From ndbecker2 at gmail.com  Thu Jun  1 16:44:16 2006
From: ndbecker2 at gmail.com (Neal Becker)
Date: Thu, 01 Jun 2006 10:44:16 -0400
Subject: [C++-sig] make_constructor + with_custodian_and_ward
Message-ID: <e5mug3$sjo$1@sea.gmane.org>

Any ideas on this?  B will hold an 'A const&'.  I think I need
'with_custodian_and_ward' to ensure the 'A' object stays alive.  This code
won't compile:

#include <boost/python/module.hpp>
#include <boost/python/def.hpp>
#include <boost/python/class.hpp>
#include <boost/python/init.hpp>
#include <boost/python/make_constructor.hpp>
#include <boost/python/with_custodian_and_ward.hpp>
using namespace boost::python;

struct A {};
struct B {
  B (A const& _a) : a (_a) {}
  A const& a;
};

B* make_B (A const& a) {
  return new B (a);
}

BOOST_PYTHON_MODULE(test)
{
  class_<B> ("B", no_init)
    .def ("__init__", make_constructor (make_B,
                                        with_custodian_and_ward_postcall<0,1>(),
                                        (arg ("A"))));
}
[...]
/usr/local/src/boost.cvs/boost/python/with_custodian_and_ward.hpp:86: error:
invalid initialization of reference of type ?PyObject* const&? from
expression of type ?const boost::python::detail::offset_args<PyObject*,
mpl_::int_<1> >?
/usr/local/src/boost.cvs/boost/python/detail/caller.hpp:50: error: in
passing argument 1 of ?unsigned int boost::python::detail::arity(PyObject*
const&)?



From yamokosk at ufl.edu  Thu Jun  1 21:00:50 2006
From: yamokosk at ufl.edu (J.D. Yamokoski)
Date: Thu, 01 Jun 2006 15:00:50 -0400
Subject: [C++-sig] trouble initializing module using Boost
Message-ID: <447F3962.8030506@ufl.edu>

I have run into a compile error which is probably very easy to fix. I am 
trying to call PyImport_AppendInittab() to add the initialization of a 
module I have created with Boost. The exact details are of the call are:

// Register the extension module
if ( PyImport_AppendInittab("shapes", initshapes) == -1 ) {
		throw python_exception( "Failed to add shapes to the 	interpreter's 
builtin modules" );
}

// Initialize the Python interpreter
Py_Initialize();
....

I believe the module is defined correctly - I have used the module in 
the Python command line, just now trying to use it through the C-API. 
Here is the definition:

---- BoxModule.h ----
#include <boost/python.hpp>
#include "Box.h"
using namespace boost::python;

---- BoxModule.cpp ----
#include "BoxModule.h"

BOOST_PYTHON_MODULE(shapes)
{
    class_<Box>("MyBox")
	.add_property("height", &Box::getHeight, &Box::setHeight)
	.add_property("width",	&Box::getWidth,	&Box::setWidth)
	.add_property("length", &Box::getLength, &Box::setLength)
	.add_property("volume", &Box::getVolume, &Box::setVolume);
}

And finally, the compile error message is:

error C2065: 'initshapes' : undeclared identifier

So is initshapes put in a namespace? Or how do I pass this function to 
PyImport_AppendInittab()?


From dfaure at klaralvdalens-datakonsult.se  Fri Jun  2 19:10:20 2006
From: dfaure at klaralvdalens-datakonsult.se (David Faure)
Date: Fri, 2 Jun 2006 19:10:20 +0200
Subject: [C++-sig] Renaming "<string>" filename for PyRun_String
Message-ID: <200606021910.21015.dfaure@klaralvdalens-datakonsult.se>

Victor Nakoryakov <nail-mail at mail.ru> wrote:
> I use PyRun_String to run script stored in some file. I can't use
> PyRun_File because of different layouts of FILE struct in my project and
> in python lib. So once I get error I see something like:
>
> Traceback (most recent call last):
>    File "<string>", line 14, in ?
> NameError: ...
>
> Is it possible somehow to rename "<string>" in this message to the real filename?

I just had the same problem, and I solved it by copying some code out of the pythonrun.c sources...

[This means it would be nice for the python developers to provide a PyRun_StringFileName().
Just like it would be nice to have a public function that calls tb_printinternal, parse_syntax_error, etc
in order to print out all the info about a parsing or runtime error just like python itself does.]

Anyway; this code does the job for me:

#include <node.h> // from python

// Copied from pythonrun.c
static PyObject *run_node(struct _node *n, const char *filename, PyObject *globals, PyObject *locals, PyCompilerFlags *flags) {
    PyCodeObject *co;
    PyObject *v;
    co = PyNode_CompileFlags(n, filename, flags);
    PyNode_Free(n);
    if (co == NULL)
        return NULL;
    v = PyEval_EvalCode(co, globals, locals);
    Py_DECREF(co);
    return v;
}

// This is missing from python: a PyRun_String that also takes a filename, to show in backtraces
static PyObject* MyPyRun_StringFileName(const char *str, const char* filename, int start, PyObject *globals, PyObject *locals) {
    struct _node* n = PyParser_SimpleParseString(str, start);
    if (!n) return 0;
    return run_node( n, filename, globals, locals, 0 );
}


-- 
David Faure -- faure at kde.org, dfaure at klaralvdalens-datakonsult.se
KDE/KOffice developer, Qt consultancy projects
Klar?lvdalens Datakonsult AB, Platform-independent software solutions


From ndbecker2 at gmail.com  Sat Jun  3 02:31:03 2006
From: ndbecker2 at gmail.com (Neal Becker)
Date: Fri, 02 Jun 2006 20:31:03 -0400
Subject: [C++-sig] Can developer find true happiness with shared_ptr?
Message-ID: <e5ql8c$n86$1@sea.gmane.org>

Continuing the quest,

I'm working on a framework for signal processing.  The plan is to create
objects in python and then compose them.  The components are (in general)
polymorphic.  From python the components can be configured and controlled. 
Also, components can be retrieved back to python from the composites.

I've been doing some experiments with boost::shared_ptr.  It appears that if
components are always stored as shared_ptr, and always passed (by value) as
shared_ptr, that life is easy.

For example:

struct cnt {
  cnt (int _n) : n (_n) {}
  int n;
};

struct A {
  A (boost::shared_ptr<cnt> _c) : c (_c) {}
  boost::shared_ptr<cnt> c;
  int get() const { return c->n; }
  void inc() { c->n++; }
};

using namespace boost::python;

BOOST_PYTHON_MODULE (testA) {
  class_<cnt> ("cnt", init<int>());
  class_<A> ("A", init<boost::shared_ptr<cnt> >())
    .def ("get", &A::get)
    .def ("inc", &A::inc)
    .def_readwrite ("c", &A::c)
    ;
}
>>> c = cnt (0)
>>> a = A(c)
>>> b = A(c)
>>> a.c
<testA.cnt object at 0x2aaaaab4b730>
>>> b.c
<testA.cnt object at 0x2aaaaab4b730>

It appears that this design meets all the requirements.  Not demonstrated
above, but also polymorphism works as expected.  Note that here only
polymorphism of components written in c++ is required, and that is all I
tested.  I did not test or require that components could be subclassed in
python.

Just wondering if I am on the right path here?  In particular, it seems
there is no need for non-default call or return policies to make this work,
which makes life a lot simpler.



From dave at boost-consulting.com  Sun Jun  4 00:27:44 2006
From: dave at boost-consulting.com (David Abrahams)
Date: Sat, 03 Jun 2006 18:27:44 -0400
Subject: [C++-sig] trouble initializing module using Boost
References: <447F3962.8030506@ufl.edu>
Message-ID: <umzctvpr3.fsf@boost-consulting.com>

"J.D. Yamokoski" <yamokosk at ufl.edu> writes:

> I have run into a compile error which is probably very easy to
> fix. I am trying to call PyImport_AppendInittab() to add the
> initialization of a module I have created with Boost. The exact
> details are of the call are:
>
> // Register the extension module
> if ( PyImport_AppendInittab("shapes", initshapes) == -1 ) {
> 		throw python_exception( "Failed to add shapes to the 	interpreter's 
> builtin modules" );
> }
>
> // Initialize the Python interpreter
> Py_Initialize();
> ....
>
> I believe the module is defined correctly - I have used the module in 
> the Python command line, just now trying to use it through the C-API. 
> Here is the definition:
>
> ---- BoxModule.h ----
> #include <boost/python.hpp>
> #include "Box.h"
> using namespace boost::python;
>
> ---- BoxModule.cpp ----
> #include "BoxModule.h"
>
> BOOST_PYTHON_MODULE(shapes)
> {
>     class_<Box>("MyBox")
> 	.add_property("height", &Box::getHeight, &Box::setHeight)
> 	.add_property("width",	&Box::getWidth,	&Box::setWidth)
> 	.add_property("length", &Box::getLength, &Box::setLength)
> 	.add_property("volume", &Box::getVolume, &Box::setVolume);
> }
>
> And finally, the compile error message is:
>
> error C2065: 'initshapes' : undeclared identifier
>
> So is initshapes put in a namespace? Or how do I pass this function to 
> PyImport_AppendInittab()?

You have to make a declaration visible before you can use it.  Unless
your PyImport_AppendInittab is in BoxModule.cpp after
BOOST_PYTHON_MODULE(shapes), you have to write

   extern "C" initshapes();

somewhere.
        
-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com



From roman.yakovenko at gmail.com  Sun Jun  4 20:59:57 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Sun, 4 Jun 2006 21:59:57 +0300
Subject: [C++-sig] Announcement: pygccxml/pyplusplus 0.8.0
Message-ID: <7465b6170606041159u1a254c28l67fe1d0d0a62c117@mail.gmail.com>

I'm glad to announce the new version of pygccxml and pyplusplus packages.

Next people contributed to the development:

    Matthias Baas
    Allen Bierbaum
    Lakin Wecker
    Georgiy Dernovoy

Thank you!

List of changes:

1. We started to use Subversion

2. Query API has been moved from pyplusplus to pygccxml. This change is fully
    backward compatible. Query API has been documented:

    http://language-binding.net/pygccxml/query_interface.html

3. pygccxml improves support for template instantiations. This is very important
    feature. Because, before this feature was introduced GCC-XML reported
    wrong names for classes and functions. Now the latest CVS version of GCCXML
    reports class name, mangled and demangled strings. pygccxml uses
    demangled string for class name.

4. pygccxml.declarations.class_t now has a list of all aliases(
typedefs) to the
    class.

    pyplusplus will use the list to guess class alias, in case the class is
    template instantiation.

5. pyplusplus has been improved a lot:

    It is knows how to export reference and pointer member variables.
    Few internal changes, that simplifies the implementation and in a near
    future those changes will allow us to introduce two very important features:

    1. call wrappers policy
        http://language-binding.net/pyplusplus/peps/call_wrapper_policies.html

    2. almost fully automated solution for std containers
        http://language-binding.net/pyplusplus/peps/indexing_suite.html

    3. Not all declarations, could be exported to Python. pyplusplus
will explain you
        why and may be will provide a link to full explanation.

     4. Not all function could be overridden in Python. pyplusplus
will explain you why.

6. Bug fixes. A lot of bugs have been fixed.

7. Documentation has been improved.

You can find more information about projects here:
http://language-binding.net/index.html

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From roman.yakovenko at gmail.com  Sun Jun  4 21:42:14 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Sun, 4 Jun 2006 22:42:14 +0300
Subject: [C++-sig] Announcement: pyboost package - first release
Message-ID: <7465b6170606041242o5c43147br999fcce9de62a7d4@mail.gmail.com>

I'm glad to announce the first release of pyboost package.

pyboost exports to Python functionality from next libraries:

        * random - a complete system for random number generation
        * date_time - date time library designed to provide a basis
for performing
                              efficient time calculations
        * crc - cyclic redundancy code computation objects
        * rational - rational number class

pyboost has good unit tests. Almost all unit tests from the libraries have been
re-written in Python. All libraries have some sample programs. All
those programs has been rewritten in Python too.

The libraries were exposed to Python using pyplusplus.

Binaries for Windows and Linux, Python 2.4 are avaiable.

You can find more information here:
http://language-binding.net/pyplusplus/examples/boost/boost.html

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From roman.yakovenko at gmail.com  Mon Jun  5 06:47:35 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Mon, 5 Jun 2006 07:47:35 +0300
Subject: [C++-sig] Announcement: pyboost package - first release
In-Reply-To: <7465b6170606041242o5c43147br999fcce9de62a7d4@mail.gmail.com>
References: <7465b6170606041242o5c43147br999fcce9de62a7d4@mail.gmail.com>
Message-ID: <7465b6170606042147n66f8895u8019df462d972de6@mail.gmail.com>

Hi.

There is a problem with Linux version -  the compressed file is corrupted.
I will upload new file this evening.

Sorry for inconvenience.

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From roman.yakovenko at gmail.com  Mon Jun  5 20:55:21 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Mon, 5 Jun 2006 21:55:21 +0300
Subject: [C++-sig] Announcement: pyboost package - first release
In-Reply-To: <7465b6170606042147n66f8895u8019df462d972de6@mail.gmail.com>
References: <7465b6170606041242o5c43147br999fcce9de62a7d4@mail.gmail.com>
	<7465b6170606042147n66f8895u8019df462d972de6@mail.gmail.com>
Message-ID: <7465b6170606051155y5009affexfcb936662b9b58ef@mail.gmail.com>

Hi. I just uploaded new Linux binaries of pyboost package.

Comments and suggestions are welcome.

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From rwgk at yahoo.com  Tue Jun  6 17:53:14 2006
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Tue, 6 Jun 2006 08:53:14 -0700 (PDT)
Subject: [C++-sig] Announcement: pyboost package - first release
In-Reply-To: <7465b6170606051155y5009affexfcb936662b9b58ef@mail.gmail.com>
Message-ID: <20060606155314.44659.qmail@web31502.mail.mud.yahoo.com>

--- Roman Yakovenko <roman.yakovenko at gmail.com> wrote:

> Hi. I just uploaded new Linux binaries of pyboost package.
> 
> Comments and suggestions are welcome.

I was hoping to get the pyplusplus-generated source code, but I can only find
.py and .so files in the linux zip file. Is there some place I can see the
generated C++ files?

Thanks!

Cheers,
        Ralf


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 


From johnbaronreid at netscape.net  Tue Jun  6 22:46:14 2006
From: johnbaronreid at netscape.net (John Reid)
Date: Tue, 06 Jun 2006 21:46:14 +0100
Subject: [C++-sig] Boost::Signals
Message-ID: <e64ph5$atl$1@sea.gmane.org>

Hi,

I'm trying to call a python function from C++. I pass the function to 
C++ as a boost::python::object. I can call functions with no arguments, 
when I try to pass a C++ created object as an argument I get the 
following error:

Traceback (most recent call last):
   File "python/test_python.py", line 8, in ?
     d.connect_new_deal_slot_py_2( f )
TypeError: No to_python (by-value) converter found for C++ type: class 
module::classname

Now the boost.python section (http://tinyurl.com/n3cpv) in the python 
wiki answers the following question:
"Is it is possible to convert pointers to existing classes to PyObjects* 
and then be able to pass an existing instance of an object directly to 
and from python?"

where the last paragraph says:

"If you really want to pass pointers around, it's certainly possible to 
tell the library to build a Python object around the pointer, but then 
you need to make sure the lifetime of the C++ object being referenced by 
the pointer extends past the lifetime of all Python references to the 
object or your program will crash."

This is what I would like to do but cannot find this elsewhere in the 
documentation. Can anyone let me know how to do this?

Thanks in advance,
John.



From johnbaronreid at netscape.net  Tue Jun  6 23:33:24 2006
From: johnbaronreid at netscape.net (John Reid)
Date: Tue, 06 Jun 2006 22:33:24 +0100
Subject: [C++-sig] Boost::Signals
In-Reply-To: <e64ph5$atl$1@sea.gmane.org>
References: <e64ph5$atl$1@sea.gmane.org>
Message-ID: <e64s9j$kke$1@sea.gmane.org>

John Reid wrote:
> Hi,
> 
> I'm trying to call a python function from C++. I pass the function to 
> C++ as a boost::python::object. I can call functions with no arguments, 
> when I try to pass a C++ created object as an argument I get the 
> following error:
> 
> Traceback (most recent call last):
>    File "python/test_python.py", line 8, in ?
>      d.connect_new_deal_slot_py_2( f )
> TypeError: No to_python (by-value) converter found for C++ type: class 
> module::classname
> 
> Now the boost.python section (http://tinyurl.com/n3cpv) in the python 
> wiki answers the following question:
> "Is it is possible to convert pointers to existing classes to PyObjects* 
> and then be able to pass an existing instance of an object directly to 
> and from python?"
> 
> where the last paragraph says:
> 
> "If you really want to pass pointers around, it's certainly possible to 
> tell the library to build a Python object around the pointer, but then 
> you need to make sure the lifetime of the C++ object being referenced by 
> the pointer extends past the lifetime of all Python references to the 
> object or your program will crash."
> 
> This is what I would like to do but cannot find this elsewhere in the 
> documentation. Can anyone let me know how to do this?
> 
> Thanks in advance,
> John.

Maybe it is better if I post a small example of what I'd like to do. 
Here is the C++...


#pragma warning( push )
# pragma warning( disable : 4099 )
# pragma warning( disable : 4267 )
# include <boost/python.hpp>
#pragma warning( pop )
#include <boost/signals.hpp>

using namespace boost::python;
using namespace boost;

struct test
	: noncopyable
{
	signal< void ( object ) > _signal;
	void connect_slot_py( object const & slot )
	{
		_signal.connect( slot );

		//call just for testing
		_signal( object( this ) );
	}
};

BOOST_PYTHON_MODULE( test_bp )
{
	class_< test, noncopyable >( "test" )
         .def( "connect_slot", &test::connect_slot_py )
	    ;
	//register_ptr_to_python< test * >();
}

  and here is the python...

import test_bp

def f( arg ):
     print 'In "f"', arg

t = test_bp.test()
t.connect_slot( f )



but I still get this error...

Traceback (most recent call last):
   File "<stdin>", line 1, in ?
TypeError: No to_python (by-value) converter found for C++ type: struct test



Any help appreciated,
John.



