Mailman 3 April 2004 - Cplusplus-sig

Problems with pure virtual methods.
by Jeff Holle April 30, 2004

April 30, 2004

In using Boost.Python version 1_31_0 with pyste version 0.9.28, and gcc version 3.3, I've run into a problem with pure virtual methods. The problem is expressed a calling a base class method that employs an abstract method also defined in the same class. I'm providing an example that demonstrates the problem. The solution is to give the abstract method an implementation in the base class (making it just a virtual method). Note that this is far from desirable in that it precludes the use of abstract methods. The actual python runtime error can vary with this problem. With the provided example the error is: ReferenceError: Attempt to return dangling reference to object of type: std::string

1 0

Re: [C++-sig] Boost.Python and gcc 3.4
by Paul F. Kunz April 29, 2004

April 29, 2004

>>>>> On Thu, 29 Apr 2004 10:14:04 -0700, "Paul F. Kunz" <Paul_Kunz(a)SLAC.Stanford.EDU> said: >>>>> On Thu, 29 Apr 2004 09:48:07 -0700 (PDT), "Ralf W. Grosse-Kunstleve" <rwgk(a)yahoo.com> said: > I was thinking of re-compiling Python in /usr/local/ with 3.4, it > was compiled with 3.3.x in /usr/local/. Follow up on my own ppst. Recompiling Python with gcc 3.4.0 fixed both my problem with the SIP and Boost.Python modules. I think the problem might have been that Python was pulling in /usr/lib/libgcc_s.so while my compiled code needed /usr/local/lib/libgcc_s.so.

1 0

RE: [C++-sig] unnamed enum bug
by Roman Yakovenko April 28, 2004

April 28, 2004

Hi. > I guess Roman means to put a constant with the same value in > the global > scope, something like this (untested): > > { > scope unnamed; > > unnamed.attr("CHANNEL_DIRECTION_ENUM_SIZE") = > CHANNEL_DIRECTION_ENUM_SIZE; > ... > } This is almost exactly what I mean. scope().attr("CHANNEL_DIRECTION_ENUM_SIZE") = int( CHANNEL_DIRECTION_ENUM_SIZE ); The same logic as VarExporter.py. > I'm not sure if everything just "works", but if it does, I guess that > change is good enough to me. Note that normal enums are exported as > usual. Roman, have you tested this? How does it work? Yes I tested this and I use this patch in "production", also I quess that there is some case I failed to test. Small description: I implemented the same logic as in VarExporter.py. Also I export unnamed enums as int. The reason is simple: MSVC 7.1 could not compile scope().attr("CHANNEL_DIRECTION_ENUM_SIZE") = CHANNEL_DIRECTION_ENUM_SIZE; for some reason. > If everyone agress, I will apply this patch. I agree :-) I attach 3 files: dummy.h - file that includes diferent enums definitions EnumExporter.py - patched file diff.txt - the difference between files( CVS version and EnumExporter.py ) Roman

1 0

Class properties with policies?
by Tony Djordjevski April 27, 2004

April 27, 2004

Hi everyone, I'm fairly new to Boost.Python and I'm trying to set a return value policy for a class property. I'm using VC7.1 with Boost 1.31.0. A simplified example of what I'm trying to do is: // Node class definition class Node { private: Node *parent; public: Node(); ~Node(); Node* getParent(); void setParent(Node *newParent); }; // The following works fine, but not what I'm looking for BOOST_PYTHON_MODULE(MyModule) { class_< Node >("Node", init< >()) .def(init< const Node& >()) .def("getParent", &Node::getParent, return_value_policy< reference_existing_object >()) .def("setParent", &Node::setParent) ; } What I'd really like to do is something like this: BOOST_PYTHON_MODULE(MyModule) { class_< Node >("Node", init< >()) .def(init< const Node& >()) .add_property("parent", &Node::getParent, return_value_policy< reference_existing_object >(), &Node::setParent) ; } Is something like this possible? The compiler complains that I'm passing in the wrong amount of arguments and after looking at the source for class.hpp, I'm led to believe that it's currently not possible. So I guess my question really is "Is there an easy way around this?". On a side note: I realize that Pyste doesn't support the add_property feature. I don't have a lot of time, but I'd like to take a crack at adding it. If someone could point me in the right direction, I'd appreciate it. Thanks, Tony Djordjevski

