Mailman 3 March 2007 - Cplusplus-sig

Boost python and import error
by Valentino Volonghi March 11, 2007

March 11, 2007

hi all, I'm trying to use boost to write python extensions in C++ under OSX. I'm trying to use the most basic example of a greet function. The extension module is inside a reactor package and it's called _reactor. Unfortunately after a compilation (that successfully finishes) I get the following error: >>> from reactor import _reactor Traceback (most recent call last): File "<stdin>", line 1, in <module> File "reactor/__init__.py", line 11, in <module> from reactor import _reactor ImportError: dynamic module does not define init function (init_reactor) I've searched for this on google and found many answers but none worked. Actually after looking at boost/python.hpp source I think that init_reactor should be there since it's inserted by BOOST_PYTHON_MODULE macro. Do I really have to use boost.jam to compile this small example? (I ask because I tried hard to use it under osx last week and failed miserably, I simply can't use it and documentation is lacking at best). Thanks for information! Here is the full source: #include <boost/python.hpp> using namespace boost::python; char const* greet() { return "hello, world"; } BOOST_PYTHON_MODULE(_reactor) { def("greet", greet); } and here's the setup.py: #!/usr/bin/python import sys, os from os.path import join from distutils.core import setup from distutils.extension import Extension from distutils import sysconfig # break abstraction to set g++ as linker - is there better way? sysconfig._init_posix() compiler = "g++ -ftemplate-depth-50" #compiler += " -fkeep-inline-functions" # for debugging # workaround for bugs in Redhat 7.3 compiler if os.popen("g++ --version").read().strip() == "2.96": compiler += " -fno-inline " sysconfig._config_vars["CC"] = compiler if sys.platform == 'darwin': sysconfig._config_vars["LDSHARED"] = "g++ -undefined dynamic_lookup" else: sysconfig._config_vars["LDSHARED"] = "g++ -shared" ext_modules = [] include_dirs = ["./include", "/usr/local/include", "/opt/local/include/"] if os.environ.has_key("BOOST_INCLUDE"): include_dirs.append(os.environ["BOOST_INCLUDE"]) libraries = [] for libpath in (os.environ.get("LD_LIBRARY_PATH", "").split(":") + ["/usr/lib", "/lib", "/usr/local/lib", "/opt/local/lib/"]): paths = [os.path.join(libpath, libname) for libname in ("libboost_python.so", "libboost_python.dylib")] for path in paths: if os.path.exists(path): libraries.append("boost_python") break if not "boost_python" in libraries: libraries.append("boost_python-gcc-mt") ext_modules.append(Extension("reactor._reactor", [join("reactor", "%s.cpp" % name) for name in ("_reactor",)], libraries=libraries, library_dirs=["/usr/local/lib", "/opt/local/lib"], include_dirs=include_dirs)) setup( name = "reactor", packages=["reactor"], ext_modules=ext_modules, ) -- Valentino Volonghi aka Dialtone Now running MacOS X 10.4 TechBlog: http://vvolonghi.blogspot.com Home Page: http://www.twisted.it

2 2

polymorphic class question
by Tim Blechmann March 11, 2007

March 11, 2007

hi all, i'm using lots of polymorphic classes in a mixed c++/python environment. the biggest problem for me are noncopyable c++ classes, that are derived from python and stored (via pointers) in c++ data structures. my_class: boost::noncopyable {}; in order to keep the reference, i introduced the following helper class: template <class T> struct self_containing_wrapper: public T { self_containing_wrapper(PyObject * self): self(self) { } virtual ~self_containing_wrapper(void){} PyObject * self; }; which is used to wrap classes in the following way: class_<my_class, self_containing_wrapper<my_class>, boost::noncopyable>("my_class"); now i'm trying to write a c++-to-python converter in order to get the correct reference when wrapping functions that return pointers to instances of my_class: template <class T> struct self_containing_wrapper_converter { static PyObject* convert(T const & x) { self_containing_wrapper<T> const & wrapper = static_cast<self_containing_wrapper<T> const &>(x); return wrapper.self; } }; to be used like: to_python_converter<my_class, self_containing_wrapper_converter<my_class> >(); now i have two questions: - is this correct or do i need to take care in terms of reference counting? - when wrapped c++ functions return my_class*, will i have to specify the return_value_policy? if so, what policy should i use? thanks in advance ... tim -- tim(a)klingt.org ICQ: 96771783 http://tim.klingt.org Which is more musical, a truck passing by a factory or a truck passing by a music school? John Cage