From roman.yakovenko at gmail.com  Wed Jun  7 06:25:17 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Wed, 7 Jun 2006 07:25:17 +0300
Subject: [C++-sig] Announcement: pyboost package - first release
In-Reply-To: <20060606155314.44659.qmail@web31502.mail.mud.yahoo.com>
References: <7465b6170606051155y5009affexfcb936662b9b58ef@mail.gmail.com>
	<20060606155314.44659.qmail@web31502.mail.mud.yahoo.com>
Message-ID: <7465b6170606062125k2301cca7h466918b86b6b26fa@mail.gmail.com>

On 6/6/06, Ralf W. Grosse-Kunstleve <rwgk at yahoo.com> wrote:
> --- Roman Yakovenko <roman.yakovenko at gmail.com> wrote:
>
> > Hi. I just uploaded new Linux binaries of pyboost package.
> >
> > Comments and suggestions are welcome.
>
> I was hoping to get the pyplusplus-generated source code, but I can only find
> .py and .so files in the linux zip file. Is there some place I can see the
> generated C++ files?

I will upload the whole pyboost project( pyplusplus scripts and
generated files ) to
SourceForge in few days.

I will let you know.

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From jesus24b at gmail.com  Wed Jun  7 18:25:29 2006
From: jesus24b at gmail.com (jesus garcia)
Date: Wed, 7 Jun 2006 18:25:29 +0200
Subject: [C++-sig] Exception Translation
Message-ID: <f24bfcdb0606070925i6f9362c1xffc28b5989849606@mail.gmail.com>

I like export the follow simple program

#define REPORT_ERROR(nerr,ErrorMsg) throw error(nerr,ErrorMsg)

where error is a class

Class FileName : public string {
...
Void compose (const string &str, int no, const string &ext)
   *this=(FileName) str;
   if (no!=-1) {
      char aux_str[5];
      sprintf(aux_str,"%05d",no);
      *this += aux_str;
   }

   if (ext!="") *this += "."+ext;
   else REPORT_ERROR(1005,"compose: need a extension");
}
...

I no have any idea for export de class string in C++ to the class
string in Python and the error I see the follow page
http://www.boost.org/libs/python/doc/tutorial/doc/html/python/exception.html
that shows how catch the error but I do it and the problem persist,
someone can help,

I put the solution in a file who contains the BOOST_PYTHON_MODULE
macro like this follow lines"

#include <boost/python.hpp>
using namespace boost::python;

struct PodBayDoorException;

void brack(PodBayDoorException const& x)
{PyErr_SetString(PyExc_UserWarning, x);}

extern void port2python_func();

BOOST_PYTHON_MODULE(_nulo) {
register_exception_translator<PodBayDoorException>(brack);
port2python_func();
...
}

Thanks,


From roman.yakovenko at gmail.com  Wed Jun  7 20:19:13 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Wed, 7 Jun 2006 21:19:13 +0300
Subject: [C++-sig] Announcement: pyboost package - first release
In-Reply-To: <7465b6170606062125k2301cca7h466918b86b6b26fa@mail.gmail.com>
References: <7465b6170606051155y5009affexfcb936662b9b58ef@mail.gmail.com>
	<20060606155314.44659.qmail@web31502.mail.mud.yahoo.com>
	<7465b6170606062125k2301cca7h466918b86b6b26fa@mail.gmail.com>
Message-ID: <7465b6170606071119w57145c8bh5307f4d460be791f@mail.gmail.com>

On 6/7/06, Roman Yakovenko <roman.yakovenko at gmail.com> wrote:
> On 6/6/06, Ralf W. Grosse-Kunstleve <rwgk at yahoo.com> wrote:
> > I was hoping to get the pyplusplus-generated source code, but I can only find
> > .py and .so files in the linux zip file. Is there some place I can see the
> > generated C++ files?
>
> I will upload the whole pyboost project( pyplusplus scripts and
> generated files ) to SourceForge in few days.

Good evening. I just upload almost all source file to SourceForge for pyboost
package.

You will not be able to create binaries, because I did not upload environment.py
script, that contains configuration details:
    gccxml location
    boost location
    python location

The main reason for this, is too much work to change location of
environment.py file,
within pyboost. I plan to move pyboost to be stand alone project.

Download page:
http://sourceforge.net/project/showfiles.php?group_id=118209

pyboost home page:
http://www.language-binding.net/pyplusplus/examples/boost/boost.html

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From amohr at pixar.com  Wed Jun  7 22:22:00 2006
From: amohr at pixar.com (Alex Mohr)
Date: Wed, 07 Jun 2006 13:22:00 -0700
Subject: [C++-sig] Overloading & keyword args bug?
Message-ID: <44873568.10004@pixar.com>

Hi folks,

Can someone tell me what's wrong with this?  As is, this code crashes in 
the example below (anytime I try to pass keyword arguments to the first 
overload).  If I reverse the order of the lines marked with ########, 
everything seems to work fine.

Thanks!

Alex


#include <boost/python.hpp>
#include <string>
#include <cstdio>

using namespace boost::python;
using std::string;

static void f1(string const &a1, int x, int y) {
     printf("f1 with %s, %d, %d\n", a1.c_str(), x, y);
}

static void f2(string const &a1, string const &a2, int x, int y) {
     printf("f2 with %s, %s, %d, %d\n", a1.c_str(), a2.c_str(), x, y);
}

BOOST_PYTHON_MODULE(overload_kw) {
     def("f", f1, (arg("x")=3, arg("y")=4)); // ########
     def("f", f2, (arg("x")=5, arg("y")=6)); // ########
}

---------------------------------------

 >>> from overload_kw import *
 >>> f('hello')
f1 with hello, 3, 4
 >>> f('hello', 'goodbye')
f2 with hello, goodbye, 5, 6
 >>> f('hello', 1, 2)
f1 with hello, 1, 2
 >>> f('hello', x=1)
Segmentation fault


From rwgk at yahoo.com  Thu Jun  8 03:44:03 2006
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Wed, 7 Jun 2006 18:44:03 -0700 (PDT)
Subject: [C++-sig] Overloading & keyword args bug?
In-Reply-To: <44873568.10004@pixar.com>
Message-ID: <20060608014403.58291.qmail@web31506.mail.mud.yahoo.com>

--- Alex Mohr <amohr at pixar.com> wrote:
> #include <boost/python.hpp>
> #include <string>
> #include <cstdio>
> 
> using namespace boost::python;
> using std::string;
> 
> static void f1(string const &a1, int x, int y) {
>      printf("f1 with %s, %d, %d\n", a1.c_str(), x, y);
> }
> 
> static void f2(string const &a1, string const &a2, int x, int y) {
>      printf("f2 with %s, %s, %d, %d\n", a1.c_str(), a2.c_str(), x, y);
> }
> 
> BOOST_PYTHON_MODULE(overload_kw) {
>      def("f", f1, (arg("x")=3, arg("y")=4)); // ########
>      def("f", f2, (arg("x")=5, arg("y")=6)); // ########
> }

I am surprised any of your tests works.
AFAIK you need to make several changes:

BOOST_PYTHON_FUNCTION_OVERLOADS(f1_overloads, f1, 1, 3)
BOOST_PYTHON_FUNCTION_OVERLOADS(f2_overloads, f2, 2, 4)

you have to give your C++ functions default arguments.

you have to define all arg(), i.e something like

  def("f", f1, f1_overloads((arg("a1"), arg("x")=3, arg("y")=4)));

At least this is what I do all the time.


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 


From roman.yakovenko at gmail.com  Thu Jun  8 07:40:03 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Thu, 8 Jun 2006 08:40:03 +0300
Subject: [C++-sig] Announcement: pyboost package - first release
In-Reply-To: <20060606155314.44659.qmail@web31502.mail.mud.yahoo.com>
References: <7465b6170606051155y5009affexfcb936662b9b58ef@mail.gmail.com>
	<20060606155314.44659.qmail@web31502.mail.mud.yahoo.com>
Message-ID: <7465b6170606072240s159dc917g5e4185bd03063710@mail.gmail.com>

On 6/6/06, Ralf W. Grosse-Kunstleve <rwgk at yahoo.com> wrote:
> I was hoping to get the pyplusplus-generated source code, but I can only find
> .py and .so files in the linux zip file. Is there some place I can see the
> generated C++ files?

I almost forgot, pyplusplus has simple GUI, that will allow you to see
generated code
without learning API. There are few users that does not use API at
all. You don't
have to install anything ( except gccxml, pygccxml and pyplusplus ).

http://www.language-binding.net/pyplusplus/tutorials/pyplusplus_demo.png

http://www.language-binding.net/pyplusplus/tutorials/tutorials.html

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From roman.yakovenko at gmail.com  Thu Jun  8 07:44:32 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Thu, 8 Jun 2006 08:44:32 +0300
Subject: [C++-sig] Distutils: setup script for binary files
In-Reply-To: <7465b6170606060502g74cba971ua41241e8e3a1acdd@mail.gmail.com>
References: <7465b6170606060502g74cba971ua41241e8e3a1acdd@mail.gmail.com>
Message-ID: <7465b6170606072244ld10cca0ya883b88fe8c539d2@mail.gmail.com>

Hi.

I asked the next question on Python list and did not get the answer.
May be I will
have more luck here.

I want to create setup script, that will install compiled extension module
plus few binaries the extension module depends on.

For example: I have package X:

X:
    __init__.py
    _x_.dll ( or so )
    some other dll's ( so's ) _x_.dll depends on.

I took a look on Python documentation,
http://docs.python.org/dist/describing-extensions.html, but it only
describes how to create
setup script for extension module from source files.

I think, I can treat _x_.dll as it was regular Python script. My
problem is, that on Linux
I should put "some other so's" in some directory, program loader can find.
How can I do this?

Any help is appreciated.

Thanks

--
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From amohr at pixar.com  Thu Jun  8 18:57:10 2006
From: amohr at pixar.com (Alex Mohr)
Date: Thu, 08 Jun 2006 09:57:10 -0700
Subject: [C++-sig] Overloading & keyword args bug?
In-Reply-To: <20060608014403.58291.qmail@web31506.mail.mud.yahoo.com>
References: <20060608014403.58291.qmail@web31506.mail.mud.yahoo.com>
Message-ID: <448856E6.4070406@pixar.com>

> I am surprised any of your tests works.
> AFAIK you need to make several changes:
> 
> BOOST_PYTHON_FUNCTION_OVERLOADS(f1_overloads, f1, 1, 3)
> BOOST_PYTHON_FUNCTION_OVERLOADS(f2_overloads, f2, 2, 4)
> 
> you have to give your C++ functions default arguments.

Why is this?  Suppose I don't want the functions to have default args in 
C++ -- only in python.  Regardless, since I'm providing default values 
for the keyword arguments at def() time, I think boost.python will 
always have to call my function with a full set of arguments.

> you have to define all arg(), i.e something like
> 
>   def("f", f1, f1_overloads((arg("a1"), arg("x")=3, arg("y")=4)));
> 
> At least this is what I do all the time.

I don't think that is (or should be) required.  Suppose I want some 
positional-only arguments.  According to the documentation for def, you 
may specify no more keyword arguments than the arity of f, so it seems 
like if you specify fewer, then they are taken as the last C++ arguments.

Note that if I remove one of the overloads (one of the def lines) then 
everything works as expected.

Perhaps David can shed some light.

Alex


From rwgk at yahoo.com  Thu Jun  8 19:22:20 2006
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Thu, 8 Jun 2006 10:22:20 -0700 (PDT)
Subject: [C++-sig] Announcement: pyboost package - first release
In-Reply-To: <7465b6170606072240s159dc917g5e4185bd03063710@mail.gmail.com>
Message-ID: <20060608172220.87029.qmail@web31502.mail.mud.yahoo.com>

--- Roman Yakovenko <roman.yakovenko at gmail.com> wrote:

> On 6/6/06, Ralf W. Grosse-Kunstleve <rwgk at yahoo.com> wrote:
> > I was hoping to get the pyplusplus-generated source code, but I can only
> find
> > .py and .so files in the linux zip file. Is there some place I can see the
> > generated C++ files?
> 
> I almost forgot, pyplusplus has simple GUI, that will allow you to see
> generated code
> without learning API. There are few users that does not use API at
> all. You don't
> have to install anything ( except gccxml, pygccxml and pyplusplus ).
> 
> http://www.language-binding.net/pyplusplus/tutorials/pyplusplus_demo.png
> 
> http://www.language-binding.net/pyplusplus/tutorials/tutorials.html

I was mainly interested in seeing if you have a better __hash__ method for your
rational type than the one I came up with, but I see you haven't gotten to that
yet. A few years ago I wanted to use rational number instances as Python
dictionary keys. In the rush of time I threw in a quick-and-dirty solution
which unfortunately persists to this day:

  http://phenix-online.org/cctbx_sources/boost_adaptbx/rational_ext.cpp

If anyone knows how to implement the hash function in a better way I'd be
grateful for advice.

Cheers,
        Ralf


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 


From christophe.pradal at cirad.fr  Thu Jun  8 18:39:13 2006
From: christophe.pradal at cirad.fr (Christophe Pradal)
Date: Thu, 08 Jun 2006 18:39:13 +0200
Subject: [C++-sig] Distutils: setup script for binary files
In-Reply-To: <7465b6170606072244ld10cca0ya883b88fe8c539d2@mail.gmail.com>
References: <7465b6170606060502g74cba971ua41241e8e3a1acdd@mail.gmail.com>
	<7465b6170606072244ld10cca0ya883b88fe8c539d2@mail.gmail.com>
Message-ID: <448852B1.9000104@cirad.fr>

Roman Yakovenko a ?crit :
> Hi.
>
> I asked the next question on Python list and did not get the answer.
> May be I will
> have more luck here.
>
> I want to create setup script, that will install compiled extension module
> plus few binaries the extension module depends on.
>
> For example: I have package X:
>
> X:
>     __init__.py
>     _x_.dll ( or so )
>     some other dll's ( so's ) _x_.dll depends on.
>
> I took a look on Python documentation,
> http://docs.python.org/dist/describing-extensions.html, but it only
> describes how to create
> setup script for extension module from source files.
>
> I think, I can treat _x_.dll as it was regular Python script. My
> problem is, that on Linux
> I should put "some other so's" in some directory, program loader can find.
> How can I do this?
>   
There is no good solution. What we do is to install binary dependencies 
as data.
Other solutions is to install first on the system the other so, then 
your package.
> Any help is appreciated.
>
> Thanks
>
> --
> Roman Yakovenko
> C++ Python language binding
> http://www.language-binding.net/
> _______________________________________________
> C++-sig mailing list
> C++-sig at python.org
> http://mail.python.org/mailman/listinfo/c++-sig
>
>   


-- 
Christophe Pradal

Equipe INRIA/CIRAD/INRA Virtual Plants - UMR AMAP
TA40/PSII, Boulevard de la Lironde
34398 MONTPELLIER CEDEX 5, France

http://www-sop.inria.fr/virtualplants
tel   : (33) 4 67 61 75 63         fax : (33) 4 67 61 56 68
email : christophe.pradal at cirad.fr  http://amap.cirad.fr




From rwgk at yahoo.com  Thu Jun  8 19:26:24 2006
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Thu, 8 Jun 2006 10:26:24 -0700 (PDT)
Subject: [C++-sig] Overloading & keyword args bug?
In-Reply-To: <448856E6.4070406@pixar.com>
Message-ID: <20060608172624.51396.qmail@web31505.mail.mud.yahoo.com>

--- Alex Mohr <amohr at pixar.com> wrote:

> > I am surprised any of your tests works.
> > AFAIK you need to make several changes:
> > 
> > BOOST_PYTHON_FUNCTION_OVERLOADS(f1_overloads, f1, 1, 3)
> > BOOST_PYTHON_FUNCTION_OVERLOADS(f2_overloads, f2, 2, 4)
> > 
> > you have to give your C++ functions default arguments.
> 
> Why is this?  Suppose I don't want the functions to have default args in 
> C++ -- only in python.

I'd simply write a thin C++ wrapper which supplies the default arguments.


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 


From amohr at pixar.com  Thu Jun  8 19:45:31 2006
From: amohr at pixar.com (Alex Mohr)
Date: Thu, 08 Jun 2006 10:45:31 -0700
Subject: [C++-sig] Overloading & keyword args bug?
In-Reply-To: <20060608172624.51396.qmail@web31505.mail.mud.yahoo.com>
References: <20060608172624.51396.qmail@web31505.mail.mud.yahoo.com>
Message-ID: <4488623B.9010600@pixar.com>

Ralf W. Grosse-Kunstleve wrote:
> --- Alex Mohr <amohr at pixar.com> wrote:
> 
>>> I am surprised any of your tests works.
>>> AFAIK you need to make several changes:
>>>
>>> BOOST_PYTHON_FUNCTION_OVERLOADS(f1_overloads, f1, 1, 3)
>>> BOOST_PYTHON_FUNCTION_OVERLOADS(f2_overloads, f2, 2, 4)
>>>
>>> you have to give your C++ functions default arguments.
>> Why is this?  Suppose I don't want the functions to have default args in 
>> C++ -- only in python.
> 
> I'd simply write a thin C++ wrapper which supplies the default arguments.

Clearly that's potentially a way to work around this, but it seems 
silly.  Another workaround is to swap the order of the defs.

But really, if it works with only one of the overloads, why shouldn't it 
work with both?  This really seems like a bug to me.

Alex


From roman.yakovenko at gmail.com  Thu Jun  8 20:07:48 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Thu, 8 Jun 2006 21:07:48 +0300
Subject: [C++-sig] Distutils: setup script for binary files
In-Reply-To: <448852B1.9000104@cirad.fr>
References: <7465b6170606060502g74cba971ua41241e8e3a1acdd@mail.gmail.com>
	<7465b6170606072244ld10cca0ya883b88fe8c539d2@mail.gmail.com>
	<448852B1.9000104@cirad.fr>
Message-ID: <7465b6170606081107r8338e6eh4a28f7f7cd3f4814@mail.gmail.com>

On 6/8/06, Christophe Pradal <christophe.pradal at cirad.fr> wrote:
> There is no good solution. What we do is to install binary dependencies
> as data.
> Other solutions is to install first on the system the other so, then
> your package.

Thanks, I though I missed something very trivial. Actually there is a
better way:
ask user where he want to put those files.

Here is relevant links how to achieve this:
http://docs.python.org/dist/node29.html

And example
http://tinyurl.com/jcvma
See doc_cmd class.

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From sebastian.redl at getdesigned.at  Fri Jun  9 14:27:20 2006
From: sebastian.redl at getdesigned.at (Sebastian Redl)
Date: Fri, 09 Jun 2006 14:27:20 +0200
Subject: [C++-sig] [Boost.Python] Extracting indirect base.
Message-ID: <44896928.20103@getdesigned.at>

Hi,

I've got a hand-wrapped plugin interface hierarchy for my program
allowing implementing plugins in Python. This looks approximately like this:

class Plugin
{
public:
  virtual void init() = 0;
  virtual PluginCategory getCategory() = 0; // Subtype of the plugin
  // ...
};

class InputPlugin : public Plugin
{
public:
  virtual void loadData(const string &resource, DataBag &target) = 0;
};

class OutputPlugin : public Plugin
{
public:
  virtual void saveData(const string &resource, const DataBag &source) = 0;
};

These are the plugin interfaces. PluginCategory is an enum identifying
the concrete subtype of the plugin, i.e. either Input or Output.