2 2

Pyste and different interface files
by Franz Kaiser April 27, 2004

April 27, 2004

Hi all, I've two classes. A base class and a derived class in two different pyste files. SOMETIMES I get an error message like "Derived::function return a pointer or reference, but no policy is specified", but the policy is specified in the pyste file of the base class. The pyste tutorial says to declare the other pyste file with Import("Base.pyste") but that doesn't work. The error message is the same. The strange thing is I tried this with many classes but the error just occurs sometimes. Can anyone help? Franz

2 2

unnamed enum bug
by Roman Yakovenko April 27, 2004

April 27, 2004

Hi. I know that this issue was raised a few month ago. I want to come back to this issue. The reason is simple: I found bug in pyste. Description: in header file I have 2 or more unnamed enums. namespace shared_consts{ enum{ CHANNEL_DIRECTION_ENUM_SIZE = 3 }; enum{ CHANNEL_FIELD_TYPE_ENUM_SIZE = 3 }; } After generation I get next code enum_< py::UniqueInt<-1292127665> >("unnamed") .value("CHANNEL_DIRECTION_ENUM_SIZE", shared_consts::CHANNEL_DIRECTION_ENUM_SIZE) .export_values() ; enum_< py::UniqueInt<-1292127665> >("unnamed") .value("CHANNEL_FIELD_TYPE_ENUM_SIZE", shared_consts::CHANNEL_FIELD_TYPE_ENUM_SIZE) .export_values() ; As you can see there is an attempt to register same class enum_ twice, every time with different definition. I am working with boost 1.31 and pyste from CVS. ( I get assertion error in registry.cpp file, registry::insert function ) Why we could not export unnamed enums as constants ? This will also solve other small problem - repeating definition of "template<int num> struct UniqueInt" each time pyste meats unnamed enum. I already have patch for doing this. Roman

3 3

boost::python and gcc 3.4
by Nick Rasmussen April 26, 2004

April 26, 2004

Now that gcc 3.4 is enforcing the standard more strictly, what's the new way of doing: handle<> main_module(borrowed( PyImport_AddModule("__main__") )); (taken from the embedding section of the tutorial) Without triggering a copy constructor call on a boost::noncopyable object. Here's the relevant traffic from the gcc guys as to why the copy constructor's getting called: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=14618 http://gcc.gnu.org/bugs.html#cxx_rvalbind -nick

