March 2008 - Boost-users - lists.preview.boost.org

Re: [Boost-users] object relational mapping library
by Pavlo Korzhyk 28 Mar '08

28 Mar '08

> Very cool indeed! > > However, from a quick glance at your tutorial, it seems you could have > folded this back into Boost.Serialization by writing a sqlite3 archive. > Is there some reason you chose not to do this? The main reason is the difference between "serialization" and "persistence". Boost.Serialization is not designed for "random access" to the data files. Consider a situation when you store several structs in one XML file. Serialization doesn't provide you any means to edit or at least extract only one specific element. All you can do is load the whole file, update some entries and write the whole set back to disc. You can build your own persistence solution on top of Boost.Serialization, but you have to do it yourself. And hiberlite is a persistene library - the main idea is to provide the smart pointers to objects inside the SQLite database. We really need some sort of SQL support in Serialization, but I don't see any good way to integrate features like primary keys, queries, tables metamodel, etc. So simply adding some sqlite output will look like a new file format, and that's not what RDBMS software is built for. But I agree, an idea of sqlite3_archive is very attractive, and also very challenging. http://code.google.com/p/hiberlite/ Yours sincerely, Pavlo Korzhyk ICQ#155870780

1 0

Need help with boost.variant and apply_visitor method overloading.
by Kevin Scarr 28 Mar '08

28 Mar '08

Hi I am trying to implement a system for ensuring producers and consumers agree on the data type they are exchanging. However, I also have "general" containers which can send or receive anything (mainly routing components). Anyway, I have tried to setup the following code to implement this scheme, however When I compile code (on MSVC 2005), I see the following error: Error 1 error C3066: there are multiple ways that an object of this type can be called with these arguments; unable to recover from previous error(s); stopping compilation c:\boost-1_34_1\boost\variant\detail\apply_visitor_binary.hpp 63 I have tracked it down to the first two overloads where I am trying to connect a consumer/producer of a specific type to a general producer/consumer. What am I doing wrong? The compiler obviously thinks that it can cast a (for example) G2Consumer<t1>* into a G2Consumer<DataTypes>*, so it cannot resolve which overload to use. How can I tell it not to cast?? Thanks for your help, Kevin typedef boost::variant<t1*,t2*,t3*> DataTypes; typedef boost::variant< G2Consumer<t1>*, G2Consumer<t2>*, G2Consumer<t3>*, G2Consumer<DataTypes>* > G2ConsumerTypes; typedef boost::variant< G2Producer<t1>*, G2Producer<t2>*, G2Producer<t3>*, G2Producer<DataTypes>* > G2ProducerTypes; class doConnectConsumerToProducer : public static_visitor<> { public: template<typename T> void operator() ( G2Consumer<DataTypes>* consumer, G2Producer<T>* producer ) const { consumer->RegisterAsConsumer( *producer ); } template<typename T> void operator() ( G2Consumer<T>* consumer, G2Producer<DataTypes>* producer ) const { consumer->RegisterAsConsumer( *producer ); } template<typename T> void operator() ( G2Consumer<T>* consumer, G2Producer<T>* producer ) const { consumer->RegisterAsConsumer( *producer ); } template<typename C, typename P> void operator() ( G2Consumer<C>* consumer, G2Producer<P>* producer ) const { G2Base* c = dynamic_cast<G2Base*>(consumer); G2Base* p = dynamic_cast<G2Base*>(producer); std::ostringstream msg; msg << "Incompatible consumer(" << typeid(*c).name() << ") and producer(" << typeid(*p).name() << ") types" << std::ends; throw G2::exception( msg.str().c_str() ); } }; // Helper function to drive this above. void connectConsumerToProducer( G2ConsumerTypes& c, G2ProducerTypes& p ) { apply_visitor( doConnectConsumerToProducer(), c, p ); }

2 1

variant cast operators?
by Daniel Oberhoff 28 Mar '08

28 Mar '08

Hi, I am playing with boost::variant to introduce a more elegant property system into my current project. I was wondering why there are no cast operatorst defined for the variant types, to avoid having to type get<>: boost::variant<double,int,string> v; string s = v; is there a basic reason why this is not possible? Best Daniel