I've got them wrapped like this:

class PluginWrap : public Plugin, public boost::python::wrapper<Plugin>
{
  virtual void init() { this->get_override("init")(); }
  // ...
};

class InputPlugin : public InputPlugin, public
boost::python::wrapper<InputPlugin>
{
  virtual void loadData(...) { ... }
};

// ...

BOOST_PYTHON_MODULE(myprogram)
{
  class_<PluginWrap, boost::noncopyable>("Plugin")
    .def("init", pure_virtual(&Plugin::init))
    .def( ... )
    ;

  class_<InputPluginWrap, bases<PluginWrap>,
boost::noncopyable>("InputPlugin")
    .def("loadData", ...)
    ;

  // ...
}

A plugin file looks like this:

from myprogram import *

class SomePlugin(InputPlugin):
  def init(self):
    // ...

  // ...

plugins = [SomePlugin()]

I then attempt to load these modules like this:

PythonLoader::load(const path &file)
{
  // Parse file and get namespace object for it. Get the "plugins" list
and extract a python::list. Get the length of that list.
  python::list plugs = ...;
  int len = // get length of plugs

  for(int i = 0; i < len; ++i) {
    try {
      py::object plug = plugs[i];
      Plugin *plugin = py::extract<Plugin *>(plug); // <-------------
      plugins.push_back(plugin);
    } catch(py::error_already_set &e) {
      // handle error
    }
  }
}



Finally, my problem: the extract<Plugin*> call fails:
TypeError: No registered converter was able to extract a C++ pointer to
type Plugin from this Python object of type SomePlugin

So here are my questions:
1) Is this expected behaviour? Does my wrapping not register the
appropriate base pointer extractors?
2) How can I manually register appropriate converters?
3) If 2 is not possible, how can I work around the problem?


Thanks in advance
Sebastian Redl



From dave at boost-consulting.com  Mon Jun 12 15:56:08 2006
From: dave at boost-consulting.com (David Abrahams)
Date: Mon, 12 Jun 2006 09:56:08 -0400
Subject: [C++-sig] [Boost.Python] Extracting indirect base.
References: <44896928.20103@getdesigned.at>
Message-ID: <u1wtuzddz.fsf@boost-consulting.com>

Sebastian Redl <sebastian.redl at getdesigned.at> writes:

>   class_<InputPluginWrap, bases<PluginWrap>,
                                  ^^^^^^
should be just Plugin, I think.

> boost::noncopyable>("InputPlugin")
>     .def("loadData", ...)
>     ;

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com



From dave at boost-consulting.com  Mon Jun 12 16:08:30 2006
From: dave at boost-consulting.com (David Abrahams)
Date: Mon, 12 Jun 2006 10:08:30 -0400
Subject: [C++-sig] Exception Translation
References: <f24bfcdb0606070925i6f9362c1xffc28b5989849606@mail.gmail.com>
Message-ID: <upshexy8x.fsf@boost-consulting.com>

"jesus garcia" <jesus24b at gmail.com> writes:

> I like export the follow simple program
>
> #define REPORT_ERROR(nerr,ErrorMsg) throw error(nerr,ErrorMsg)
>
> where error is a class
>
> Class FileName : public string {
> ...
> Void compose (const string &str, int no, const string &ext)
>    *this=(FileName) str;
>    if (no!=-1) {
>       char aux_str[5];
>       sprintf(aux_str,"%05d",no);
>       *this += aux_str;
>    }
>
>    if (ext!="") *this += "."+ext;
>    else REPORT_ERROR(1005,"compose: need a extension");
> }
> ...

Driving from string is a bad idea for many reasons, but anyway.

> I no have any idea for export de class string in C++ to the class
> string in Python 

You can't; it's already done automatically, but you can't derive from
it, because std::string is mapped into python's "str" type, which
isn't a Boost.Python wrapper class.

If you derive your C++ exception types from std::exception you'll get
automatic translation of the what() string into Python.

> and the error I see the follow page
> http://www.boost.org/libs/python/doc/tutorial/doc/html/python/exception.html
> that shows how catch the error but I do it and the problem persist,

You haven't told us what the problem is.

> someone can help,
>
> I put the solution in a file who contains the BOOST_PYTHON_MODULE
> macro like this follow lines"
>
> #include <boost/python.hpp>
> using namespace boost::python;
>
> struct PodBayDoorException;
>
> void brack(PodBayDoorException const& x)
> {PyErr_SetString(PyExc_UserWarning, x);}
>
> extern void port2python_func();
>
> BOOST_PYTHON_MODULE(_nulo) {
> register_exception_translator<PodBayDoorException>(brack);
> port2python_func();
> ...
> }

I'm not sure what you're trying to do here.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com



From dave at boost-consulting.com  Mon Jun 12 16:12:57 2006
From: dave at boost-consulting.com (David Abrahams)
Date: Mon, 12 Jun 2006 10:12:57 -0400
Subject: [C++-sig] Overloading & keyword args bug?
References: <44873568.10004@pixar.com>
Message-ID: <uk67mxy1i.fsf@boost-consulting.com>

Alex Mohr <amohr at pixar.com> writes:

> Hi folks,
>
> Can someone tell me what's wrong with this?  As is, this code crashes in 
> the example below (anytime I try to pass keyword arguments to the first 
> overload).  If I reverse the order of the lines marked with ########, 
> everything seems to work fine.
>
> Thanks!
>
> Alex
>
>
> #include <boost/python.hpp>
> #include <string>
> #include <cstdio>
>
> using namespace boost::python;
> using std::string;
>
> static void f1(string const &a1, int x, int y) {
>      printf("f1 with %s, %d, %d\n", a1.c_str(), x, y);
> }
>
> static void f2(string const &a1, string const &a2, int x, int y) {
>      printf("f2 with %s, %s, %d, %d\n", a1.c_str(), a2.c_str(), x, y);
> }
>
> BOOST_PYTHON_MODULE(overload_kw) {
>      def("f", f1, (arg("x")=3, arg("y")=4)); // ########
>      def("f", f2, (arg("x")=5, arg("y")=6)); // ########
> }
>
> ---------------------------------------
>
>  >>> from overload_kw import *
>  >>> f('hello')
> f1 with hello, 3, 4
>  >>> f('hello', 'goodbye')
> f2 with hello, goodbye, 5, 6
>  >>> f('hello', 1, 2)
> f1 with hello, 1, 2
>  >>> f('hello', x=1)
> Segmentation fault

Broken compiler code generation?  The enclosed works for me without
errors.  Got a stack backtrace?

-------------- next part --------------
A non-text attachment was scrubbed...
Name: alex_mohr.zip
Type: application/zip
Size: 1875 bytes
Desc: not available
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20060612/fccb8840/attachment.zip>
-------------- next part --------------

myjam -sTOOLS=vc-8_0 --verbose-test test
...found 1671 targets...
...updating 6 targets...
vc-C++ c:\tmp\build\bin\boost\libs\python\user\quack_ext.pyd\vc-8_0\debug\threading-multi\quack.obj
quack.cpp
vc-Link c:\tmp\build\bin\boost\libs\python\user\quack_ext.pyd\vc-8_0\debug\threading-multi\quack_ext.pyd c:\tmp\build\bin\boost\libs\python\user\quack_ext.pyd\vc-8_0\debug\threading-multi\quack_ext.lib
   Creating library c:\tmp\build\bin\boost\libs\python\user\quack_ext.pyd\vc-8_0\debug\threading-multi\quack_ext.lib and object c:\tmp\build\bin\boost\libs\python\user\quack_ext.pyd\vc-8_0\debug\threading-multi\quack_ext.exp
Microsoft (R) Manifest Tool version 5.2.3790.2014.
Copyright (c) Microsoft Corporation 2005. .
All rights reserved..
execute-test c:\tmp\build\bin\boost\libs\python\user\quack.test\vc-8_0\debug\threading-multi\quack.run
        1 file(s) copied.
====== BEGIN OUTPUT ======
# installing zipimport hook
import zipimport # builtin
# installed zipimport hook
# c:\Python24\lib\site.pyc matches c:\Python24\lib\site.py
import site # precompiled from c:\Python24\lib\site.pyc
# c:\Python24\lib\os.pyc matches c:\Python24\lib\os.py
import os # precompiled from c:\Python24\lib\os.pyc
import nt # builtin
# c:\Python24\lib\ntpath.pyc matches c:\Python24\lib\ntpath.py
import ntpath # precompiled from c:\Python24\lib\ntpath.pyc
# c:\Python24\lib\stat.pyc matches c:\Python24\lib\stat.py
import stat # precompiled from c:\Python24\lib\stat.pyc
# c:\Python24\lib\UserDict.pyc matches c:\Python24\lib\UserDict.py
import UserDict # precompiled from c:\Python24\lib\UserDict.pyc
# c:\Python24\lib\copy_reg.pyc matches c:\Python24\lib\copy_reg.py
import copy_reg # precompiled from c:\Python24\lib\copy_reg.pyc
# c:\Python24\lib\types.pyc matches c:\Python24\lib\types.py
import types # precompiled from c:\Python24\lib\types.pyc
# c:\Python24\lib\locale.pyc matches c:\Python24\lib\locale.py
import locale # precompiled from c:\Python24\lib\locale.pyc
import _locale # builtin
# c:\Python24\lib\codecs.pyc matches c:\Python24\lib\codecs.py
import codecs # precompiled from c:\Python24\lib\codecs.pyc
import _codecs # builtin
import encodings # directory c:\Python24\lib\encodings
# c:\Python24\lib\encodings\__init__.pyc matches c:\Python24\lib\encodings\__init__.py
import encodings # precompiled from c:\Python24\lib\encodings\__init__.pyc
# c:\Python24\lib\encodings\aliases.pyc matches c:\Python24\lib\encodings\aliases.py
import encodings.aliases # precompiled from c:\Python24\lib\encodings\aliases.pyc
# c:\Python24\lib\encodings\cp1252.pyc matches c:\Python24\lib\encodings\cp1252.py
import encodings.cp1252 # precompiled from c:\Python24\lib\encodings\cp1252.pyc
# c:\Python24\lib\warnings.pyc matches c:\Python24\lib\warnings.py
import warnings # precompiled from c:\Python24\lib\warnings.pyc
# c:\Python24\lib\linecache.pyc matches c:\Python24\lib\linecache.py
import linecache # precompiled from c:\Python24\lib\linecache.pyc
Python 2.4.2 (#67, Sep 28 2005, 12:41:11) [MSC v.1310 32 bit (Intel)] on win32
Type "help", "copyright", "credits" or "license" for more information.
import quack_ext # dynamically loaded from c:\tmp\build\bin\boost\libs\python\user\quack_ext.pyd\vc-8_0\debug\threading-multi\quack_ext.pyd
# clear __builtin__._
# clear sys.path
# clear sys.argv
# clear sys.ps1
# clear sys.ps2
# clear sys.exitfunc
# clear sys.exc_type
# clear sys.exc_value
# clear sys.exc_traceback
# clear sys.last_type
# clear sys.last_value
# clear sys.last_traceback
# clear sys.path_hooks
# clear sys.path_importer_cache
# clear sys.meta_path
# restore sys.stdin
# restore sys.stdout
# restore sys.stderr
# cleanup __main__
# cleanup[1] locale
# cleanup[1] site
# cleanup[1] encodings
# cleanup[1] quack_ext
# cleanup[1] nt
# cleanup[1] zipimport
# cleanup[1] warnings
# cleanup[1] _codecs
# cleanup[1] encodings.cp1252
# cleanup[1] codecs
# cleanup[1] types
# cleanup[1] _locale
# cleanup[1] signal
# cleanup[1] linecache
# cleanup[1] encodings.aliases
# cleanup[1] exceptions
# cleanup[2] copy_reg
# cleanup[2] os.path
# cleanup[2] ntpath
# cleanup[2] UserDict
# cleanup[2] stat
# cleanup[2] os
# cleanup sys
# cleanup __builtin__
# cleanup ints: 7 unfreed ints in 1 out of 5 blocks
# cleanup floats
f1 with hello, 3, 4
f2 with hello, goodbye, 5, 6
f1 with hello, 1, 2
f1 with hello, 1, 4
 
EXIT STATUS: 0 
====== END OUTPUT ======
**passed** c:\tmp\build\bin\boost\libs\python\user\quack.test\vc-8_0\debug\threading-multi\quack.test
...updated 6 targets...



-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com

From amohr at pixar.com  Mon Jun 12 19:41:21 2006
From: amohr at pixar.com (Alex Mohr)
Date: Mon, 12 Jun 2006 10:41:21 -0700
Subject: [C++-sig] Overloading & keyword args bug?
In-Reply-To: <uk67mxy1i.fsf@boost-consulting.com>
References: <44873568.10004@pixar.com> <uk67mxy1i.fsf@boost-consulting.com>
Message-ID: <448DA741.6010705@pixar.com>

> Broken compiler code generation?  The enclosed works for me without
> errors.  Got a stack backtrace?

Interesting.  Thanks for looking at it, David.

The stack from gdb is bogus (2 empty frames).  This is with gcc 4.0.1 on 
linux.   I'll try with 4.0.1 on OSX and see if it's any different. 
We're supposed to upgrade to 4.0.3 this week, I think.  I'll try that 
when I can.

Anyone else out there with who's willing to try this?  I'm particularly 
interested to confirm/deny with another gcc 4.0.x user.

Alex


From dave at boost-consulting.com  Mon Jun 12 20:48:45 2006
From: dave at boost-consulting.com (David Abrahams)
Date: Mon, 12 Jun 2006 14:48:45 -0400
Subject: [C++-sig] Overloading & keyword args bug?
References: <44873568.10004@pixar.com> <uk67mxy1i.fsf@boost-consulting.com>
	<448DA741.6010705@pixar.com>
Message-ID: <u64j6w6pe.fsf@boost-consulting.com>

Alex Mohr <amohr at pixar.com> writes:

>> Broken compiler code generation?  The enclosed works for me without
>> errors.  Got a stack backtrace?
>
> Interesting.  Thanks for looking at it, David.
>
> The stack from gdb is bogus (2 empty frames).  This is with gcc 4.0.1 on 
> linux.   I'll try with 4.0.1 on OSX and see if it's any different. 
> We're supposed to upgrade to 4.0.3 this week, I think.  I'll try that 
> when I can.
>
> Anyone else out there with who's willing to try this?  I'm particularly 
> interested to confirm/deny with another gcc 4.0.x user.

Works for me with g++-4.0.2/4.0.0 on Cygwin also (with the boost CVS HEAD).

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com



From s_sourceforge at nedprod.com  Tue Jun 13 00:25:54 2006
From: s_sourceforge at nedprod.com (Niall Douglas)
Date: Mon, 12 Jun 2006 23:25:54 +0100
Subject: [C++-sig] Code generation bug in pyplusplus
Message-ID: <448DF802.5375.2134119@s_sourceforge.nedprod.com>

Back onto getting the python bindings for TnFOX working again. I'm 
using the latest from SVN.

Pyplusplus generates bad code here:

    FXPrimaryButton_wrapper(::FX::FXComposite * parent, 
::FX::FXString const & text, ::FX::FXIcon * icon=0, ::FX::FXObject * 
tgt=0, ::FX::FXSelector sel=0, ::FX::FXuint 
opts=::FX::FXuint(::FX::PBUTTON_NORMAL), ::FX::FXint x=0, ::FX::FXint 
y=0, ::FX::FXint w=0, ::FX::FXint h=0, ::FX::FXint 
pl=(FX::FXWindow::defaultPadding()() * 4), ::FX::FXint 
pr=(FX::FXWindow::defaultPadding()() * 4), ::FX::FXint 
pt=FX::FXWindow::defaultPadding( ), ::FX::FXint 
pb=FX::FXWindow::defaultPadding( ), ::FX::FXint sr=0 )
    : FX::FXPrimaryButton( parent, text, icon, tgt, sel, opts, x, y, 
w, h, pl, pr, pt, pb, sr )
      , bp::wrapper< FX::FXPrimaryButton >()
    {}

It should be:

    FXPrimaryButton_wrapper(::FX::FXComposite * parent, 
::FX::FXString const & text, ::FX::FXIcon * icon=0, ::FX::FXObject * 
tgt=0, ::FX::FXSelector sel=0, ::FX::FXuint 
opts=::FX::FXuint(::FX::PBUTTON_NORMAL), ::FX::FXint x=0, ::FX::FXint 
y=0, ::FX::FXint w=0, ::FX::FXint h=0, ::FX::FXint 
pl=(FX::FXWindow::defaultPadding() * 4), ::FX::FXint 
pr=(FX::FXWindow::defaultPadding() * 4), ::FX::FXint 
pt=FX::FXWindow::defaultPadding( ), ::FX::FXint 
pb=FX::FXWindow::defaultPadding( ), ::FX::FXint sr=0 )
    : FX::FXPrimaryButton( parent, text, icon, tgt, sel, opts, x, y, 
w, h, pl, pr, pt, pb, sr )
      , bp::wrapper< FX::FXPrimaryButton >()
    {}

Note the difference? FXWindow::defaultPadding() has an extra () after 
it. It seems pyplusplus only makes this mistake when it's an 
expression eg; when multiplied by 4 as above.

Cheers,
Niall




From s_sourceforge at nedprod.com  Tue Jun 13 03:53:00 2006
From: s_sourceforge at nedprod.com (Niall Douglas)
Date: Tue, 13 Jun 2006 02:53:00 +0100
Subject: [C++-sig] Inserting custom code into pyplusplus output
Message-ID: <448E288C.7303.2D0DF55@s_sourceforge.nedprod.com>

How do I insert custom code into the BOOST_PYTHON_MODULE() output 
from pyplusplus?


I want:

extern void InitialiseTnFOXPython();

BOOST_PYTHON_MODULE(TnFOX){
    InitialiseTnFOXPython();

    register_enumerations();

    register_FXVec3f_class();

...

