Mailman 3 October 2007 - Cplusplus-sig

No registered converter was able to produce a C++ rvalue...
by Neal Becker Oct. 10, 2007

Oct. 10, 2007

I know this comes up frequently, but this one has me baffled. u = array1_complex_int ((1,2)) TypeError: No registered converter was able to produce a C++ rvalue of type std::complex<int> from this Python object of type int But there is a converter, it can be invoked from the python side: >>> complex_int (2) <blitz.complex_int object at 0x2aaaaab69db8> or like this: u = array1_complex_int (complex_int(e) for e in (1,2)) It seems that extract<std::complex<int> > is not finding the conversion. Any hints? The conversion is there, because there is this minimal class_: bp::class_<std::complex<int> > ("complex_int", bp::init<int,optional<int> >()) ;

1 0

interaction of boost::shared_ptr with boost::python
by Neal Becker Oct. 9, 2007

Oct. 9, 2007

A bit puzzled about this one. struct X { X() { std::cout << "X()" << '\n'; } ~X() { std::cout << "~X()" << '\n'; } }; boost::shared_ptr<X> make_x () { return boost::shared_ptr<X> (new X); } boost::shared_ptr<X> make_x_from_shared (boost::shared_ptr<X> x) { return boost::shared_ptr<X> (new X); } X* clone1 (X const& x) { return new X; } boost::shared_ptr<X> clone2 (X const& x) { return boost::shared_ptr<X> (new X); } BOOST_PYTHON_MODULE (stuff) { class_<X> ("X", no_init) .def ("__init__", make_constructor (make_x)) .def ("__init__", make_constructor (make_x_from_shared)) .def ("__copy__", &clone1, return_value_policy<manage_new_object>()) .def ("__copy2__", &clone2) ; def ("clone1", &clone1, return_value_policy<manage_new_object>()); def ("clone2", &clone2); >>> x = X() >>> clone2(x) Traceback (most recent call last): File "<stdin>", line 1, in <module> TypeError: No to_python (by-value) converter found for C++ type: boost::shared_ptr<X>

1 0

Re: [C++-sig] R: Re: R: Re: Boost.Python type registry
by Ralf W. Grosse-Kunstleve Oct. 8, 2007

Oct. 8, 2007

Would it help you if you had an interface to extract an array of type_info objects known to the registry? If you send patches with docs I'll review them. BTW: What you need seems very unusual. Could you explain the situation? Did you consider alternative approaches? Ralf ----- Original Message ---- From: François Duranleau <duranlef(a)iro.umontreal.ca> To: jumpjoe(a)fastwebnet.it; Development of Python/C++ integration <c++-sig(a)python.org> Sent: Friday, October 5, 2007 8:22:20 AM Subject: Re: [C++-sig] R: Re: R: Re: Boost.Python type registry On Fri, 5 Oct 2007, jumpjoe(a)fastwebnet.it wrote: >> I think that you can search the registry for entry with m_class_object >> that matches your PyObject's type or search the lvalue_chain(s) for a >> converter that returns !=0. But there is no api for that so far. > > This is another interesting approach, but after a few tries, I > discovered the registry is really hard to get by, as it is a static > variable defined in a "hidden" function in an anonymous namespace in the > registry.cpp file. Any suggestions on how to extract it? As an "extern" > of some kind? Without modifying the source code, there is no way. The only way to search the registry externally is via the boost::python::converter::query function, but then you need to provide a type_info, and to do this exhaustively is not really simpler than using an "inverse" registry like I mentionned. -- Francois Duranleau _______________________________________________ C++-sig mailing list C++-sig(a)python.org http://mail.python.org/mailman/listinfo/c++-sig

3 2

Multiple-inheritance problem
by Piotr Jaroszyński Oct. 6, 2007

Oct. 6, 2007

Hello, Seems that when one creates a Python class, which multiple-inherits from classes exposed by boost.python then it can be only converted(casted) to the first class in the inheritance list. Simple example code follows. C++: #include <boost/python.hpp> namespace bp = boost::python; struct A { }; struct B { }; void test_A(const A &) { } void test_B(const B &) { } BOOST_PYTHON_MODULE(multiple_inheritance) { bp::def("test_A", &test_A); bp::def("test_B", &test_B); bp::class_<A>("A", bp::init<>()); bp::class_<B>("B", bp::init<>()); } Python: class C(A, B): pass test_A(C()) test_B(C()) And I get: Traceback (most recent call last): File "./multiple_inheritance.py", line 10, in <module> test_B(C()) Boost.Python.ArgumentError: Python argument types in multiple_inheritance.test_B(C) did not match C++ signature: test_B(B) Am I missing something or is it a known problem? Any workaround if so? -- Best Regards, Piotr Jaroszyński