2 1

[statechart] post_event() and Boost.Bind
by Commander Pirx 28 Mar '08

28 Mar '08

Currently I don't understand why the following code dosn't compile (with VC 8.0). I simply try to bind the post_event() method from the state_machine to a function object. To illustrate the problem I've added the following code to the Camera.hpp file of the Camera example from the Statechart library: #include <boost/function.hpp> #include <boost/bind.hpp> <snip> struct Camera : sc::state_machine< Camera, NotShooting > { // @test start typedef boost::function< void( const event_base_type& ) >process_event_fun; typedef boost::function< void( const event_base_type& ) > post_event_fun; process_event_fun process_event_fun_; post_event_fun post_event_fun_; Camera() { process_event_fun_ = boost::bind( &Camera::process_event, boost::ref( *this ), _1 ); // the following line doesn't compile with VC 8.0 // post_event_fun_ = boost::bind( &Camera::post_event, boost::ref( *this ), _1 ); } // @test end bool IsMemoryAvailable() const { return true; } bool IsBatteryLow() const { return false; } }; After uncommenting the marked line the compiler produce more than 200 error messages: A function template has too few arguments. But the line above compiles very well and the process_event() method has the same signature. The post_event() method is polymorphic. But this shouldn't be a problem. Should it?

2 2

[pointer_container] and std::back_inserter
by Jeff Flinn 28 Mar '08

28 Mar '08

Are pointer_container's intended to be compatible with std::back_inserter? I've also tried calling insert and assign. I've tried: #include <boost/ptr_container/ptr_vector.hpp> #include <iterator> #include <vector> #include <memory> int main() { boost::ptr_vector<int> int_ptrs; *std::back_inserter(int_ptrs) = new int(123); *std::back_inserter(int_ptrs) = std::auto_ptr<int>(new int(456)); std::vector<int*> ptrs(1, new int(123)); int_ptrs.insert(int_ptrs.end(), ptrs.begin(), ptrs.end()); return 0; } Thanks, Jeff

3 5

Boost Pool - error: segmentation fault
by Joost Meijles 28 Mar '08

28 Mar '08

Hi All, I am trying to use the Boost Pool library, more specifically I am trying to use the singleton pool interface. Everything seems to work fine, till the moment I use it to create an object from a class containing a (STL) map, which results in a segmentation fault. Does anyone know why this happens, or how to work around it? What I find also strange is that the constructor and destructor of the class doesn't seem to be called when using the pool. Is this normal behavior? Please find below the example that causes the segmentation fault (and the not calling of the constructor/destructor issue). Thanks a lot in advance. Kind regards, Joost ========================================== class MyClass { public: int x; vector y; map z; }; int main() { typedef boost::singleton_pool myPool; MyClass * const t = (MyClass *) myPool::malloc(); t->x = 1; t->y.push_back(2); t->z.insert(make_pair(3,3));//causes segmentation fault myPool::purge_memory(); } ========================================== _________________________________________________________________ Probeer Live Search: de zoekmachine van de makers van MSN! http://www.live.com/?searchOnly=true

4 5

Build CSR graph from a file in graphviz format
by Johan Oudinet 28 Mar '08

28 Mar '08

Hi, I read the csr-example.cpp that explain how to use compressed_sparse_row_graph and output it in graphviz format (.dot), and I'd like building a csr graph from a file in DOT format. So I tried using read_graphviz but actually it doesn't work with CSR graph (see below for my code). Does someone successfully build a CSR graph from a DOT file ? Regards, //===== csr.cc ===== # include <iostream> # include <string> # include <fstream> # include <boost/graph/compressed_sparse_row_graph.hpp> # include <boost/graph/iteration_macros.hpp> # include <boost/graph/graphviz.hpp> struct Node { std::string name; }; int main () { typedef boost::compressed_sparse_row_graph<boost::directedS, Node> graph_t; // Try to read the CSR graph g from foo.dot graph_t g; std::ifstream in("foo.dot", std::ios::in); boost::dynamic_properties dp; dp.property ("node_id", boost::get (&Node::name, g)); bool status = read_graphviz(in, g, dp, "node_id"); // doesn't work, see below for the error message BGL_FORALL_VERTICES (v, g, graph_t) std::cout << g[v].name << std::endl; BGL_FORALL_EDGES (e, g, graph_t) std::cout << g[source (e, g)].name << " -> " << g[target (e, g)].name << std::endl; } //================ The error message with g++ 4.1.3: /usr/include/boost/graph/graphviz.hpp:721: error: conversion from 'boost::csr_edge_descriptor<unsigned int, unsigned int>' to non-scalar type 'std::pair<boost::csr_edge_descriptor<unsigned int, unsigned int>, bool>' requested <8==== foo.dot ===== digraph { 0 -> 3; 0 -> 4; 1 -> 3; 1 -> 4; 2 -> 3; 2 -> 4; 3 -> 4; } ================= -- Johan Oudinet

1 0

nondynamic: A utility class that promotes safe programming
by Daniel Lidström 28 Mar '08