I can't see any way from the sources on how to do this :(



Next bit of custom code:

In FXApp::init(), it takes mandatory parameters of void init( int & 
argc, char * * argv, ::FX::FXbool connect ). Obviously, BPL doesn't 
know to match a list of strings to a char **argv, so under pyste I 
inserted the following custom code:

DEFINE_MAKECARRAYITER(FXApp, const FX::FXchar *, getArgv, (), 
c.getArgc())
static void FXApp_init(FXApp &app, int argc, list argv, unsigned char 
connect=TRUE)
{
	int n, size=PyList_Size(argv.ptr());
	static QMemArray<const char *> array;
	array.resize(size+1);
	for(n=0; n<size; n++)
	{
		array[n]=PyString_AsString(PyList_GetItem(argv.ptr(), n));
	}
	array[n]=0;
	app.init(argc, (char **)(array.data()), connect);
}
static void FXApp_init2(FXApp &app, int argc, list argv)
{
	FXApp_init(app, argc, argv);
}

... and then I replaced:

-        .def("init", &FX::FXApp::init, 
&TnFOX::FX_FXApp_Wrapper::default_init_3)
-        .def("init", &TnFOX::FX_FXApp_Wrapper::default_init_2)
+        .def("init", &FXApp_init)
+        .def("init", &FXApp_init2)

-        .def("getArgv", &FX::FXApp::getArgv, 
return_internal_reference< 1 >())
+        .def("getArgv", &FXApp_getArgv)

So can pyplusplus be extended to automatically generate this kind of 
accessor code, if not how can I insert this custom code via 
pyplusplus?

(We are very close to a working TnFOX python bindings. Yay!)

Cheers,
Niall





From roman.yakovenko at gmail.com  Tue Jun 13 10:40:19 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Tue, 13 Jun 2006 11:40:19 +0300
Subject: [C++-sig] Code generation bug in pyplusplus
In-Reply-To: <448DF802.5375.2134119@s_sourceforge.nedprod.com>
References: <448DF802.5375.2134119@s_sourceforge.nedprod.com>
Message-ID: <7465b6170606130140g4e447d22g4c617705780f3e43@mail.gmail.com>

On 6/13/06, Niall Douglas <s_sourceforge at nedprod.com> wrote:
> Back onto getting the python bindings for TnFOX working again.

Good.

> I'm using the latest from SVN.

It has same functionality and bugs ( :-) ) as 0.8.0 version

> Note the difference? FXWindow::defaultPadding() has an extra () after
> it. It seems pyplusplus only makes this mistake when it's an
> expression eg; when multiplied by 4 as above.

I agree that this is a bug, but this is GCC-XML bug. If you will take
a look on generated xml file you will find that all "defaultPadding"
have extra (). pygccxml
aware of such bug ( and some other ) and fixes them. In this case it can do
nothing. Fortunately you have the way to fix the bug:

mb = module_builder_t( ... )
constr = mb.constructor( 'FXPrimaryButton'
                                         , arg_types=[None]*number of
arguments )

constr.arguments[x].default_value = new default value

If the query, I wrote, does not work for you, you will have to refine it.
You can read more about "query interface" here:

http://language-binding.net/pygccxml/query_interface.html

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From roman.yakovenko at gmail.com  Tue Jun 13 10:59:10 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Tue, 13 Jun 2006 11:59:10 +0300
Subject: [C++-sig] Inserting custom code into pyplusplus output
In-Reply-To: <448E288C.7303.2D0DF55@s_sourceforge.nedprod.com>
References: <448E288C.7303.2D0DF55@s_sourceforge.nedprod.com>
Message-ID: <7465b6170606130159n55bba9faw9377ebe74d159c0d@mail.gmail.com>

On 6/13/06, Niall Douglas <s_sourceforge at nedprod.com> wrote:
> How do I insert custom code into the BOOST_PYTHON_MODULE() output
> from pyplusplus?
>
>
> I want:
>
> extern void InitialiseTnFOXPython();
>
> BOOST_PYTHON_MODULE(TnFOX){
>     InitialiseTnFOXPython();

from pyplusplus import code_creators

mb = module_builder_t( ... )
mb.build_code_creator( ... )

code = your code
mb.code_creator.body.adopt_creator( code_creators.custom_text_t( code ), 0 )

> I can't see any way from the sources on how to do this :(

I will try to document code creators before next release

>
> Next bit of custom code:
>
> In FXApp::init(), it takes mandatory parameters of void init( int &
> argc, char * * argv, ::FX::FXbool connect ). Obviously, BPL doesn't
> know to match a list of strings to a char **argv, so under pyste I
> inserted the following custom code:
>
> DEFINE_MAKECARRAYITER(FXApp, const FX::FXchar *, getArgv, (),
> c.getArgc())
> static void FXApp_init(FXApp &app, int argc, list argv, unsigned char
> connect=TRUE)
> {
>         int n, size=PyList_Size(argv.ptr());
>         static QMemArray<const char *> array;
>         array.resize(size+1);
>         for(n=0; n<size; n++)
>         {
>                 array[n]=PyString_AsString(PyList_GetItem(argv.ptr(), n));
>         }
>         array[n]=0;
>         app.init(argc, (char **)(array.data()), connect);
> }
> static void FXApp_init2(FXApp &app, int argc, list argv)
> {
>         FXApp_init(app, argc, argv);
> }
>
> ... and then I replaced:
>
> -        .def("init", &FX::FXApp::init,
> &TnFOX::FX_FXApp_Wrapper::default_init_3)
> -        .def("init", &TnFOX::FX_FXApp_Wrapper::default_init_2)
> +        .def("init", &FXApp_init)
> +        .def("init", &FXApp_init2)
>
> -        .def("getArgv", &FX::FXApp::getArgv,
> return_internal_reference< 1 >())
> +        .def("getArgv", &FXApp_getArgv)
>
> So can pyplusplus be extended to automatically generate this kind of
> accessor code

We are working on this:
http://language-binding.net/pyplusplus/peps/call_wrapper_policies.html

> if not how can I insert this custom code via pyplusplus?

1. You should exclude FXApp::init functions from being exposed, right?

mb = module_builder_t( ... _
FXApp = mb.class_( 'FXApp' )
FXApp.member_functions( 'init' ).exclude()
init = """def("init", &FX::FXApp::init,
&TnFOX::FX_FXApp_Wrapper::default_init_3)"""
FXApp.add_code( init )
...

> We are very close to a working TnFOX python bindings. Yay!

It is really good news :-). I am glad you find the time to work on this project.
I have few ideas about new GUI to pyplusplus. May be I will implement it using
Python bindings to TnFOX, we'll see.

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From mat at matbray.com  Tue Jun 13 14:23:37 2006
From: mat at matbray.com (Matthew Bray)
Date: Tue, 13 Jun 2006 14:23:37 +0200
Subject: [C++-sig] missing procedure entry point in boost_python.dll
Message-ID: <e6mae9$a7j$1@sea.gmane.org>

I just downloaded the 1.33.1 distribution of the boost libraries for 
Windows and quickly set up the Boost.Python Hello example in VC++ 6.5 by 
following the documentation on the boost site. Everything appears to 
compile well enough in both debug and release versions however when 
attempting to use the library I get this error:

The procedure entry point 
??1py_function_impl_base at objects@python at boost@@UAE at XZ could not be 
located in the dynamic link library boost_python.dll

The same error occurs when attempting to use python 2.2.3 and 2.4.

The name isn't entirely mangled so perhaps it can be of some help - has 
anyway has anyone come across this issue before, is it easily solvable? 
I didnt see this error when using a much older boost 1.29

Regards, Mat.



From s_sourceforge at nedprod.com  Tue Jun 13 16:15:51 2006
From: s_sourceforge at nedprod.com (Niall Douglas)
Date: Tue, 13 Jun 2006 15:15:51 +0100
Subject: [C++-sig] Inserting custom code into pyplusplus output
In-Reply-To: <7465b6170606130159n55bba9faw9377ebe74d159c0d@mail.gmail.com>
References: <448E288C.7303.2D0DF55@s_sourceforge.nedprod.com>
Message-ID: <448ED6A7.2219.145840@s_sourceforge.nedprod.com>

On 13 Jun 2006 at 11:59, Roman Yakovenko wrote:

> > So can pyplusplus be extended to automatically generate this kind of
> > accessor code
> 
> We are working on this:
> http://language-binding.net/pyplusplus/peps/call_wrapper_policies.html

You may find the code defining the macro DEFINE_MAKECARRAYITER() very 
useful in TnFOX. It allows access to dynamic arrays which can be 
modified and extended by a python list interface.

Also, you can't get pyplusplus to generate thread safe code alone. 
You need alterations to BPL because the iterator interface needs GIL 
unlocking, and nowhere does BPL pass through pyplusplus code to allow 
this.

The patches required to BPL are moderate however and would cost 
little in runtime overhead. You basically need two installable 
function pointers to do the GIL locking and unlocking.

Cheers,
Niall





From s_sourceforge at nedprod.com  Tue Jun 13 16:15:51 2006
From: s_sourceforge at nedprod.com (Niall Douglas)
Date: Tue, 13 Jun 2006 15:15:51 +0100
Subject: [C++-sig] Inserting custom code into pyplusplus output
In-Reply-To: <7465b6170606130159n55bba9faw9377ebe74d159c0d@mail.gmail.com>
References: <448E288C.7303.2D0DF55@s_sourceforge.nedprod.com>
Message-ID: <448ED6A7.3101.1457E2@s_sourceforge.nedprod.com>

On 13 Jun 2006 at 11:59, Roman Yakovenko wrote:

> > if not how can I insert this custom code via pyplusplus?
> 
> 1. You should exclude FXApp::init functions from being exposed, right?
> 
> mb = module_builder_t( ... _
> FXApp = mb.class_( 'FXApp' )
> FXApp.member_functions( 'init' ).exclude()
> init = """def("init", &FX::FXApp::init,
> &TnFOX::FX_FXApp_Wrapper::default_init_3)"""
> FXApp.add_code( init )
> ...

Unfortunately, FXApp::init is a required function for all GUI 
programs. It also mandatorily takes argc and argv.

> > We are very close to a working TnFOX python bindings. Yay!
> 
> It is really good news :-). I am glad you find the time to work on this project.
> I have few ideas about new GUI to pyplusplus. May be I will implement it using
> Python bindings to TnFOX, we'll see.

If pyplusplus doesn't yet support these things, I can also patch them 
in and supply a diff file to apply on the resulting bindings in the 
short term.

It would be really great if we had a real world user of the TnFOX 
python bindings.

Cheers,
Niall





From roman.yakovenko at gmail.com  Tue Jun 13 20:13:20 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Tue, 13 Jun 2006 21:13:20 +0300
Subject: [C++-sig] Inserting custom code into pyplusplus output
In-Reply-To: <448ED6A7.3101.1457E2@s_sourceforge.nedprod.com>
References: <448E288C.7303.2D0DF55@s_sourceforge.nedprod.com>
	<7465b6170606130159n55bba9faw9377ebe74d159c0d@mail.gmail.com>
	<448ED6A7.3101.1457E2@s_sourceforge.nedprod.com>
Message-ID: <7465b6170606131113x39df721ena98821790150ce9f@mail.gmail.com>

On 6/13/06, Niall Douglas <s_sourceforge at nedprod.com> wrote:
> On 13 Jun 2006 at 11:59, Roman Yakovenko wrote:
>
> > > if not how can I insert this custom code via pyplusplus?
> >
> > 1. You should exclude FXApp::init functions from being exposed, right?
> >
> > mb = module_builder_t( ... _
> > FXApp = mb.class_( 'FXApp' )
> > FXApp.member_functions( 'init' ).exclude()
> > init = """def("init", &FX::FXApp::init,
> > &TnFOX::FX_FXApp_Wrapper::default_init_3)"""
> > FXApp.add_code( init )
> > ...
>
> Unfortunately, FXApp::init is a required function for all GUI
> programs. It also mandatorily takes argc and argv.

I think I was not clear. I showed you the way to replace code that would
be generated by pyplusplus for FXApp::init member function, with your own code.

> > > We are very close to a working TnFOX python bindings. Yay!
> >
> > It is really good news :-). I am glad you find the time to work on this project.
> > I have few ideas about new GUI to pyplusplus. May be I will implement it using
> > Python bindings to TnFOX, we'll see.
>
> If pyplusplus doesn't yet support these things, I can also patch them
> in and supply a diff file to apply on the resulting bindings in the
> short term.

It is not as simple as I'd like it to be. I still try to find elegant
way to implement
this.

> It would be really great if we had a real world user of the TnFOX
> python bindings.

:-) I understand you.

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From niki at vintech.bg  Tue Jun 13 20:08:10 2006
From: niki at vintech.bg (Niki)
Date: Tue, 13 Jun 2006 21:08:10 +0300
Subject: [C++-sig] Inserting custom code into pyplusplus output
In-Reply-To: <448ED6A7.3101.1457E2@s_sourceforge.nedprod.com>
References: <448E288C.7303.2D0DF55@s_sourceforge.nedprod.com>
	<448ED6A7.3101.1457E2@s_sourceforge.nedprod.com>
Message-ID: <448EFF0A.8010501@vintech.bg>

Niall Douglas wrote:
> It would be really great if we had a real world user of the TnFOX 
> python bindings.
>
>   
When will they be ready?

Niki Spahiev



From roman.yakovenko at gmail.com  Tue Jun 13 20:19:25 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Tue, 13 Jun 2006 21:19:25 +0300
Subject: [C++-sig] Inserting custom code into pyplusplus output
In-Reply-To: <448ED6A7.2219.145840@s_sourceforge.nedprod.com>
References: <448E288C.7303.2D0DF55@s_sourceforge.nedprod.com>
	<7465b6170606130159n55bba9faw9377ebe74d159c0d@mail.gmail.com>
	<448ED6A7.2219.145840@s_sourceforge.nedprod.com>
Message-ID: <7465b6170606131119o338fe43bl51ae366074b860af@mail.gmail.com>

On 6/13/06, Niall Douglas <s_sourceforge at nedprod.com> wrote:
> On 13 Jun 2006 at 11:59, Roman Yakovenko wrote:
>
> > > So can pyplusplus be extended to automatically generate this kind of
> > > accessor code
> >
> > We are working on this:
> > http://language-binding.net/pyplusplus/peps/call_wrapper_policies.html
>
> You may find the code defining the macro DEFINE_MAKECARRAYITER() very
> useful in TnFOX. It allows access to dynamic arrays which can be
> modified and extended by a python list interface.

Thank you, I will take a look on the code and will find out how to integrate it
with pyplusplus.

> Also, you can't get pyplusplus to generate thread safe code alone.
> You need alterations to BPL because the iterator interface needs GIL
> unlocking, and nowhere does BPL pass through pyplusplus code to allow
> this.
>
> The patches required to BPL are moderate however and would cost
> little in runtime overhead. You basically need two installable
> function pointers to do the GIL locking and unlocking.

I understand this. I just thought that sometimes it will be enough to make some
function thread safe.


-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From dave at boost-consulting.com  Wed Jun 14 01:11:50 2006
From: dave at boost-consulting.com (David Abrahams)
Date: Tue, 13 Jun 2006 19:11:50 -0400
Subject: [C++-sig] missing procedure entry point in boost_python.dll
References: <e6mae9$a7j$1@sea.gmane.org>
Message-ID: <u3be8lkg9.fsf@boost-consulting.com>

Matthew Bray <mat at matbray.com> writes:

> I just downloaded the 1.33.1 distribution of the boost libraries for 
> Windows and quickly set up the Boost.Python Hello example in VC++ 6.5 by 
> following the documentation on the boost site. Everything appears to 
> compile well enough in both debug and release versions however when 
> attempting to use the library I get this error:
>
> The procedure entry point 
> ??1py_function_impl_base at objects@python at boost@@UAE at XZ could not be 
> located in the dynamic link library boost_python.dll
>
> The same error occurs when attempting to use python 2.2.3 and 2.4.
>
> The name isn't entirely mangled so perhaps it can be of some help - has 
> anyway has anyone come across this issue before, is it easily solvable? 
> I didnt see this error when using a much older boost 1.29

Did you use Boost.Build to compile your test?  Boost.Python passed all
its tests with vc6.5 for the 1.33.1 release, and the tests are all run
through Boost.Build.  Probably you have the wrong combination of
compiler/linker switches.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com



From s_sourceforge at nedprod.com  Wed Jun 14 23:24:48 2006
From: s_sourceforge at nedprod.com (Niall Douglas)
Date: Wed, 14 Jun 2006 22:24:48 +0100
Subject: [C++-sig] Inserting custom code into pyplusplus output
In-Reply-To: <7465b6170606131113x39df721ena98821790150ce9f@mail.gmail.com>
References: <448ED6A7.3101.1457E2@s_sourceforge.nedprod.com>
Message-ID: <44908CB0.961.968DBF@s_sourceforge.nedprod.com>

On 13 Jun 2006 at 21:13, Roman Yakovenko wrote:

> > > mb = module_builder_t( ... _
> > > FXApp = mb.class_( 'FXApp' )
> > > FXApp.member_functions( 'init' ).exclude()
> > > init = """def("init", &FX::FXApp::init,
> > > &TnFOX::FX_FXApp_Wrapper::default_init_3)"""
> > > FXApp.add_code( init )
> > > ...
> >
> > Unfortunately, FXApp::init is a required function for all GUI
> > programs. It also mandatorily takes argc and argv.
> 
> I think I was not clear. I showed you the way to replace code that would
> be generated by pyplusplus for FXApp::init member function, with your own code.

Ah right. I did get that bit, but I didn't see how to custom insert 
some text before the RegisterFXApp() function. For example, if using 
the above trying:

FXApp.parent.adopt_creator(code_creators.custom_text_t(<FXApp::init 
replacement definition>))

... it turns out FXApp.parent is a namespace, and it wouldn't let me 
adopt custom code into it.

Once again, consulting the source left me pretty clueless :(

Cheers,
Niall





From s_sourceforge at nedprod.com  Wed Jun 14 23:24:48 2006
From: s_sourceforge at nedprod.com (Niall Douglas)
Date: Wed, 14 Jun 2006 22:24:48 +0100
Subject: [C++-sig] Inserting custom code into pyplusplus output
In-Reply-To: <448EFF0A.8010501@vintech.bg>
References: <448ED6A7.3101.1457E2@s_sourceforge.nedprod.com>
Message-ID: <44908CB0.30130.968D13@s_sourceforge.nedprod.com>

On 13 Jun 2006 at 21:08, Niki wrote:

> Niall Douglas wrote:
> > It would be really great if we had a real world user of the TnFOX 
> > python bindings.
> >
> >   
> When will they be ready?

It'll be soon. But I'll have to say "when it's done".

I'm currently seeing a strange segfault inside python. Going to have 
to compile my own with debug symbols to find out what's going on.

