name_t>::type ReverseVertexNameMapType; GraphType g; VertexNameMapType nameMap=get(vertex_name,g); // get nameMap from graph g reverse_graph<GraphType> rg(g); // create the reverse_graph ReverseVertexNameMapType rnameMap=boost::get(vertex_name,rg); // get rnameMap frmo rg, but this line fails to compile The VC6 compile error message, with long substrings substituted for string ( STR ) and adjacency list ( ADJL ) looks like: graph\reverse_graph.hpp(241) : error C2664: '__thiscall boost::vec_adj_list_vertex_property_map<A,A*,S,S&,vertex_name_t>::boost::vec_adj_list_vertex_property_map<A,A*,S,S&,vertex_name_t><class boost::ADJL *>' : cannot convert parameter 1 from 'struct boost::vec_adj_list_vertex_property_map<class boost::ADJL,class boost::ADJL const *,class STR,class STR const &,enum boost::vertex_name_t>' to 'class boost::ADJL *' No user-defined-conversion operator available that can perform this conversion, or the operator cannot be called example\testreverse.cpp(97) : see reference to function template instantiation 'struct boost::vec_adj_list_vertex_property_map<A,A*,S,S&,vertex_name_t> __cdecl boost::get<enum boost::vertex_name_t,class boost::reverse_graph<class boost::ADJL,class boost::ADJL const &> &>' being compiled graph\reverse_graph.hpp(241) : error C2553: no legal conversion of return value to return type 'struct boost::vec_adj_list_vertex_property_map<A,A*,S,S&,vertex_name_t> *' example\testreverse.cpp(97) : see reference to function template instantiation 'struct boost::vec_adj_list_vertex_property_map<A,A*,S,S&,vertex_name_t> __cdecl boost::get<enum boost::vertex_name_t,class boost::reverse_graph<class boost::ADJL,class boost::ADJL const &> &>' being compiled I am trying to determine whether I have made the call incorrectly or is this a limitation of the reverse_graph?

1 0

[BGL SUGGESTION] about source() and target()
by mycia 08 Nov '02

08 Nov '02

to Boost Graph Library designers: in the boost/graph/graph_traits.hpp (line 104): //--begin---------------------------------------------------------------- // Since pair is in namespace std, Koenig lookup will find source and // target if they are also defined in namespace std. This is illegal, // but the alternative is to put source and target in the global // namespace which causes name conflicts with other libraries (like // SUIF). namespace std { template <class T, class G> T source(pair<T,T> p, const G&) { return p.first; } template <class T, class G> T target(pair<T,T> p, const G&) { return p.second; } } //--end---------------------------------------------------------------- why not change it with: //--begin---------------------------------------------------------------- template <class E, class G> typename boost::graph_traits<G>::vertex_descriptor source(E e, const G& g) { return boost::source(e,g); } template <class E, class G> typename boost::graph_traits<G>::vertex_descriptor target(E e, const G& g) { return boost::target(e,g); } //--end---------------------------------------------------------------- and add the following in boost/graph/vector_as_graph.hpp //--begin---------------------------------------------------------------- template <class T, class EdgeList, class Alloc> T source(std::pair<T,T> p, const std::vector<EdgeList, Alloc>&) { return p.first; } template <class T, class EdgeList, class Alloc> T target(std::pair<T,T> p, const std::vector<EdgeList, Alloc>&) { return p.second; } //---end--------------------------------------------------------------- which means, vector_as_graph handles source() and target() inside its class (and in boost namespace) and makes std::source() and std::target() more general the reason why I want to modify this is that the original design makes trouble when I try to develop a subgraph view adaptor (something like filtered_graph, but accept vector_as_graph type) I need to define my own source() and target() to deal with my own subgraph view class, as shown below: namespace boost { template <class G> typename subview_graph<G>::vertex_descriptor target(typename subview_graph<G>::edge_descriptor e, subview_graph<G>& v) {...} } however, when dealing with vector_as_graph (which means the 1st argument, edge_descriptor, will be std::pair<int,int>), the compiler complains ambiguity between std::target() and boost::target() in fact, it will make trouble to everyone who wants to define his/her own source() & target() function but the 1st argument is std::pair<T,T> if it's possible, please fix this strange design or any other opinions will be appreciated best regards, hcyeh

1 0