2 2

operator<< problems
by Piotr Jaroszynski March 10, 2007

March 10, 2007

Hello, I have written a simple hello.cpp - [1], which doesn't compile - [2]. After some "investigation" I have tracked down the problem to boost/python/str.hpp header. Removing str.hpp and numeric.hpp, which includes it, from python.hpp make the hello.cpp compile and work just fine: >>> import hello >>> blah = hello.World() >>> blah.set("foo") >>> print blah foo System info: boost=1.33.1, gcc-4.1.2, glibc-2.5, kernel 2.6.20-gentoo x86_64, python-2.4.4 [1] = http://dev.gentooexperimental.org/~peper/boost/hello.cpp [2] - http://dev.gentooexperimental.org/~peper/boost/hello.error -- Best Regards, Piotr Jaroszynski

2 2

Memory issues
by Paul Guse March 8, 2007

March 8, 2007

Hello. I have a few boost::python issues I need help with . Firstly, I'm getting some memory leaks by just including the boost libraries. (Dynamically linked dll, VS2005. Using boost to embed python in c++) >From what I gather this is still an outstanding issue with no, as of yet, solution? Is that correct? Second, this is also causing some issues as I need to free the resources allocated during the running of scripts. Specifically, I'm using the intrusive_ptr to wrap my own object's reference counting. When these objects are referenced in a python script the reference count is incremented, but not decremented when the script ends. As expected, this is causing some issues. Is there a way to clean up the variables allocated in a namespace? Or to effectively clear the main namespace? Thanks. Paul

2 2

passing python derived instance back to C++
by Kirill Lapshin March 7, 2007

March 7, 2007

I've spent couple of hours trying to diagnose quite a basic problem, finally solved it, and just for reference documenting it here, since my googling around did not reveal any solution. Hopefully it will save someone a bit of time. Suppose you have abstract C++ interface Base properly exposed to python via wrapper, etc. Suppose you have C++ function taking pointer to Base. Now suppose you want to implement Base interface in Python and pass it to C++ function. // C++ struct Base { virtual ~Base(){} virtual void foo() =0; }; void bar(Base * b) { } // Python class Derived(module.Base): def __init__(self): pass def foo(): print 'Derived.foo' module.bar(Derived()) At this point call to module.bar will fail complaining that it can't find matching C++ function. It turns out that solution is quite obvious -- just call base class constructor in Derived constructor, i.e. class Derived(module.Base): def __init__(self): module.Base.__init__(self) def foo(): print 'Derived.foo'

1 0

C++ assertion
by Simon Wood March 6, 2007

March 6, 2007

I am in the process of wrapping a C++ class using Boost.Python. Things are progressing well, however the underlying C++ library uses assert() for various error checks. Out of curiousity I triggerred one of the asserts() on an exported C++ function from within Python. Sure enough the assertion error printed (o.k.) AND the Python interpreter quit (not o.k.). So, is there a way to implement some Python code to capture the underlying C++ assertions, or do I have to modify the underlying library code? -Simon

2 2

module destructor?
by Neal Becker March 5, 2007

March 5, 2007

It is easy to put some code to run when a module is loaded: BOOST_PYTHON_MODULE(foo) { do_some_init(); ... But how can I do some cleanup when module is unloaded?

2 3

c++ to python object conversion problem
by Tim Blechmann March 4, 2007

March 4, 2007

hi all, i've wrapped a simple c++ class with boost python, which looks like this: class simple: boost::noncopyable{}; an instance of this class (or a derived one) is allocated from python and a pointer is stored in a class with a virtual function: class callback: boost::noncopyable { virtual void overloadme(simple const &) = 0 void run(void) { this->overloadme(s_); } void set(simple * s) {s_ = s;} simple * s_; }; this callback class is wrapped like it's described in the tutorial: class cbwrapper: callback, wrapper<callback> { virtual void overloadme(simple const & s) { this->get_override("overloadme")(s); } }; callback::run is called from the c++ code later in the program asynchronously. however, when trying to call the overridden function, the program crashes: > 0 PyErr_Restore PC=0xb7f30a9f, FP=0xb45461c8 [/usr/lib/libpython2.4.so.1.0] 1 PyErr_SetObject PC=0xb7f30b30, FP=0xb45461c8 [/usr/lib/libpython2.4.so.1.0] 2 _ZNK5boost6python9converter12registration9to_pythonEPVKv PC=0xb519c4d6, FP=0xb45461c8 [/usr/lib/libboost_python-mt.so.1.33.1] 3 _ZN5boost6python9converter6detail18arg_to_python_baseC2EPVKvRKNS1_12registrationE PC=0xb51accdf, FP=0xb45461e8 [/usr/lib/libboost_python-mt.so.1.33.1] 4 boost::python::converter::detail::value_arg_to_python<simple>::not-in-charge value_arg_to_python PC=0xb678648b, FP=0xb4546208 [/usr/include/boost/python/converter/arg_to_python.hpp#209] 5 boost::python::converter::arg_to_python<simple>::arg_to_python PC=0xb67864b5, FP=0xb4546228 [/usr/include/boost/python/converter/arg_to_python.hpp#256] 6 boost::python::override::operator ()<simple> PC=0xb67864e0, FP=0xb4546268 [/usr/include/boost/python/override.hpp#138] the instance of the wrapped object is still valid, but the conversion from the c++ object to the python object seems to be the reason of the crash ... any idea, what i am missing? thanks, tim -- tim(a)klingt.org ICQ: 96771783 http://tim.klingt.org Just what the hell is the experimental tradition? Morton Feldman