Cheers,
Niall





From roman.yakovenko at gmail.com  Thu Jun 15 07:55:13 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Thu, 15 Jun 2006 08:55:13 +0300
Subject: [C++-sig] Inserting custom code into pyplusplus output
In-Reply-To: <44908CB0.961.968DBF@s_sourceforge.nedprod.com>
References: <448ED6A7.3101.1457E2@s_sourceforge.nedprod.com>
	<7465b6170606131113x39df721ena98821790150ce9f@mail.gmail.com>
	<44908CB0.961.968DBF@s_sourceforge.nedprod.com>
Message-ID: <7465b6170606142255h7e5ccabfha8a64ba9e8ee2120@mail.gmail.com>

On 6/15/06, Niall Douglas <s_sourceforge at nedprod.com> wrote:
> On 13 Jun 2006 at 21:13, Roman Yakovenko wrote:
>
> > > > mb = module_builder_t( ... _
> > > > FXApp = mb.class_( 'FXApp' )
> > > > FXApp.member_functions( 'init' ).exclude()
> > > > init = """def("init", &FX::FXApp::init,
> > > > &TnFOX::FX_FXApp_Wrapper::default_init_3)"""
> > > > FXApp.add_code( init )
> > > > ...
> > >
> > > Unfortunately, FXApp::init is a required function for all GUI
> > > programs. It also mandatorily takes argc and argv.
> >
> > I think I was not clear. I showed you the way to replace code that would
> > be generated by pyplusplus for FXApp::init member function, with your own code.
>
> Ah right. I did get that bit, but I didn't see how to custom insert
> some text before the RegisterFXApp() function. For example, if using
> the above trying:
>
> FXApp.parent.adopt_creator(code_creators.custom_text_t(<FXApp::init
> replacement definition>))
>
> ... it turns out FXApp.parent is a namespace, and it wouldn't let me
> adopt custom code into it.
>
> Once again, consulting the source left me pretty clueless :(

I did not write FXApp.parent - this will not work, you already know this.

If you want to insert a code in some specific place, then :

module.body.adopt_creator( code_creators.custom_text_t( ... ), position )
( By default the code creator will be appended to the end of the list )

module is top level, root, main code creator. It derives from compound_t code
creator. So it keeps a list of other code creators, including module_body_t.

If you take a look in debugger on module_t code creator ( mb.code_creator )
you will see something like this:
<module_t ....>
    <precompiled_header_t ... >
    <include_t ... >
    <class_wrapper_t ...>
    <class_wrapper_t ...>
    ...
    <module_body_t ...>#<= it generates BOOST_PYTHON_MODULE macro
        <enumeration_t ...>
        <enumeration_t ...>
        <class_t ...>
        <class_t ... >
        ...
        <global_variable_t ...>
        <free_function_t ....>

FXApp is not code creator, but declaration. Declarations tree is used
as input to
code creators factory. They are different beast. Am I clear? If you need more
explanation, please ask some specific question, okay?

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From s_sourceforge at nedprod.com  Thu Jun 15 13:40:37 2006
From: s_sourceforge at nedprod.com (Niall Douglas)
Date: Thu, 15 Jun 2006 12:40:37 +0100
Subject: [C++-sig] More pyplusplus custom stuff
Message-ID: <44915545.5613.272220@s_sourceforge.nedprod.com>

The good news is I now have a set of fully working bindings, or at 
least to the stage of where TnFOX was at the v0.85 release. The next 
problem is the orphaned child problem.

TnFOX's GUI objects, much like most C++ GUI toolkits, manage their 
own children. If you delete the parent, it deletes all its children. 
Needless to say, this is a problem for Python.

The solution is to upcall any child deletions so BPL can unhook the 
relevent python object from its C++ sibling. This is done by 
replacing the virtual functions:
   virtual void *getPythonObject() const;
   virtual void decouplePythonObject() const;

Therefore, we need to do the following:

For every class in the FX namespace which has FXObject as a base 
class, insert into EVERY constructor of the wrapper class the 
following code:

FXPython::int_pythonObjectCreated((int_decouplingFunctor=Generic::Bind
ObjN(*this, &"+wrapper_name+"::decouplePythonObject)));

Into private member variables of the wrapper class:

Generic::BoundFunctorV *int_decouplingFunctor;

Into every destructor of the wrapper class:

~"+wrapper_name+"() {
    FXPython::int_pythonObjectDeleted(int_decouplingFunctor);
    using namespace boost::python;
    using namespace std;
    PyObject *py_self=get_owner(this);
    // We must be careful not to reinvoke object destruction!
    py_self->ob_refcnt=1<<16;
    {
        object me(boost::python::handle<>(borrowed(py_self)));
        auto_ptr<"+wrapper_name+"> 
&autoptr=extract<auto_ptr<"+wrapper_name+"> &>(me);
        autoptr.release();
    }
    py_self->ob_refcnt=0
}

Into every body of the wrapper class:

virtual void *getPythonObject() const {
    return (void *) get_owner(this);
}

virtual void decouplePythonObject() const {
    using namespace boost::python;
    using namespace std;
    object me(boost::python::handle<>(borrowed(py_self)));
    auto_ptr<"+wrapper_name+"> 
&autoptr=extract<auto_ptr<"+wrapper_name+"> &>(me);
    autoptr.reset(0);
}


So, how does one achieve this then?

Cheers,
Niall





From dave at boost-consulting.com  Thu Jun 15 15:15:58 2006
From: dave at boost-consulting.com (David Abrahams)
Date: Thu, 15 Jun 2006 09:15:58 -0400
Subject: [C++-sig] More pyplusplus custom stuff
References: <44915545.5613.272220@s_sourceforge.nedprod.com>
Message-ID: <uac8eh84x.fsf@boost-consulting.com>

"Niall Douglas" <s_sourceforge at nedprod.com> writes:

> The solution is to upcall any child deletions so BPL can unhook the 
> relevent python object from its C++ sibling.

You could simplify that system a lot by using shared_ptr to hold the
children in C++, and building a new smart pointer around a core
weak_ptr, that you use as the HeldType in your Python classes.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com



From dave at boost-consulting.com  Thu Jun 15 17:44:20 2006
From: dave at boost-consulting.com (David Abrahams)
Date: Thu, 15 Jun 2006 11:44:20 -0400
Subject: [C++-sig] Boost::Signals
References: <e64ph5$atl$1@sea.gmane.org>
Message-ID: <ulkryfmp7.fsf@boost-consulting.com>

John Reid <johnbaronreid at netscape.net> writes:

> Hi,
>
> I'm trying to call a python function from C++. I pass the function to 
> C++ as a boost::python::object. I can call functions with no arguments, 
> when I try to pass a C++ created object as an argument I get the 
> following error:
>
> Traceback (most recent call last):
>    File "python/test_python.py", line 8, in ?
>      d.connect_new_deal_slot_py_2( f )
> TypeError: No to_python (by-value) converter found for C++ type: class 
> module::classname

Actually that would be namespacename::classname.  It's a C++ class
name, most probably of the return type of the C++ function
implementing your connect_new_deal_slot_py_2 method.  Another
possibility is that it's the name of a C++ class of which you're
passing an instance to f inside that C++ function implementing
connect_new_deal_slot_py_2.  Either way, you haven't wrapped that C++
class.

> Now the boost.python section (http://tinyurl.com/n3cpv) in the python 
> wiki answers the following question:
> "Is it is possible to convert pointers to existing classes to PyObjects* 
> and then be able to pass an existing instance of an object directly to 
> and from python?"
>
> where the last paragraph says:
>
> "If you really want to pass pointers around, it's certainly possible to 
> tell the library to build a Python object around the pointer, but then 
> you need to make sure the lifetime of the C++ object being referenced by 
> the pointer extends past the lifetime of all Python references to the 
> object or your program will crash."

I don't see how that relates to the error message you're seeing above.

> This is what I would like to do

which?  "tell the library to build a Python object around the
pointer?"  http://boost.org/libs/python/doc/v2/ptr.html#examples.

Ensuring the appropriate lifetime relationships is a more complicated
topic than I can explain here, though.          

> but cannot find this elsewhere in the 
> documentation. Can anyone let me know how to do this?
>
> Thanks in advance,
> John.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com



From s_sourceforge at nedprod.com  Thu Jun 15 17:46:58 2006
From: s_sourceforge at nedprod.com (Niall Douglas)
Date: Thu, 15 Jun 2006 16:46:58 +0100
Subject: [C++-sig] Inserting custom code into pyplusplus output
In-Reply-To: <7465b6170606142255h7e5ccabfha8a64ba9e8ee2120@mail.gmail.com>
References: <44908CB0.961.968DBF@s_sourceforge.nedprod.com>
Message-ID: <44918F02.20434.108ABEB@s_sourceforge.nedprod.com>

On 15 Jun 2006 at 8:55, Roman Yakovenko wrote:

> If you want to insert a code in some specific place, then :
> 
> module.body.adopt_creator( code_creators.custom_text_t( ... ), position )
> ( By default the code creator will be appended to the end of the list )

No, this appends into TnFOX.main.cpp. I need it appended into 
FXApp.pypp.cpp. If you examine file_writers/multiple_files.py, you 
can see it writes license, precompiled header, global headers, global 
namespace aliases, wrapper classes and the register function for the 
class in question. Nowhere is it letting you write custom text 
outside the bodies of these items. I suppose you could add a custom 
form of wrapper code creator, but this seems inflexible - you should 
be able to insert text anywhere in any output file.

Cheers,
Niall





From roman.yakovenko at gmail.com  Thu Jun 15 20:08:07 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Thu, 15 Jun 2006 21:08:07 +0300
Subject: [C++-sig] Inserting custom code into pyplusplus output
In-Reply-To: <44918F02.20434.108ABEB@s_sourceforge.nedprod.com>
References: <44908CB0.961.968DBF@s_sourceforge.nedprod.com>
	<7465b6170606142255h7e5ccabfha8a64ba9e8ee2120@mail.gmail.com>
	<44918F02.20434.108ABEB@s_sourceforge.nedprod.com>
Message-ID: <7465b6170606151108w561f5883x33fbebd63add0149@mail.gmail.com>

On 6/15/06, Niall Douglas <s_sourceforge at nedprod.com> wrote:
> On 15 Jun 2006 at 8:55, Roman Yakovenko wrote:
>
> > If you want to insert a code in some specific place, then :
> >
> > module.body.adopt_creator( code_creators.custom_text_t( ... ), position )
> > ( By default the code creator will be appended to the end of the list )
>
> No, this appends into TnFOX.main.cpp. I need it appended into
> FXApp.pypp.cpp. If you examine file_writers/multiple_files.py, you
> can see it writes license, precompiled header, global headers, global
> namespace aliases, wrapper classes and the register function for the
> class in question. Nowhere is it letting you write custom text
> outside the bodies of these items. I suppose you could add a custom
> form of wrapper code creator, but this seems inflexible - you should
> be able to insert text anywhere in any output file.

:-). This is one of my big problems - I don't know what interface to
give to the user.
It is not very difficult to implement what you are asking for.  I am
sure you have an
idea, right ?

As a temporal solution you can create custom file writer. Or may be to write
small script that will add the code to the file.

--
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From roman.yakovenko at gmail.com  Thu Jun 15 20:30:14 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Thu, 15 Jun 2006 21:30:14 +0300
Subject: [C++-sig] More pyplusplus custom stuff
In-Reply-To: <44915545.5613.272220@s_sourceforge.nedprod.com>
References: <44915545.5613.272220@s_sourceforge.nedprod.com>
Message-ID: <7465b6170606151130l23eb0fan54fd7bf3fa5d8ba9@mail.gmail.com>

On 6/15/06, Niall Douglas <s_sourceforge at nedprod.com> wrote:
> The good news is I now have a set of fully working bindings, or at
> least to the stage of where TnFOX was at the v0.85 release.

Congratulation !!!

>The next problem is the orphaned child problem.
>
> TnFOX's GUI objects, much like most C++ GUI toolkits, manage their
> own children. If you delete the parent, it deletes all its children.
> Needless to say, this is a problem for Python.
>
> The solution is to upcall any child deletions so BPL can unhook the
> relevent python object from its C++ sibling. This is done by
> replacing the virtual functions:
>    virtual void *getPythonObject() const;
>    virtual void decouplePythonObject() const;
>
> Therefore, we need to do the following:
>
> For every class in the FX namespace which has FXObject as a base
> class, insert into EVERY constructor of the wrapper class the
> following code:
>
> FXPython::int_pythonObjectCreated((int_decouplingFunctor=Generic::Bind
> ObjN(*this, &"+wrapper_name+"::decouplePythonObject)));
>
> Into private member variables of the wrapper class:
>
> Generic::BoundFunctorV *int_decouplingFunctor;
>
> Into every destructor of the wrapper class:
>
> ~"+wrapper_name+"() {
>     FXPython::int_pythonObjectDeleted(int_decouplingFunctor);
>     using namespace boost::python;
>     using namespace std;
>     PyObject *py_self=get_owner(this);
>     // We must be careful not to reinvoke object destruction!
>     py_self->ob_refcnt=1<<16;
>     {
>         object me(boost::python::handle<>(borrowed(py_self)));
>         auto_ptr<"+wrapper_name+">
> &autoptr=extract<auto_ptr<"+wrapper_name+"> &>(me);
>         autoptr.release();
>     }
>     py_self->ob_refcnt=0
> }
>
> Into every body of the wrapper class:
>
> virtual void *getPythonObject() const {
>     return (void *) get_owner(this);
> }
>
> virtual void decouplePythonObject() const {
>     using namespace boost::python;
>     using namespace std;
>     object me(boost::python::handle<>(borrowed(py_self)));
>     auto_ptr<"+wrapper_name+">
> &autoptr=extract<auto_ptr<"+wrapper_name+"> &>(me);
>     autoptr.reset(0);
> }
>
>
> So, how does one achieve this then?

Niall, there is pretty simple solution, but I am not sure I want it
within pyplusplus.
May be as temporal patch, but I will remove it after we will implement
"call wrapper
policies" ( http://language-binding.net/pyplusplus/peps/call_wrapper_policies.html
)

Description: ( solution for destructor )

cls = mb.class_( 'X' )
cls.add_wrapper_code( destructor declaration and definition )

You can do it with your version of pyplusplus

Description for constructor:

1. add pyplusplus.decl_wrappers.constructor_t new property
    body ( string )

2.  add code that insert this member variable into generated code in
next classes:
     pyplusplus.code_creators.
         constructor_wrapper_t
         special_constructor_wrapper_t
         trivial_constructor_wrapper_t

3. initialize the body property
    > For every class in the FX namespace which has FXObject as a base
    > class, insert into EVERY constructor of the wrapper class the
    > following code:

    mb = module_builder_t( ... )
    FX = mb.namespace( 'FX' )
    FXObject = FX.class_( 'FXObject' )
    classes = FX.classes(
          lambda decl: pygccxml.declarations.is_based_and_derived(
FXObject, decl ) )
    for cls in classes:
        cls.constructors().body = ...
        cls.add_wrapper_code( ... )

Pretty simple and readable, right :-)?

Enjoy

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From s_sourceforge at nedprod.com  Fri Jun 16 00:14:04 2006
From: s_sourceforge at nedprod.com (Niall Douglas)
Date: Thu, 15 Jun 2006 23:14:04 +0100
Subject: [C++-sig] More pyplusplus custom stuff
In-Reply-To: <uac8eh84x.fsf@boost-consulting.com>
Message-ID: <4491E9BC.32342.26B10C6@s_sourceforge.nedprod.com>

On 15 Jun 2006 at 9:15, David Abrahams wrote:

> > The solution is to upcall any child deletions so BPL can unhook the 
> > relevent python object from its C++ sibling.
> 
> You could simplify that system a lot by using shared_ptr to hold the
> children in C++, and building a new smart pointer around a core
> weak_ptr, that you use as the HeldType in your Python classes.

Can you elaborate?

Doesn't shared_ptr basically hold a lock on a weak_ptr which is held 
on a main shared_ptr? I suppose this means I can avoid the refcount 
tricks I was using?

Cheers,
Niall





From j.reid at mail.cryst.bbk.ac.uk  Fri Jun 16 09:55:27 2006
From: j.reid at mail.cryst.bbk.ac.uk (John Reid)
Date: Fri, 16 Jun 2006 08:55:27 +0100
Subject: [C++-sig] Boost::Signals
In-Reply-To: <ulkryfmp7.fsf@boost-consulting.com>
References: <e64ph5$atl$1@sea.gmane.org> <ulkryfmp7.fsf@boost-consulting.com>
Message-ID: <e6to6a$db6$1@sea.gmane.org>



David Abrahams wrote:
> John Reid <johnbaronreid at netscape.net> writes:
> 
> 
>>Hi,
>>
>>I'm trying to call a python function from C++. I pass the function to 
>>C++ as a boost::python::object. I can call functions with no arguments, 
>>when I try to pass a C++ created object as an argument I get the 
>>following error:
>>
>>Traceback (most recent call last):
>>   File "python/test_python.py", line 8, in ?
>>     d.connect_new_deal_slot_py_2( f )
>>TypeError: No to_python (by-value) converter found for C++ type: class 
>>module::classname
> 
> 
> Actually that would be namespacename::classname.  It's a C++ class
> name, most probably of the return type of the C++ function
> implementing your connect_new_deal_slot_py_2 method.  Another
> possibility is that it's the name of a C++ class of which you're
> passing an instance to f inside that C++ function implementing
> connect_new_deal_slot_py_2.  Either way, you haven't wrapped that C++
> class.
> 
> 
I was pretty sure I had wrapped it. Was that not what I was doing with 
the following code?

	class_< test, noncopyable >( "test" )
          .def( "connect_slot", &test::connect_slot_py )
          ;



Anyhow I tracked the problem down a little more and found out it did not 
like the noncopyable base class. Without this everything was OK. From 
the documentation it wasn't clear that this would make it difficult for 
me to pass C++ objects to python. I'm guessing there is a way to do this 
even with the noncopyable base class but I couldn't work it out.

It is clear that I had some problems understanding what was going wrong. 
I'm not really sure if that was a problem with me, the error messages or 
the documentation.

Thanks for your help.

Best,
John.



From lee.will at gmail.com  Fri Jun 16 16:29:58 2006
From: lee.will at gmail.com (Will Lee)
Date: Fri, 16 Jun 2006 09:29:58 -0500
Subject: [C++-sig] Reducing size of binary
Message-ID: <7f03db650606160729u4dc6bc7em4bf795d0b0cb1287@mail.gmail.com>

Hi,

I'm currently using Boost Python in a fairly large project using Linux and
g++.  The combined .so code that I tried to wrap is around 9MB (-O with no
symbol).  I got to a point where the python wrapper binary gets pretty big
(sometimes the wrapper debug .so with symbols can get to 350 MB and the -O
version without symobls would be around 18 MB).  Compile time can get up to
15-20 minutes.  As far as I see, I'm not doing anything too extraordinary.
I am using a lot of add_property and def methods in the wrapper code.

Does anyone have hints on reducing the size of the wrapper binary?

Thanks,

Will
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20060616/3acff3de/attachment.htm>

From lee.will at gmail.com  Fri Jun 16 19:17:26 2006
From: lee.will at gmail.com (Will Lee)
Date: Fri, 16 Jun 2006 12:17:26 -0500
Subject: [C++-sig]  Bug with custom convertor and stl container?
Message-ID: <7f03db650606161017k28035c1dw6ba2a57d2017cf4@mail.gmail.com>

It seems like there's a bug when I use a custom converter (as described in
the FAQ for the custom string) and a wrapped stl container.  If I define a
custom converter and use that type in a std::vector or std::map, I'll get an
"TypeError: No Python class registered for C++" error for that converted
type.