28 Mar '08

Hi there! (I recently posted a similar message to microsoft.public.vc.language, but the response was so bad I thought I would post here instead.) I have created a utility class that I think is useful, called nondynamic. The purpose is sort of like boost::noncopyable, but I disallow any dynamic/ pointer usages. For example, & and -> are disabled. Operator new/delete as well. Let me show nondynamic: // all members are private class nondynamic { // disallow operator new static void* operator new(std::size_t); static void* operator new(std::size_t, const std::nothrow_t&) throw(); static void* operator new[](std::size_t); static void* operator new[](std::size_t, const std::nothrow_t&) throw(); // disallow operator delete static void operator delete(void*){} static void operator delete(void*, const std::nothrow_t&) throw(){} static void operator delete[](void*){} static void operator delete[](void*, const std::nothrow_t&) throw(){} // disallow address-of const nondynamic* operator&() const; /* */ nondynamic* operator&() /* */; // disallow -> const nondynamic* operator->() const; /* */ nondynamic* operator->() /* */; }; Using this class is as simple as using boost::noncopyable: struct DataIO : private nondynamic { ... }; Now you are only able to use DataIO as a value type. All other usages are disabled. I think this promotes safe programming, in that pointer usage is minimized (it seems with pointers, new/delete/null is never far away). Let me illustrate how this class can be used most effectively. I thought it would be nice to use nondynamic when you have a class hierarchy, and a factory method. Here's the base class: struct IDataIO { virtual void Write(const std::string& str) = 0; virtual void Read(char* buffer, std::size_t size) = 0; }; typedef boost::shared_ptr<IDataIO> IDataIOPtr; A sample subclass: struct SerialDataIO : IDataIO { void Write(const std::string&) { std::cout << __FUNCTION__ << std::endl; } void Read(char*, std::size_t) { std::cout << __FUNCTION__ << std::endl; } }; Here's the wrapper class that inherits from nondynamic: struct DataIO : IDataIO, private nondynamic { DataIO(IDataIOPtr dataIO) : mDataIO(dataIO) {} void Write(const std::string& str) { mDataIO->Write(str); } void Read(char* buffer, std::size_t size) { mDataIO->Read(buffer, size); } private: IDataIOPtr mDataIO; }; And finally, the factory: DataIO CreateDataIO(int i) { if( i==0 ) return IDataIOPtr(new SerialDataIO); else throw std::runtime_error("Could not create DataIO"); } Now the clients of the class hierarchy uses the wrapper, and are completely isolated from pointers: DataIO dataIO = CreateDataIO(0); dataIO.Write("Hello"); char buffer[1024]; dataIO.Read(buffer, sizeof(buffer)); It should be impossible to misuse DataIO. What do you think about nondynamic? Is it implemented correctly? Named appropriately? Useful at all? Any comments gladly accepted :-) Thanks in advance! -- Daniel

2 1

[move] How to use BOOST_LVALUE_ASSIGN?
by Terry G 28 Mar '08

28 Mar '08

How are BOOST_LVALUE_ASSIGN and BOOST_LVALUE_ASSIGN2 used? How do I know which one to use? Is David Abrahams move.hpp the best library to use for move semantics? I noticed something similar under interprocess. Is there a version of unique_ptr that works with Dave's move.hpp? If not, does it make sense to create one? Its been several years since this was a topic of discussion here. Unfortunately, I didn't need to use it then. Now I do, and C++0X is still over the horizon for a few years for embedded systems. I'm not sure, but I think the compiler I'll be using is GCC 3.4 or later. terry

1 0

intrusive_ptr and incomplete types
by gast128 27 Mar '08

27 Mar '08

Dear all. I was googling about incomplete types and smart pointers but couldn't find a definitive statement about intrusive pointers. >From the header file of intrusive_ptr it seems that it can deal with incomplete types. The constructor and destructor use the 'intrusive_ptr_add_ref' or 'intrusive_ptr_release'. If these are not invoked, use of an intrusive_ptr with incomplete types should be ok. So it seems that below example will compile: //foo.h: #include <boost/intrusive_ptr.hpp> class A; class Foo { public: Foo(); ~Foo(); //leave the dtor out, and it won't work boost::intrusive_ptr<A> m_ptr; }; //bla.cpp: #include "foo.h" void F() { Foo foo; } Is this correct? Or is my (non confirming?) compiler giving me the wrong impression. A good article about incomplete types is 'The Standard Librarian: Containers of Incomplete Types' from Austern. He states that member function which are not referenced in the client code are not checked for incomplete types. For intrusive_ptr this would be the constructor and destructor. wkr, me

2 2