1 0

R: Re: R: Re: Boost.Python type registry
by jumpjoe＠fastwebnet.it Oct. 5, 2007

Oct. 5, 2007

>I think that you can search the registry for entry with m_class_object >that matches >your PyObject's type or search the lvalue_chain(s) for a converter that >returns !=0. >But there is no api for that so far. This is another interesting approach, but after a few tries, I discovered the registry is really hard to get by, as it is a static variable defined in a "hidden" function in an anonymous namespace in the registry.cpp file. Any suggestions on how to extract it? As an "extern" of some kind? Cheers, Giovanni Beltrame

2 1

Re: [C++-sig] inner class pickle issue
by Ralf W. Grosse-Kunstleve Oct. 4, 2007

Oct. 4, 2007

> Or only the top level class can be pickled? Yes. Try this: import pickle class outer(object): class inner(object): pass oi = outer.inner() pickle.dumps(oi) You'll get a similar error message: pickle.PicklingError: Can't pickle <class '__main__.inner'>: it's not found as __main__.inner Ralf

1 0

inner class pickle issue
by Grant Oct. 4, 2007

Oct. 4, 2007

I am trying to write pickle support for an inner class by using boost.pyhton. But it always fail with this error message: >>> pickle.dump(k,mm) Traceback (most recent call last): File "<stdin>", line 1, in <module> File "/usr/lib64/python2.5/pickle.py", line 1362, in dump Pickler(file, protocol).dump(obj) File "/usr/lib64/python2.5/pickle.py", line 224, in dump self.save(obj) File "/usr/lib64/python2.5/pickle.py", line 331, in save self.save_reduce(obj=obj, *rv) File "/usr/lib64/python2.5/pickle.py", line 400, in save_reduce save(func) File "/usr/lib64/python2.5/pickle.py", line 295, in save self.save_global(obj) File "/usr/lib64/python2.5/pickle.py", line 748, in save_global (obj, module, name)) pickle.PicklingError: Can't pickle <class 'libpyUtils2.KaiserBessel'>: it's not found as libpyUtils2.KaiserBessel Can I pickle an inner class? Or only the top level class can be pickled? Grant

1 0

R: Re: Boost.Python type registry
by jumpjoe＠fastwebnet.it Oct. 4, 2007

Oct. 4, 2007

>AFAIK, there is nothing in the API to help you there. Boost.Python doesn't >work directly this way internally. It provides a way to test if an >instance holds a given type, but nothing to return the contained type >info. Is there at least a way to get a raw pointer to the wrapped object from a PyObject? I think extract<> is doing it. Cheers, Giovanni Beltrame

3 2

Extansion compilation in Linux
by Paul Vicenti Oct. 4, 2007

Oct. 4, 2007

Hi, I'm moving a project from Windows and I'm having problems to compile. I get a: 'undefined reference to 'PyThreadState_Get' and whatever I use in my code of Python library. In windows I include Python.h and it links with python library to resolve all symbols. Here, it looks that the include file is found (otherwise, I expect 'cannot open file Python.h') but the library seems not to be linked. Any help for this? thanks, paul

2 1

Boost.Python type registry
by Giovanni Beltrame Oct. 3, 2007

Oct. 3, 2007

Hello everybody, I am trying to understand how the Boost.Python type registry works... What I want to do is, given a python object (let's say a PyObject*), determine its wrapper C++ type_info. Theoretically, what I want to do is going name->PyObject->C++ type. Notice that the traditional extract<T> won't work for me as I want to determine which type is being wrapped. I know Boost.Python is doing this internally in some way, but I haven't figured it out yet. Can anybody help? Thanks in advance, -- Giovanni Beltrame TEC-EDM Research Fellow European Space Agency

2 1