For example, in the following case where I'm converting from type A to a
python integer, std::vector<A> and std::map<int, A> are not working
propertly.  The testMap and testVec unit tests will both fail.  This is
somewhat a showstopper so it would be great if you have any idea on how to
get around this.

Thanks!

ATest.cpp:
#include <boost/python.hpp>
#include <boost/python/suite/indexing/map_indexing_suite.hpp>
#include <boost/python/suite/indexing/vector_indexing_suite.hpp>

#include <vector>
#include <map>

struct A
{
  int value;
  A() : value(0){};
  A(int v) : value(v) {};
};

bool operator==(const A& v1, const A& v2)
{
  return (v1.value == v2.value);
}

struct B
{
  A a;
};

// Converter from A to python int
struct AToPython
{
  static PyObject* convert(const A& s)
  {
    return boost::python::incref(boost::python::object((int)s.value).ptr());
  }
};

// Conversion from python int to A
struct AFromPython
{
  AFromPython()
  {
    boost::python::converter::registry::push_back(
        &convertible,
        &construct,
        boost::python::type_id< A >());
  }

  static void* convertible(PyObject* obj_ptr)
  {
    if (!PyInt_Check(obj_ptr)) return 0;
    return obj_ptr;
  }

  static void construct(
      PyObject* obj_ptr,
      boost::python::converter::rvalue_from_python_stage1_data* data)
  {
    void* storage = (
        (boost::python::converter::rvalue_from_python_storage< A >*)
        data)->storage.bytes;

    new (storage) A((int)PyInt_AsLong(obj_ptr));
    data->convertible = storage;
  }
};

BOOST_PYTHON_MODULE(atest)
{
  using namespace boost::python;
  to_python_converter< A , AToPython >();
  AFromPython();

  class_<std::vector< A > >("AVec")
    .def(vector_indexing_suite<std::vector<A>, false>())
    ;

  class_< std::map<int, A> >("AMap")
    .def(map_indexing_suite<std::map<int, A>, false >())
    ;

  class_< B >("B")
    .add_property("a", make_getter(&B::a,
return_value_policy<return_by_value>()),
        make_setter(&B::a, return_value_policy<return_by_value>()))
    ;
}



ATest.py:
#!/usr/bin/env python

import unittest
import atest

class ATest(unittest.TestCase):
    def testB(self):
        b = atest.B()
        self.assert_(b.a == 0)
        b.a = 35
        self.assert_(b.a == 35)

    def testMap(self):
        amap = atest.AMap()
        amap[3] = 4
        self.assert_(amap[3] == 4)

    def testVec(self):
        amap = atest.AVec()
        amap.append(4)
        self.assert_(amap[0] == 4)

if __name__ == "__main__":
    unittest.main()
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20060616/8bb3dcc0/attachment.htm>

From s_sourceforge at nedprod.com  Fri Jun 16 19:37:46 2006
From: s_sourceforge at nedprod.com (Niall Douglas)
Date: Fri, 16 Jun 2006 18:37:46 +0100
Subject: [C++-sig] Inserting custom code into pyplusplus output
In-Reply-To: <7465b6170606151108w561f5883x33fbebd63add0149@mail.gmail.com>
References: <44918F02.20434.108ABEB@s_sourceforge.nedprod.com>
Message-ID: <4492FA7A.20748.26B63F7@s_sourceforge.nedprod.com>

On 15 Jun 2006 at 21:08, Roman Yakovenko wrote:

> > No, this appends into TnFOX.main.cpp. I need it appended into
> > FXApp.pypp.cpp. If you examine file_writers/multiple_files.py, you
> > can see it writes license, precompiled header, global headers, global
> > namespace aliases, wrapper classes and the register function for the
> > class in question. Nowhere is it letting you write custom text
> > outside the bodies of these items. I suppose you could add a custom
> > form of wrapper code creator, but this seems inflexible - you should
> > be able to insert text anywhere in any output file.
> 
> :-). This is one of my big problems - I don't know what interface to
> give to the user.
> It is not very difficult to implement what you are asking for.  I am
> sure you have an
> idea, right ?

I was thinking that you should replace the wrapper property in 
class_t with a more generic code_creators, of which one is the 
wrapper class (if needed). Call it 'classfilecreators' or something. 
Then to add things per class file, you simply adopt a custom text 
creator into it.

You'd need to be careful when using the single file writer as some 
items may duplicate. Of course, this is probably an end user problem.

> As a temporal solution you can create custom file writer. Or may be to write
> small script that will add the code to the file.

I'll wait for your solution and use diff and patch in the meantime.

Cheers,
Niall





From roman.yakovenko at gmail.com  Fri Jun 16 19:57:11 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Fri, 16 Jun 2006 20:57:11 +0300
Subject: [C++-sig] Reducing size of binary
In-Reply-To: <7f03db650606160729u4dc6bc7em4bf795d0b0cb1287@mail.gmail.com>
References: <7f03db650606160729u4dc6bc7em4bf795d0b0cb1287@mail.gmail.com>
Message-ID: <7465b6170606161057i4bf0c10bjc52ae32e76b3d51c@mail.gmail.com>

On 6/16/06, Will Lee <lee.will at gmail.com> wrote:
> Hi,
>
> I'm currently using Boost Python in a fairly large project using Linux and
> g++.  The combined .so code that I tried to wrap is around 9MB (-O with no
> symbol).  I got to a point where the python wrapper binary gets pretty big
> (sometimes the wrapper debug .so with symbols can get to 350 MB and the -O
> version without symobls would be around 18 MB).  Compile time can get up to
> 15-20 minutes.  As far as I see, I'm not doing anything too extraordinary.
> I am using a lot of add_property and def methods in the wrapper code.
>
> Does anyone have hints on reducing the size of the wrapper binary?

It could be useful to know your g++ version and compilation line.
Meanwhile you can
read this: http://www.nedprod.com/programs/gccvisibility.html

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From s_sourceforge at nedprod.com  Fri Jun 16 21:34:17 2006
From: s_sourceforge at nedprod.com (Niall Douglas)
Date: Fri, 16 Jun 2006 20:34:17 +0100
Subject: [C++-sig] More pyplusplus custom stuff
In-Reply-To: <7465b6170606151130l23eb0fan54fd7bf3fa5d8ba9@mail.gmail.com>
References: <44915545.5613.272220@s_sourceforge.nedprod.com>
Message-ID: <449315C9.9331.2D611F6@s_sourceforge.nedprod.com>

On 15 Jun 2006 at 21:30, Roman Yakovenko wrote:

> 1. add pyplusplus.decl_wrappers.constructor_t new property
>     body ( string )
> 
> 2.  add code that insert this member variable into generated code in
> next classes:
>      pyplusplus.code_creators.
>          constructor_wrapper_t
>          special_constructor_wrapper_t
>          trivial_constructor_wrapper_t

Firstly, I attach modified pyplusplus source implementing your 
suggestion.

For special_constructor_wrapper_t and trivial_constructor_wrapper_t, 
self.declaration refers to a decl_wrapper.class_t rather than 
decl_wrapper.constructor_t. AFAICS trivial constructors are really 
compiler generated null constructors while special constructors are 
really copy constructors. Therefore, you'd probably want to be able 
to set each individually so I split them into a series of properties. 
I have a new function which sets all constructor bodies.

Cheers,
Niall



-------------- next part --------------
The following section of this message contains a file attachment
prepared for transmission using the Internet MIME message format.
If you are using Pegasus Mail, or any other MIME-compliant system,
you should be able to save it or view it from within your mailer.
If you cannot, please ask your system administrator for assistance.

   ---- File information -----------
     File:  calldef.py
     Date:  16 Jun 2006, 15:03
     Size:  48750 bytes.
     Type:  Unknown
-------------- next part --------------
A non-text attachment was scrubbed...
Name: calldef.py
Type: application/octet-stream
Size: 48749 bytes
Desc: not available
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20060616/06f3bb59/attachment.obj>
-------------- next part --------------
The following section of this message contains a file attachment
prepared for transmission using the Internet MIME message format.
If you are using Pegasus Mail, or any other MIME-compliant system,
you should be able to save it or view it from within your mailer.
If you cannot, please ask your system administrator for assistance.

   ---- File information -----------
     File:  class_wrapper.py
     Date:  16 Jun 2006, 15:02
     Size:  5402 bytes.
     Type:  Unknown
-------------- next part --------------
A non-text attachment was scrubbed...
Name: class_wrapper.py
Type: application/octet-stream
Size: 5401 bytes
Desc: not available
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20060616/06f3bb59/attachment-0001.obj>
-------------- next part --------------
The following section of this message contains a file attachment
prepared for transmission using the Internet MIME message format.
If you are using Pegasus Mail, or any other MIME-compliant system,
you should be able to save it or view it from within your mailer.
If you cannot, please ask your system administrator for assistance.

   ---- File information -----------
     File:  calldef_wrapper.py
     Date:  16 Jun 2006, 14:27
     Size:  13426 bytes.
     Type:  Unknown
-------------- next part --------------
A non-text attachment was scrubbed...
Name: calldef_wrapper.py
Type: application/octet-stream
Size: 13426 bytes
Desc: not available
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20060616/06f3bb59/attachment-0002.obj>

From s_sourceforge at nedprod.com  Fri Jun 16 21:34:18 2006
From: s_sourceforge at nedprod.com (Niall Douglas)
Date: Fri, 16 Jun 2006 20:34:18 +0100
Subject: [C++-sig] Reducing size of binary
In-Reply-To: <7f03db650606160729u4dc6bc7em4bf795d0b0cb1287@mail.gmail.com>
Message-ID: <449315CA.12403.2D6163C@s_sourceforge.nedprod.com>

On 16 Jun 2006 at 9:29, Will Lee wrote:

> I'm currently using Boost Python in a fairly large project using Linux and
> g++.  The combined .so code that I tried to wrap is around 9MB (-O with no
> symbol).  I got to a point where the python wrapper binary gets pretty big
> (sometimes the wrapper debug .so with symbols can get to 350 MB and the -O
> version without symobls would be around 18 MB).  Compile time can get up to
> 15-20 minutes.  As far as I see, I'm not doing anything too extraordinary.
> I am using a lot of add_property and def methods in the wrapper code.
> 
> Does anyone have hints on reducing the size of the wrapper binary?

That's nothing. I'm regularly generating 25Mb release build bindings 
on Linux. They take several hours to compile.

If you're using gcc 4.x, use -fvisibility=hidden. Also optimise for 
smallest possible size at expense of speed. Use 3.4 or later rather 
than 3.3 or earlier.

Ultimately though, ELF is a fat binary format and how GCC stores 
typeinfo is space expensive. Mach-O on Apple and PE on Windows are 
far far smaller for the same source.

Cheers,
Niall





From rwgk at yahoo.com  Sat Jun 17 01:48:18 2006
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Fri, 16 Jun 2006 16:48:18 -0700 (PDT)
Subject: [C++-sig] Bug with custom convertor and stl container?
In-Reply-To: <7f03db650606161017k28035c1dw6ba2a57d2017cf4@mail.gmail.com>
Message-ID: <20060616234818.62358.qmail@web31501.mail.mud.yahoo.com>

--- Will Lee <lee.will at gmail.com> wrote:

> It seems like there's a bug when I use a custom converter (as described in
> the FAQ for the custom string) and a wrapped stl container.  If I define a
> custom converter and use that type in a std::vector or std::map, I'll get an
> "TypeError: No Python class registered for C++" error for that converted
> type.
> 
> For example, in the following case where I'm converting from type A to a
> python integer, std::vector<A> and std::map<int, A> are not working
> propertly.  The testMap and testVec unit tests will both fail.  This is
> somewhat a showstopper so it would be great if you have any idea on how to
> get around this.

You probably have to tell the vector_indexing_suite and the map_indexing_suite
to also return the result of getitem by value. I don't know how to do this but
I believe it is possible.


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 


From s_sourceforge at nedprod.com  Sun Jun 18 05:14:40 2006
From: s_sourceforge at nedprod.com (Niall Douglas)
Date: Sun, 18 Jun 2006 04:14:40 +0100
Subject: [C++-sig] Another bug in pyplusplus
Message-ID: <4494D330.5974.34333B4@s_sourceforge.nedprod.com>

If you disable use of keywords using 
mb.calldefs().use_keywords=False, pyplusplus also stops inserting 
default argument values into its wrappers eg;

int FXDialogBox::execute(int placement=DEFAULT)

... should become

    virtual ::FX::FXuint execute( ::FX::FXuint placement = 
FX::DEFAULT ) {
        if( bp::override func_execute = this->get_override( "execute" 
) )
            return func_execute( placement );
        else
            return FX::FXDialogBox::execute( placement );
    }

Unfortunately, it's missing the default argument. To fix this, 
replace calldef_wrapper_t.args_declaration() with:

    def args_declaration( self ):
        args = []
        boost_obj = algorithm.create_identifier( self, 
'::boost::python::object' )
        for index, arg in enumerate( self.declaration.arguments ):
            result = arg.type.decl_string + ' ' + 
self.argument_name(index)
            if arg.default_value:
                if not declarations.is_pointer( arg.type ) or 
arg.default_value != '0':
                    result += '=%s' % arg.default_value
                else:
                    result += '=%s()' % boost_obj
            args.append( result )
        if len( args ) == 1:
            return args[ 0 ] 
        return ', '.join( args )

Another point - I believe the default_X function doesn't need to be 
virtual. By making it so, you're introducing an unnecessary 
inefficiency:    
    
    virtual ::FX::FXuint default_execute( ::FX::FXuint placement ) {
        return FX::FXDialogBox::execute( placement );
    }

Cheers,
Niall





From roman.yakovenko at gmail.com  Sun Jun 18 07:44:59 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Sun, 18 Jun 2006 08:44:59 +0300
Subject: [C++-sig] Another bug in pyplusplus
In-Reply-To: <4494D330.5974.34333B4@s_sourceforge.nedprod.com>
References: <4494D330.5974.34333B4@s_sourceforge.nedprod.com>
Message-ID: <7465b6170606172244o80943d4kfd7d6259ca163b1b@mail.gmail.com>

On 6/18/06, Niall Douglas <s_sourceforge at nedprod.com> wrote:
> If you disable use of keywords using
> mb.calldefs().use_keywords=False, pyplusplus also stops inserting
> default argument values into its wrappers eg;

May I correct you? All wrappers have been generated without default
arguments, right :-)?
It has nothing to do with use_keywords

>     def args_declaration( self ):
>         args = []
>         boost_obj = algorithm.create_identifier( self,
> '::boost::python::object' )
>         for index, arg in enumerate( self.declaration.arguments ):
>             result = arg.type.decl_string + ' ' +
> self.argument_name(index)
>             if arg.default_value:
>                 if not declarations.is_pointer( arg.type ) or
> arg.default_value != '0':
>                     result += '=%s' % arg.default_value
>                 else:
>                     result += '=%s()' % boost_obj

I could be wrong, but it seems to me that we don't need
" if not declarations.is_pointer( arg.type ) or arg.default_value != '0': "
condition, just plain
"result += '=%s' % arg.default_value"
will do the job, right? Am I missing something?

> Another point - I believe the default_X function doesn't need to be
> virtual. By making it so, you're introducing an unnecessary
> inefficiency:

Fixed.

Thank you!

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From roman.yakovenko at gmail.com  Sun Jun 18 08:42:21 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Sun, 18 Jun 2006 09:42:21 +0300
Subject: [C++-sig] More pyplusplus custom stuff
In-Reply-To: <449315C9.9331.2D611F6@s_sourceforge.nedprod.com>
References: <44915545.5613.272220@s_sourceforge.nedprod.com>
	<7465b6170606151130l23eb0fan54fd7bf3fa5d8ba9@mail.gmail.com>
	<449315C9.9331.2D611F6@s_sourceforge.nedprod.com>
Message-ID: <7465b6170606172342x1d5f32c8g528f46bbcdc43453@mail.gmail.com>

On 6/16/06, Niall Douglas <s_sourceforge at nedprod.com> wrote:

Niall, thank you for patches! I committed them.

> For special_constructor_wrapper_t and trivial_constructor_wrapper_t,
> self.declaration refers to a decl_wrapper.class_t rather than
> decl_wrapper.constructor_t. AFAICS trivial constructors are really
> compiler generated null constructors while special constructors are
> really copy constructors.

I switched to your terminology.

> Therefore, you'd probably want to be able
> to set each individually so I split them into a series of properties.
> I have a new function which sets all constructor bodies.

Now class_t have null_constructor_body and copy_constructor_body. If you use
set_constructors_body method you should be fine.

If you are going to check out latest source code - be careful. Latest
version has new
feature - indexing suite.
http://language-binding.net/pyplusplus/peps/indexing_suite.html - this
has been implemented.

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From s_sourceforge at nedprod.com  Sun Jun 18 14:48:00 2006
From: s_sourceforge at nedprod.com (Niall Douglas)
Date: Sun, 18 Jun 2006 13:48:00 +0100
Subject: [C++-sig] Another bug in pyplusplus
In-Reply-To: <7465b6170606172244o80943d4kfd7d6259ca163b1b@mail.gmail.com>
References: <4494D330.5974.34333B4@s_sourceforge.nedprod.com>
Message-ID: <44955990.15451.D1398@s_sourceforge.nedprod.com>

On 18 Jun 2006 at 8:44, Roman Yakovenko wrote:

> On 6/18/06, Niall Douglas <s_sourceforge at nedprod.com> wrote:
> > If you disable use of keywords using
> > mb.calldefs().use_keywords=False, pyplusplus also stops inserting
> > default argument values into its wrappers eg;
> 
> May I correct you? All wrappers have been generated without default
> arguments, right :-)?
> It has nothing to do with use_keywords

It's quite possible. However, I know when use_keywords is true, it 
puts arg("x")=5 and such into declarations.

> >     def args_declaration( self ):
> >         args = []
> >         boost_obj = algorithm.create_identifier( self,
> > '::boost::python::object' )
> >         for index, arg in enumerate( self.declaration.arguments ):
> >             result = arg.type.decl_string + ' ' +
> > self.argument_name(index)
> >             if arg.default_value:
> >                 if not declarations.is_pointer( arg.type ) or
> > arg.default_value != '0':
> >                     result += '=%s' % arg.default_value
> >                 else:
> >                     result += '=%s()' % boost_obj
> 
> I could be wrong, but it seems to me that we don't need
> " if not declarations.is_pointer( arg.type ) or arg.default_value != '0': "
> condition, just plain
> "result += '=%s' % arg.default_value"
> will do the job, right? Am I missing something?