2 2

Boost.Python looks broken on debian/arm
by Domenico Andreoli April 26, 2004

April 26, 2004

[ i'm not subscribed to any of the two ml, please CC me as well as 245893(a)bugs.debian.org ] hi all, debian arm autobuilder fails to build the only two applications in main archive which make use Boost.Python, they are kdeedu [0] and regina-normal [1]. i'm trying to understand where the chain of errors reported in debian bug #245893 [2] really starts, problem is that i still have no simple program to reproduce this bug and not arm machine to play with. i'm asking to debian-arm mailing list some help reproducing the bug and to Boost.Python mailing list some help understanding the outcome. here follows what i think is the culprit error... /usr/include/boost/python/detail/def_helper.hpp: At global scope: /usr/include/boost/python/detail/def_helper.hpp: In instantiation of `boost::mpl::apply1<boost::python::detail::is_reference_to_member_function_pointer<boost::mpl::_1>, const char* const&>': /usr/include/boost/python/detail/def_helper.hpp:78: instantiated from `boost::python::detail::tuple_extract_base_select<boost::tuples::tuple<const char* const&, const boost::python::detail::not_specified&, const boost::python::detail::not_specified&, const boost::python::detail::not_specified&, boost::python::default_call_policies, boost::python::detail::keywords<0>, const char*, void (boost::python::detail::not_specified::*)(), boost::tuples::null_type, boost::tuples::null_type>, boost::python::detail::is_reference_to_member_function_pointer<boost::mpl::_1> >' /usr/include/boost/python/detail/def_helper.hpp:173: instantiated from `boost::python::detail::tuple_extract<boost::tuples::tuple<const char* const&, const boost::python::detail::not_specified&, const boost::python::detail::not_specified&, const boost::python::detail::not_specified&, boost::python::default_call_policies, boost::python::detail::keywords<0>, const char*, void (boost::python::detail::not_specified::*)(), boost::tuples::null_type, boost::tuples::null_type>, boost::python::detail::is_reference_to_member_function_pointer<boost::mpl::_1> >' /usr/include/boost/python/detail/def_helper.hpp:173: instantiated from `boost::python::detail::default_implementation_extract<boost::tuples::tuple<const char* const&, const boost::python::detail::not_specified&, const boost::python::detail::not_specified&, const boost::python::detail::not_specified&, boost::python::default_call_policies, boost::python::detail::keywords<0>, const char*, void (boost::python::detail::not_specified::*)(), boost::tuples::null_type, boost::tuples::null_type> >' /usr/include/boost/python/detail/def_helper.hpp:173: instantiated from `boost::python::detail::def_helper<const char*, boost::python::detail::not_specified, boost::python::detail::not_specified, boost::python::detail::not_specified>' /usr/include/boost/python/class.hpp:304: instantiated from `boost::python::class_<T, X1, X2, X3>& boost::python::class_<T, X1, X2, X3>::def(const char*, F) [with F = void (PythonOutputStream::*)(), T = PythonOutputStream, X1 = boost::noncopyable, X2 = boost::python::detail::not_specified, X3 = boost::python::detail::not_specified]' pythoninterpreter.cpp:89: instantiated from here /usr/include/boost/python/detail/def_helper.hpp:78: error: no class template named `apply' in `struct boost::python::detail::is_reference_to_member_function_pointer<boost::mpl::_1> ' any hint? thanks domenico [0] http://buildd.debian.org/fetch.php?&pkg=kdeedu&ver=4%3A3.2.2-1&arch=arm&sta… [1] http://buildd.debian.org/fetch.php?&pkg=regina-normal&ver=4.1.1-1&arch=arm&… [2] http://bugs.debian.org/245893 -----[ Domenico Andreoli, aka cavok --[ http://people.debian.org/~cavok/gpgkey.asc ---[ 3A0F 2F80 F79C 678A 8936 4FEE 0677 9033 A20E BC50

1 0

return_existing_py_object policy
by Nikolay Mladenov April 26, 2004

April 26, 2004

Hi, I made myself policy that searches for existing python object holding the returned from a function pointer or reference and then returns it to python. I am posting it in case anybody is interested. It is declared like that: template < class BackRef = BackReference , class FallbackPolicies = detail::default_fallback_policy , class Base = default_call_policies > struct return_existing_py_object; The BackRef parameter is the class that holds self pointer. So if you have class A{ //A has to be polymorphic virual void f(){} }; class A *func(); struct A_wrap:A { A_wrap(PyObject *s):self(s){} PyObject *self; }; class_<A,A_wrap>("A"); then you can do def("func", &func, return_existing_py_object<A_wrap>()); The FallbackPolicies parameter is used in case no python object object holding the result is found. The default is to raise an error. I also use it with return_internal_reference<>. The default BackReference class can bu used for defining wrapper classes, like struct A_wrap: A, BackReference { A_wrap(PyObject *s):BackReference(s){} }; in which case the def call can look like that: def("func", &func, return_existing_py_object<>()); -- Nikolay Mladenov Sitius Automation Inc. www.sitius.com

1 0

Can you send me example?
by Antonello April 26, 2004

April 26, 2004

Hi i'm a italian student. I must create a virtual machine which call python script from Visual C++ program. My teacher tell me use swig fror make this work. I use Swig for import in Python class, object, variables of C++, but my problem are when in C++ i must call script Python. I find many article, but i don't find a working example. Can you advise me how i call python script from C++? Have you example which can you send me (in my email)? (my script use only int, float, string and object, class). Thanks.

1 0