2 5

boost::python and custom smart pointer
by Gabriel Becedillas March 2, 2007

March 2, 2007

Hi, I'd like to have my custom smart_ptr behave like boost::shared_ptr regarding from_python conversions. To_python converstions work fine, but the problem arises when I have an instance to a derived class via a smart_ptr to the base class MySmartPtr<Base> and this has to be passed to a function receiving MySmartPtr<Derived>. If I use boost::shared_ptr this works just fine. I think I'm doing something wrong in MySmartPtr_from_python. I'm pasting a simplified version of my code: ------ BEGIN C++ CODE ------ // Dummy smart pointer class. template <typename T> class MySmartPtr { public: MySmartPtr() : m_P(NULL) { } MySmartPtr(T* a_P) : m_P(a_P) { } T* m_P; }; namespace boost { namespace python { template <typename T> T* get_pointer(MySmartPtr<T> const& a_P) { return a_P.m_P; } template <typename T> struct pointee< MySmartPtr<T> > { typedef T type; }; } // namespace python } // namespace boost template <class T> struct MySmartPtr_from_python { static void* convertible(PyObject* p) { if (p == Py_None) return p; return boost::python::converter::get_lvalue_from_python(p, boost::python::converter::registered< MySmartPtr<T> >::converters); } static void construct(PyObject* source, boost::python::converter::rvalue_from_python_stage1_data* data) { void* const storage = ((boost::python::converter::rvalue_from_python_storage<MySmartPtr<T> >*)data)->storage.bytes; if (data->convertible == source) // This is because convertible returned p if p was Py_None. new (storage) MySmartPtr<T>(); else new (storage) MySmartPtr<T>( *((MySmartPtr<T>*)data->convertible) ); data->convertible = storage; } }; class Base { public: virtual ~Base() {} }; class Derived : public Base { }; boost::shared_ptr<Base> build_ok() { return boost::shared_ptr<Base>(new Derived()); } MySmartPtr<Base> build_fail() { return MySmartPtr<Base>(new Derived()); } void test_ok(boost::shared_ptr<Derived> a_Derived) {} void test_fail(MySmartPtr<Derived> a_Derived) {} template <typename T> void register_MySmartPtr_conversions() { boost::python::register_ptr_to_python< MySmartPtr<Base> >(); boost::python::converter::registry::insert( &MySmartPtr_from_python<T>::convertible, &MySmartPtr_from_python<T>::construct, boost::python::type_id< MySmartPtr<T> >() ); } BOOST_PYTHON_MODULE(bptest) { boost::python::class_<Base, boost::noncopyable> class_Base("Base", boost::python::no_init); boost::python::register_ptr_to_python< boost::shared_ptr<Base> >(); register_MySmartPtr_conversions<Base>(); boost::python::class_<Derived, boost::noncopyable, boost::python::bases<Base> > class_Derived("Derived", boost::python::no_init); boost::python::register_ptr_to_python< boost::shared_ptr<Derived> >(); register_MySmartPtr_conversions<Derived>(); boost::python::def("build_ok", build_ok); boost::python::def("test_ok", test_ok); boost::python::def("build_fail", build_fail); boost::python::def("test_fail", test_fail); } ------ END C++ CODE ------ This python code fails to work: import bptest bptest.test_ok(bptest.build_ok()) bptest.test_fail(bptest.build_fail()) And this is the error that I get: Boost.Python.ArgumentError: Python argument types in bptest.test_fail(Derived) did not match C++ signature: test_fail(class MySmartPtr<class Derived>) Thanks in advance. -- Gabriel Becedillas Developer CORE SECURITY TECHNOLOGIES

2 4

Re: [C++-sig] boost.python metaclass problem
by Ralf W. Grosse-Kunstleve March 2, 2007

March 2, 2007

> I don't know pyqt, and I'm not sure you get access to boost.python's > own metaclass. (David ?) Simply use the __class__ attribute of a Boost.Python wrapped class. I.e. if the class is "bar" (and you construct objects with "bar()"), bar.__class__ will give you the metaclass. See also: boost/libs/python/doc/tutorial/doc/tutorial.qbk Look for: __metaclass__ But, like Hans I'm also a bit skeptical ... Ralf ____________________________________________________________________________________ We won't tell. Get more on shows you hate to love (and love to hate): Yahoo! TV's Guilty Pleasures list. http://tv.yahoo.com/collections/265

2 1