No you're absolutely right. I wrote that email at something like 3am 
last night and I hadn't quite tested it properly. What you wrote is 
correct.

> > Another point - I believe the default_X function doesn't need to be
> > virtual. By making it so, you're introducing an unnecessary
> > inefficiency:
> 
> Fixed.
> 
> Thank you!

No problem. I have the imageviewer.py example very nearly working. 
Once it is working, I would consider the python bindings finished 
apart from three items:

1. Named parameters do not work which is annoying because BPL throws 
an exception if its type is as yet unregistered. This problem can't 
be fixed by pyplusplus - it requires fixing BPL.

2. FXWinShellLink causes pyplusplus to generate bad code. This is 
because it uses unions.

3. FXImage subclasses don't provide a working getData(). This 
requires custom body text insertion to be added to pyplusplus.

Otherwise, they're finished! Thanks for all the help!

Cheers,
Niall





From roman.yakovenko at gmail.com  Sun Jun 18 15:46:41 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Sun, 18 Jun 2006 16:46:41 +0300
Subject: [C++-sig] Another bug in pyplusplus
In-Reply-To: <44955990.15451.D1398@s_sourceforge.nedprod.com>
References: <4494D330.5974.34333B4@s_sourceforge.nedprod.com>
	<7465b6170606172244o80943d4kfd7d6259ca163b1b@mail.gmail.com>
	<44955990.15451.D1398@s_sourceforge.nedprod.com>
Message-ID: <7465b6170606180646x6db4b173w635d09287a839407@mail.gmail.com>

On 6/18/06, Niall Douglas <s_sourceforge at nedprod.com> wrote:
> No problem. I have the imageviewer.py example very nearly working.
> Once it is working, I would consider the python bindings finished
> apart from three items:
>
> 1. Named parameters do not work which is annoying because BPL throws
> an exception if its type is as yet unregistered. This problem can't
> be fixed by pyplusplus - it requires fixing BPL.

class_t has property - always_expose_using_scope. If you set it to true, then
it can help you in next case:

struct A{
    void do_smth( const A& a=A() ){...}
};

Generated code will not throw exception, while registering the function.

> 2. FXWinShellLink causes pyplusplus to generate bad code. This is
> because it uses unions.

Not just unions, but unnamed ones :-)

I work on a project that has same problem, you have good chance that I
will add missing
functionality.

> 3. FXImage subclasses don't provide a working getData(). This
> requires custom body text insertion to be added to pyplusplus.

Can you send short example what pyplusplus generates now and the code you need
to be generated, may be I will find a solution.

> Otherwise, they're finished! Thanks for all the help!

My pleasure :-)


-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From s_sourceforge at nedprod.com  Sun Jun 18 16:38:23 2006
From: s_sourceforge at nedprod.com (Niall Douglas)
Date: Sun, 18 Jun 2006 15:38:23 +0100
Subject: [C++-sig] Default arguments in pyplusplus
Message-ID: <4495736F.11535.722233@s_sourceforge.nedprod.com>

Hmm, pyplusplus doesn't tell BPL about default arguments at all! 
Surely it should output BOOST_PYTHON_MEMBER_FUNCTION_OVERLOADS() for 
each overloaded function?

Needless to say, the TnFOX python bindings won't work without default 
arguments working :(

I did try adding this on my own, but it's complex. I'm afraid the 
class hierarchy in pyplusplus is fairly complex, complex enough that 
without a debugger it's hard to modify. So I'll have to leave it to 
you Roman!

Cheers,
Niall





From s_sourceforge at nedprod.com  Sun Jun 18 16:47:28 2006
From: s_sourceforge at nedprod.com (Niall Douglas)
Date: Sun, 18 Jun 2006 15:47:28 +0100
Subject: [C++-sig] Another bug in pyplusplus
In-Reply-To: <7465b6170606180646x6db4b173w635d09287a839407@mail.gmail.com>
References: <44955990.15451.D1398@s_sourceforge.nedprod.com>
Message-ID: <44957590.13157.7A707B@s_sourceforge.nedprod.com>

On 18 Jun 2006 at 16:46, Roman Yakovenko wrote:

> On 6/18/06, Niall Douglas <s_sourceforge at nedprod.com> wrote:
> > No problem. I have the imageviewer.py example very nearly working.
> > Once it is working, I would consider the python bindings finished
> > apart from three items:
> >
> > 1. Named parameters do not work which is annoying because BPL throws
> > an exception if its type is as yet unregistered. This problem can't
> > be fixed by pyplusplus - it requires fixing BPL.
> 
> class_t has property - always_expose_using_scope. If you set it to true, then
> it can help you in next case:
> 
> struct A{
>     void do_smth( const A& a=A() ){...}
> };
> 
> Generated code will not throw exception, while registering the function.

I already have this set to true. I more meant:

void foo(int a=1, int b=2, int c=3)

... and then from python do:

foo(c=8)

You need to declare parameter names to do this, but BPL faults when 
the type is (say) an enum defined in the class it's about to declare.

> > 3. FXImage subclasses don't provide a working getData(). This
> > requires custom body text insertion to be added to pyplusplus.
> 
> Can you send short example what pyplusplus generates now and the code you need
> to be generated, may be I will find a solution.

All I need is the ability to insert custom code into the body of 
class files. Put simply, I need to insert DEFINEMAKECARRAYITER at the 
top of every FXImage.pypp.cpp, FXBMPImage.pypp.cpp, 
FXJPGImage.pypp.cpp etc.

It needs to go before or after the wrapper class. Declared inside the 
body.

If you add an overloads section (see previous email) containing all 
the BOOST_PYTHON_MEMBER_FUNCTION_OVERLOADS declarations (which MUST 
come after the wrapper class but before the register function), then 
I could add them there as a custom text code creator.

Cheers,
Niall





From roman.yakovenko at gmail.com  Sun Jun 18 19:41:08 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Sun, 18 Jun 2006 20:41:08 +0300
Subject: [C++-sig] Default arguments in pyplusplus
In-Reply-To: <4495736F.11535.722233@s_sourceforge.nedprod.com>
References: <4495736F.11535.722233@s_sourceforge.nedprod.com>
Message-ID: <7465b6170606181041n41fd837et4c01db41d978c323@mail.gmail.com>

On 6/18/06, Niall Douglas <s_sourceforge at nedprod.com> wrote:
> Hmm, pyplusplus doesn't tell BPL about default arguments at all!

Niall, I just added the tests that check your statement. I can not reproduce
the behaviour you are talking about.

namespace default_args{

struct data{
    int sum( int i=0 ){ return i; }
    int sum( int i, int j, int k=3 ){ return i + j + k; }
};

}

pyplusplus will generate next code:

namespace bp = boost::python;

BOOST_PYTHON_MODULE(default_args){
    bp::class_< default_args::data >( "data" )
        .def("sum"
                , (int ( ::default_args::data::* )( int ) )(
&::default_args::data::sum )
                , ( bp::arg("i")=0 )
                , bp::default_call_policies() )
        .def("sum"
                , (int ( ::default_args::data::* )( int,int,int ) )(
&::default_args::data::sum )
                , ( bp::arg("i"), bp::arg("j"), bp::arg("k")=3 )
                , bp::default_call_policies() );
}

My test is next:

d = module.data()
self.failUnless( d.sum() == 0 )
self.failUnless( d.sum( 1, 2 ) == 6 )

I would like to see your use case

> Surely it should output BOOST_PYTHON_MEMBER_FUNCTION_OVERLOADS() for
> each overloaded function?

BOOST_PYTHON_MEMBER_FUNCTION_OVERLOADS has limit: it can not be used
for functions that takes different type of arguments:

double sum( int );
double sum( double, double );

You can't not expose them using the macro.

> Needless to say, the TnFOX python bindings won't work without default
> arguments working :(

I agree with you, may be you have some where in the code use_keywords = False?

> I did try adding this on my own, but it's complex. I'm afraid the
> class hierarchy in pyplusplus is fairly complex, complex enough that
> without a debugger it's hard to modify. So I'll have to leave it to
> you Roman!

Please provide test case and I will try to fix ASAP.

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From roman.yakovenko at gmail.com  Sun Jun 18 19:57:43 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Sun, 18 Jun 2006 20:57:43 +0300
Subject: [C++-sig] Another bug in pyplusplus
In-Reply-To: <44957590.13157.7A707B@s_sourceforge.nedprod.com>
References: <44955990.15451.D1398@s_sourceforge.nedprod.com>
	<7465b6170606180646x6db4b173w635d09287a839407@mail.gmail.com>
	<44957590.13157.7A707B@s_sourceforge.nedprod.com>
Message-ID: <7465b6170606181057p5abea875h5a6aa17cf0a323be@mail.gmail.com>

On 6/18/06, Niall Douglas <s_sourceforge at nedprod.com> wrote:
> > > 3. FXImage subclasses don't provide a working getData(). This
> > > requires custom body text insertion to be added to pyplusplus.
> >
> > Can you send short example what pyplusplus generates now and the code you need
> > to be generated, may be I will find a solution.
>
> All I need is the ability to insert custom code into the body of
> class files. Put simply, I need to insert DEFINEMAKECARRAYITER at the
> top of every FXImage.pypp.cpp, FXBMPImage.pypp.cpp,
> FXJPGImage.pypp.cpp etc.
>
> It needs to go before or after the wrapper class. Declared inside the
> body.
>
> If you add an overloads section (see previous email) containing all
> the BOOST_PYTHON_MEMBER_FUNCTION_OVERLOADS declarations (which MUST
> come after the wrapper class but before the register function), then
> I could add them there as a custom text code creator.

There is a solution. You can generate all code that uses DEFINEMAKECARRAYITER
in some header file and then to include it.

mb = module_builder_t( ... )
for cls in mb.classes( lambda decl: decl.name.endswith( 'Image' ) ):
    generate code to some header file
mb.build_code_creator( ... )
mb.code_creator.add_include( that header file )

I have another solution, but I don't have time to implement it or even check:
pyplusplus can divide every generated cpp file to sections:

1. include
2. space
3. wrapper
4. space
5. class

Then, pyplusplus can define next class:

class on_cpp_file_write_i:
    def on_include( self, declaration ):
     ...
    def on_first_space
    def on_wrapper
    def on_second space
    def on_class

Every function return some text or None. In case method returns None
pyplusplus will write the code it have to write. In case method returns some
text, this text will be written instead.

Thoughts?

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From s_sourceforge at nedprod.com  Mon Jun 19 02:34:39 2006
From: s_sourceforge at nedprod.com (Niall Douglas)
Date: Mon, 19 Jun 2006 01:34:39 +0100
Subject: [C++-sig] Another bug in pyplusplus
In-Reply-To: <7465b6170606181057p5abea875h5a6aa17cf0a323be@mail.gmail.com>
References: <44957590.13157.7A707B@s_sourceforge.nedprod.com>
Message-ID: <4495FF2F.6696.103EB1@s_sourceforge.nedprod.com>

On 18 Jun 2006 at 20:57, Roman Yakovenko wrote:

> There is a solution. You can generate all code that uses DEFINEMAKECARRAYITER
> in some header file and then to include it.

I was even thinking of just generating the header file from within 
the python script.

> I have another solution, but I don't have time to implement it or even check:
> pyplusplus can divide every generated cpp file to sections:
> 
> 1. include
> 2. space
> 3. wrapper
> 4. space
> 5. class
> 
> Then, pyplusplus can define next class:
> 
> class on_cpp_file_write_i:
>     def on_include( self, declaration ):
>      ...
>     def on_first_space
>     def on_wrapper
>     def on_second space
>     def on_class
> 
> Every function return some text or None. In case method returns None
> pyplusplus will write the code it have to write. In case method returns some
> text, this text will be written instead.
> 
> Thoughts?

Looks reasonable.

Cheers,
Niall





From s_sourceforge at nedprod.com  Mon Jun 19 02:34:38 2006
From: s_sourceforge at nedprod.com (Niall Douglas)
Date: Mon, 19 Jun 2006 01:34:38 +0100
Subject: [C++-sig] Default arguments in pyplusplus
In-Reply-To: <7465b6170606181041n41fd837et4c01db41d978c323@mail.gmail.com>
References: <4495736F.11535.722233@s_sourceforge.nedprod.com>
Message-ID: <4495FF2E.10980.103E62@s_sourceforge.nedprod.com>

On 18 Jun 2006 at 20:41, Roman Yakovenko wrote:

> > Needless to say, the TnFOX python bindings won't work without default
> > arguments working :(
> 
> I agree with you, may be you have some where in the code use_keywords = False?

That's precisely the case. I am talking about default parameters not 
working when use_keywords is False. Sorry, I thought I mentioned that 
earlier in a previous bug report.

I think surely you will need to use 
BOOST_PYTHON_MEMBER_FUNCTION_OVERLOADS() for when keywords are not 
enabled?

> > I did try adding this on my own, but it's complex. I'm afraid the
> > class hierarchy in pyplusplus is fairly complex, complex enough that
> > without a debugger it's hard to modify. So I'll have to leave it to
> > you Roman!
> 
> Please provide test case and I will try to fix ASAP.

I think that I have use_keywords as False is enough? Try your unit 
test with use_keywords as False, if that still works I will modify 
one of your unit tests to show the behaviour.

Cheers,
Niall





From roman.yakovenko at gmail.com  Mon Jun 19 07:17:48 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Mon, 19 Jun 2006 08:17:48 +0300
Subject: [C++-sig] Default arguments in pyplusplus
In-Reply-To: <4495FF2E.10980.103E62@s_sourceforge.nedprod.com>
References: <4495736F.11535.722233@s_sourceforge.nedprod.com>
	<7465b6170606181041n41fd837et4c01db41d978c323@mail.gmail.com>
	<4495FF2E.10980.103E62@s_sourceforge.nedprod.com>
Message-ID: <7465b6170606182217l2827f921g90ddba239a5e3b8e@mail.gmail.com>

On 6/19/06, Niall Douglas <s_sourceforge at nedprod.com> wrote:
> On 18 Jun 2006 at 20:41, Roman Yakovenko wrote:
>
> > > Needless to say, the TnFOX python bindings won't work without default
> > > arguments working :(
> >
> > I agree with you, may be you have some where in the code use_keywords = False?
>
> That's precisely the case. I am talking about default parameters not
> working when use_keywords is False. Sorry, I thought I mentioned that
> earlier in a previous bug report.
>
> I think surely you will need to use
> BOOST_PYTHON_MEMBER_FUNCTION_OVERLOADS() for when keywords are not
> enabled?

As I already said: this macro has limitations. It can not be used in
general case.

> > > I did try adding this on my own, but it's complex. I'm afraid the
> > > class hierarchy in pyplusplus is fairly complex, complex enough that
> > > without a debugger it's hard to modify. So I'll have to leave it to
> > > you Roman!
> >
> > Please provide test case and I will try to fix ASAP.
>
> I think that I have use_keywords as False is enough?

Yes this is enough.

> Try your unit
> test with use_keywords as False, if that still works I will modify
> one of your unit tests to show the behaviour.

Why don't you resolve the issue with enums. I see few ways to go:

To find places where enum used as default arguments and enum still not exported
>From here you have 2 solutions: to remove default arguments or to change default
arguments to have integer type with value of enum. This should work.

Another solution is to move enum to from class to FX namespace. In
this case it will be
exported earlier.

There is third approach:
Every function has "overloads" property. You can find all member
functions, that are
overloaded. Then you can exclude them. After this, generate code that
uses the macro
in one file ( may be it is a time to implement interface that will
allow to edit generated cpp
file ). And finally you can add the code to the class using add_code.

What do you think?

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From roman.yakovenko at gmail.com  Mon Jun 19 07:53:04 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Mon, 19 Jun 2006 08:53:04 +0300
Subject: [C++-sig] Ogre bindings - good news
Message-ID: <7465b6170606182253h64d73148od61cffb0e9272f72@mail.gmail.com>

Hi. I just want to share good news.

Lakin Wecker recently worked on creating Python bindings for Ogre (
Object-Oriented Graphics Rendering Engine http://www.ogre3d.org/ )
using pyplusplus and boost.python.
Ogre project already has Python bindings using SWIG, but maintenance
process was and still
is a nightmare and very time consuming. But now when he started to use
pyplusplus it is a
different story.

For full story please read:

http://lakin.weckers.net/index_ogre_python.html
http://www.ogre3d.org/phpBB2addons/viewtopic.php?t=1478

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From lee.will at gmail.com  Mon Jun 19 17:33:53 2006
From: lee.will at gmail.com (Will Lee)
Date: Mon, 19 Jun 2006 10:33:53 -0500
Subject: [C++-sig] Bug with custom convertor and stl container?
In-Reply-To: <20060616234818.62358.qmail@web31501.mail.mud.yahoo.com>
References: <7f03db650606161017k28035c1dw6ba2a57d2017cf4@mail.gmail.com>
	<20060616234818.62358.qmail@web31501.mail.mud.yahoo.com>
Message-ID: <7f03db650606190833l578ba5e5h77a1aacc73710453@mail.gmail.com>

I have resolved this problem when I set the template NoProxy to true.
However, another problem comes up when I try to use the map in python:

If I do the following in my unit test:

    def testMap(self):
        amap = atest.AMap()
        amap[3] = 4
        self.assert_(amap[3] == 4)

        # This should not fail?
        for i in amap:
            self.assert_(i.data() == 4)

I'd get a error that looks like:

======================================================================
ERROR: testMap (__main__.ATest)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "ATest.py", line 20, in testMap
    self.assert_(i.data() == 4)
TypeError: No Python class registered for C++ class A

----------------------------------------------------------------------
Ran 3 tests in 0.005s

FAILED (errors=1)

Note the first assert would pass after I set NoProxy to false, but the
data() call would get me the same error.  The type of "i" is <class '
atest.map_indexing_suite_AMap_entry'>.  Is there a bug in the generated
entry object if I use a custom converter?

Will

On 6/16/06, Ralf W. Grosse-Kunstleve <rwgk at yahoo.com> wrote:
>
> --- Will Lee <lee.will at gmail.com> wrote:
>
> > It seems like there's a bug when I use a custom converter (as described
> in
> > the FAQ for the custom string) and a wrapped stl container.  If I define
> a
> > custom converter and use that type in a std::vector or std::map, I'll
> get an
> > "TypeError: No Python class registered for C++" error for that converted
> > type.
> >
> > For example, in the following case where I'm converting from type A to a
> > python integer, std::vector<A> and std::map<int, A> are not working
> > propertly.  The testMap and testVec unit tests will both fail.  This is
> > somewhat a showstopper so it would be great if you have any idea on how
> to
> > get around this.
>
> You probably have to tell the vector_indexing_suite and the
> map_indexing_suite
> to also return the result of getitem by value. I don't know how to do this
> but
> I believe it is possible.
>
>
> __________________________________________________
> Do You Yahoo!?
> Tired of spam?  Yahoo! Mail has the best spam protection around
> http://mail.yahoo.com
> _______________________________________________
> C++-sig mailing list
> C++-sig at python.org
> http://mail.python.org/mailman/listinfo/c++-sig
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20060619/011a2e99/attachment.htm>

From cyril.bazin at info.unicaen.fr  Mon Jun 19 17:35:39 2006
From: cyril.bazin at info.unicaen.fr (Cyril Bazin)
Date: Mon, 19 Jun 2006 17:35:39 +0200
Subject: [C++-sig] ArgumentError...
Message-ID: <a10b8d620606190835w5fa04d3cg98789bb2cfef9b2d@mail.gmail.com>

Hello,

I am trying to use boost.python in order to bind a library called CGAL (and
also to test if it's diffcult to bind such a library ;-).
http://www.cgal.org/
This library provide a class CGAL::Delaunay_Triangulation_2 which represent
a Delaunay triangulation in two dimensions.

I want to build a delaunay triangultion using the "push_back" method (which
add a Point to the mesh).
Then, I want to iter over the vertexes, the faces and the edges of the mesh
using a Python iterator which use the following methods :
 - Finite_vertices_iterator    t.finite_vertices_begin ()
 - Finite_vertices_iterator    t.finite_vertices_end ()


I added the following line in my C++ boost code :

   class_<Delaunay>("Delaunay")
       //...
       .def("finite_vertices", range(&Delaunay::finite_vertices_begin,
&Delaunay::finite_vertices_end))


And the following code in my unit tests :

class TestSequenceFunctions(unittest.TestCase):
   def init_delaunay(self):
        d = pygale.Delaunay()
        for x, y in self.seq:
            d.push_back(pygale.Point(x, y))
        return d

    def test_delaunay_2(self):
        d = self.init_delaunay()
        it = d.finite_vertices()


But when I call the method "finite_vertices", I get the following error :

Traceback (most recent call last):
   File "testPygale.py", line 40, in test_delaunay_2
     it = d.finite_vertices()
ArgumentError: Python argument types in
     Delaunay.finite_vertices(Delaunay)
did not match C++ signature:

finite_vertices(boost::python::back_reference<CGAL::Triangulation_2<CGAL::Cartesian<double>,
CGAL::Tds2<CGAL::Tvb<CGAL::Cartesian<double>, CGAL::Tdsvb<void> >,
CGAL::Tdsfb<void> > >&>)


If you can help me to solve this error, please...

Thanks in advance.


Cyril BAZIN


Annexes:
The classes to bind
The C++ code for boost
The python unittests
The execution trace


--------------------------------------------------------- The classes to
bind ---------------------------------------------
#include <CGAL/Cartesian.h>
#include <CGAL/squared_distance_2.h>
#include <CGAL/Delaunay_triangulation_2.h>
#include <CGAL/Triangulation_2.h>
#include <CGAL/Triangulation_face_base_2.h>
#include <CGAL/Triangulation_euclidean_traits_2.h>


#include <CGAL/Delaunay_triangulation_2.h>

#include <iostream>
#include <sstream>

typedef CGAL::Cartesian<double>    Gt;
typedef Gt::FT                     F_T;
typedef Gt::RT                     R_T;
typedef Gt::Point_2                Point;
typedef Gt::Segment_2              Segment;
typedef Gt::Line_2                 Line;
typedef Gt::Triangle_2             Triangle;
///typedef Gt::Bbox_2                 Bbox;
typedef CGAL::Delaunay_triangulation_2<Gt>      Delaunay;
typedef Delaunay::Finite_vertices_iterator   Finite_vertices_iterator;
typedef Delaunay::Vertex_handle              Vertex_handle;
typedef Delaunay::Vertex                     Vertex;

-----------------------------------------------------------------------------------------------------------------------------


---------------------------------------------------------- C++ code for
boost --------------------------------------------
#include <boost/python.hpp>
#include "pygale2.cpp"
using namespace boost::python;

BOOST_PYTHON_MODULE(pygale2)
{
   //......

   class_<Vertex>("Vertex")
    ;

   class_<Vertex_handle>("Vertex_handle")
    ;

   class_<Finite_vertices_iterator>("Finite_vertices_iterator")
    ;

   class_<Delaunay>("Delaunay")
        .def(init<>())
        .def("number_of_vertices", &Delaunay::number_of_vertices )
        .def("number_of_faces", &Delaunay::number_of_faces )
        .def("push_back", &Delaunay::push_back )
        .def("finite_vertices", range(&Delaunay::finite_vertices_begin,
&Delaunay::finite_vertices_end))
    ;
   //.....
}
------------------------------------------------------------------------------------------------------------------------------------


----------------------------------------------------The python unit tests
----------------------------------------------------
import pygale2 as pygale
import unittest

class TestSequenceFunctions(unittest.TestCase):

    def setUp(self):
        self.seq = [(5, 8), (9, 6), (5, 2), (12, 9), (0, 0), (2, 7)]

    #...

    def init_delaunay(self):
        d = pygale.Delaunay()
        for x, y in self.seq:
            d.push_back(pygale.Point(x, y))
        return d

    def test_delaunay(self):
        d = self.init_delaunay()
        self.assertEqual(d.number_of_vertices(), len(self.seq))

    def test_delaunay_2(self):
        d = self.init_delaunay()
        it = d.finite_vertices()


if __name__ == '__main__':
    suite = unittest.makeSuite(TestSequenceFunctions)
    unittest.TextTestRunner(verbosity=2).run(suite)
    #unittest.main()
-----------------------------------------------------------------------------------------------------------------------------------------

------------------------------------------- The execution trace
--------------------------------------------------------------
$make all test
g++ -I/export/home/cbazin/python/Python-2.4/ -I/usr/include/python2.4
-I/usr/local/include/boost-1_33_1/boost/ -I/usr/local/include
pygale2.boost.cpp -c -o pygale2.o
g++ pygale2.o -L/usr/local/lib/ -lboost_python -L/usr/lib/CGAL/ -lCGAL -o
pygale2.so --shared
python testPygale.py
test_delaunay (__main__.TestSequenceFunctions) ... ok
test_delaunay_2 (__main__.TestSequenceFunctions) ... ERROR
test_points (__main__.TestSequenceFunctions) ... ok
test_segments (__main__.TestSequenceFunctions) ... ok

======================================================================
ERROR: test_delaunay_2 (__main__.TestSequenceFunctions)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "testPygale.py", line 40, in test_delaunay_2
    it = d.finite_vertices()
ArgumentError: Python argument types in
    Delaunay.finite_vertices(Delaunay)
did not match C++ signature:
    finite_vertices(boost::python::back_reference<CGAL::Triangulation_2<CGAL::Cartesian<double>,
CGAL::Tds2<CGAL::Tvb<CGAL::Cartesian<double>, CGAL::Tdsvb<void> >,
CGAL::Tdsfb<void> > >&>)

----------------------------------------------------------------------
Ran 4 tests in 0.003s

FAILED (errors=1)
----------------------------------------------------------------------------------------------------------------------------------------
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20060619/a67253a9/attachment.htm>

From dave at boost-consulting.com  Mon Jun 19 23:35:55 2006
From: dave at boost-consulting.com (David Abrahams)
Date: Mon, 19 Jun 2006 17:35:55 -0400
Subject: [C++-sig] ArgumentError...
References: <a10b8d620606190835w5fa04d3cg98789bb2cfef9b2d@mail.gmail.com>
Message-ID: <uveqwomkk.fsf@boost-consulting.com>

"Cyril Bazin" <cyril.bazin at info.unicaen.fr> writes:

> But when I call the method "finite_vertices", I get the following error :
>
> Traceback (most recent call last):
>   File "testPygale.py", line 40, in test_delaunay_2
>     it = d.finite_vertices()
> ArgumentError: Python argument types in
>     Delaunay.finite_vertices(Delaunay)
> did not match C++ signature:
>
> finite_vertices(boost::python::back_reference<CGAL::Triangulation_2<CGAL::Cartesian<double>,
> CGAL::Tds2<CGAL::Tvb<CGAL::Cartesian<double>, CGAL::Tdsvb<void> >,
> CGAL::Tdsfb<void> > >&>)
>
>
> If you can help me to solve this error, please...

My best guess is that Delaunay is derived from 

  CGAL::Triangulation_2<
       CGAL::Cartesian<double>
     , CGAL::Tds2<
           CGAL::Tvb<CGAL::Cartesian<double>, CGAL::Tdsvb<void> >
         , CGAL::Tdsfb<void> 
       > 
  >

which supplies finite_vertices_begin and finite_vertices_end.  Since
you haven't wrapped that, and Boost.Python doesn't know it's a base
class of Delaunay, you get the error.  

You have two choices:

1. pre-assign finite_vertices_end/begin to variables (or cast to
   temporaries) of type 

      Finite_vertices_iterator (Delaunay::*)()

   and pass those to range(...)

2. Use the first form of range<NextPolicies,Target>(...) described
   here: http://boost.org/libs/python/doc/v2/iterator.html#range-spec

HTH,

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com



From joel at boost-consulting.com  Tue Jun 20 02:33:31 2006
From: joel at boost-consulting.com (Joel de Guzman)
Date: Tue, 20 Jun 2006 08:33:31 +0800
Subject: [C++-sig] Bug with custom convertor and stl container?
In-Reply-To: <7f03db650606190833l578ba5e5h77a1aacc73710453@mail.gmail.com>
References: <7f03db650606161017k28035c1dw6ba2a57d2017cf4@mail.gmail.com>	<20060616234818.62358.qmail@web31501.mail.mud.yahoo.com>
	<7f03db650606190833l578ba5e5h77a1aacc73710453@mail.gmail.com>
Message-ID: <4497425B.2030802@boost-consulting.com>

Will Lee wrote:
> I have resolved this problem when I set the template NoProxy to true.  
> However, another problem comes up when I try to use the map in python:

That's been fixed sometime ago. To be sure, I added some tests
to map_indexing_suite.cpp/py in the test directory. Please check
out the CVS.

Cheers!
-- 
Joel de Guzman
http://www.boost-consulting.com
http://spirit.sf.net



From s_sourceforge at nedprod.com  Tue Jun 20 04:22:30 2006
From: s_sourceforge at nedprod.com (Niall Douglas)
Date: Tue, 20 Jun 2006 03:22:30 +0100
Subject: [C++-sig] Default arguments in pyplusplus
Message-ID: <mailman.98.1223937214.18936.cplusplus-sig@python.org>

On 19 Jun 2006 at 8:17, Roman Yakovenko wrote:

> Why don't you resolve the issue with enums. I see few ways to go:
> 
> To find places where enum used as default arguments and enum still not exported
> >From here you have 2 solutions: to remove default arguments or to change default
> arguments to have integer type with value of enum. This should work.

I tried this one using the following:

    # Patch default args with enums to be number (to avoid undeclared 
enum problem)
    allfuncs = mb.calldefs()
    for func in allfuncs:
        print type(func), type(func.parent), func
        for arg in func.arguments:
            # Is it an argument with a default value?
            if arg.default_value:
                # Is it an enumeration?
                if isinstance( arg.type, 
declarations.cpptypes.declarated_t ) and isinstance( 
arg.type.declaration, declarations.enumeration.enumeration_t ):
                    # Is it an embedded enum rather than global?
                    scopecount = arg.default_value.count('::') - 
arg.default_value.startswith('::')
                    if scopecount>1:
                        # Be lazy, and just lookup the last part
                        value = arg.default_value[ 
arg.default_value.rfind('::')+2: ]
                        arg.default_value = 
arg.type.declaration.values[ value ] + '/*' + arg.default_value + 
'*/'

And this works fine. EXCEPT that class wrappers also take the int 
converted default parameter, which leads unfortunately to code like 
this:

    QBlkSocket_wrapper(::FX::QBlkSocket::Type 
type=0/*::FX::QBlkSocket::Stream*/, ::FX::FXushort port=0 )
    : FX::QBlkSocket( type, port )
      , bp::wrapper< FX::QBlkSocket >()
    {   // Normal constructor
    
    }

Of course, this won't compile. Any ideas?

Thanks for the enum idea. If I can get keyword parameters working I'd 
be very happy.

Cheers,
Niall



From roman.yakovenko at gmail.com  Tue Jun 20 07:49:42 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Tue, 20 Jun 2006 08:49:42 +0300
Subject: [C++-sig] Default arguments in pyplusplus
In-Reply-To: <200606200226.k5K2QjwT025305@smtp-vbr6.xs4all.nl>
References: <200606200226.k5K2QjwT025305@smtp-vbr6.xs4all.nl>
Message-ID: <7465b6170606192249uf696ff3u4ecd0beeca76a946@mail.gmail.com>

On 6/20/06, Niall Douglas <s_sourceforge at nedprod.com> wrote:
# Is it an enumeration?
if isinstance( arg.type, declarations.cpptypes.declarated_t ) \
  and isinstance( arg.type.declaration,
declarations.enumeration.enumeration_t ):

You are working too hard here. pygccxml has functionality similar to
boost.type_traits
Read more here:
http://language-binding.net/pygccxml/design.html#type-traits

pygccxml.declarations.is_enum( arg.type ) will do the job

> Of course, this won't compile. Any ideas?

Obviously we need to supply different default arguments for wrapper
and for code that
actually exposes the function. The idea is simple we need new class variable on
argument_t class default_value_xxx ( I explain later why xxx, but you
will have to rename it
to something readable ). And then to generate code that in wrapper
uses original value, but
in boost.python code uses modified one, right?

I attached the solution to your problem. I made small refactoring + added new
functionality ( enum_declaration ) so you will need the latest version.

> Thanks for the enum idea. If I can get keyword parameters working I'd
> be very happy.

Hope I made you very, very happy :-).

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: default_value.py
Type: application/x-python
Size: 1396 bytes
Desc: not available
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20060620/2b88fc6d/attachment.bin>

From roman.yakovenko at gmail.com  Tue Jun 20 13:56:20 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Tue, 20 Jun 2006 14:56:20 +0300
Subject: [C++-sig] [ vector | list ] indexing suite
Message-ID: <7465b6170606200456w53eb66ecrad80540888e176@mail.gmail.com>

Hi. While working on adding support to pyplusplus for indexing suite,
I took a look on
vector_indexing_suite.

I think that with a small change of code it is possible to make
"vector_indexing_suite" class
to work with std::list. The change is simple: to replace operator[]
and iterator += ( Distance )
with std::advance( iterator, Distance ). I am almost sure that there
is no performance
penalties for this change.

What do you think? Am I missing some trivial reason why this change
could not be done?

I attach the modified file.

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: vector_indexing_suite.hpp
Type: application/octet-stream
Size: 8335 bytes
Desc: not available
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20060620/38138b61/attachment.obj>

From cyril.bazin at info.unicaen.fr  Tue Jun 20 18:11:10 2006
From: cyril.bazin at info.unicaen.fr (Cyril Bazin)
Date: Tue, 20 Jun 2006 18:11:10 +0200
Subject: [C++-sig] ArgumentError...
In-Reply-To: <uveqwomkk.fsf@boost-consulting.com>
References: <a10b8d620606190835w5fa04d3cg98789bb2cfef9b2d@mail.gmail.com>
	<uveqwomkk.fsf@boost-consulting.com>
Message-ID: <a10b8d620606200911h1405ce8dt4fac4ed22fd51085@mail.gmail.com>

Hello,

Thanks, It works!!!


>
> My best guess is that Delaunay is derived from
>
>   CGAL::Triangulation_2<
>        CGAL::Cartesian<double>
>      , CGAL::Tds2<
>            CGAL::Tvb<CGAL::Cartesian<double>, CGAL::Tdsvb<void> >
>          , CGAL::Tdsfb<void>
>        >
>   >
>
> which supplies finite_vertices_begin and finite_vertices_end.  Since
> you haven't wrapped that, and Boost.Python doesn't know it's a base
> class of Delaunay, you get the error.


Yes. You are right.

You have two choices:
>
> 1. pre-assign finite_vertices_end/begin to variables (or cast to
>    temporaries) of type
>
>       Finite_vertices_iterator (Delaunay::*)()
>
>    and pass those to range(...)


This method works fine. Thanks !!!

2. Use the first form of range<NextPolicies,Target>(...) described
>    here: http://boost.org/libs/python/doc/v2/iterator.html#range-spec


The mechanisms of NextPolicies and Terget are a little obscur to me...
Today, I don't understand everything about this kind of stuffs.

Just a question :
Is there a third method where I can define the iterator and member of the
class Triangulation2<...> and define the inheritance between Triangulation_2
and Delaunay using boost?
That would let the Python binding looks more like the C++ original
library...

HTH,


Happy Being Helped!!!

Cyril
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20060620/342a4bab/attachment.htm>

From s_sourceforge at nedprod.com  Tue Jun 20 21:59:51 2006
From: s_sourceforge at nedprod.com (Niall Douglas)
Date: Tue, 20 Jun 2006 20:59:51 +0100
Subject: [C++-sig] [ vector | list ] indexing suite
In-Reply-To: <7465b6170606200456w53eb66ecrad80540888e176@mail.gmail.com>
Message-ID: <449861C7.15757.1D9833D@s_sourceforge.nedprod.com>

On 20 Jun 2006 at 14:56, Roman Yakovenko wrote:

> Hi. While working on adding support to pyplusplus for indexing suite,
> I took a look on
> vector_indexing_suite.
> 
> I think that with a small change of code it is possible to make
> "vector_indexing_suite" class
> to work with std::list. The change is simple: to replace operator[]
> and iterator += ( Distance )
> with std::advance( iterator, Distance ). I am almost sure that there
> is no performance
> penalties for this change.
> 
> What do you think? Am I missing some trivial reason why this change
> could not be done?

You would be far better off using the new and improved indexing 
suite. It's still sitting in CVS on some experimental branch. It is a 
lot more efficient, already automatically interfaces with std::list 
and std::vector and also can be easily extended to handle any 
arbitrary kind of container.

TnFOX uses the new indexing suite. Saved me hours of extra work. 
Search the mailing list archives for Raoul Gough.

Cheers,
Niall





From dave at boost-consulting.com  Tue Jun 20 23:16:19 2006
From: dave at boost-consulting.com (David Abrahams)
Date: Tue, 20 Jun 2006 17:16:19 -0400
Subject: [C++-sig] ArgumentError...
References: <a10b8d620606190835w5fa04d3cg98789bb2cfef9b2d@mail.gmail.com>
	<uveqwomkk.fsf@boost-consulting.com>
	<a10b8d620606200911h1405ce8dt4fac4ed22fd51085@mail.gmail.com>
Message-ID: <u4pyffrz0.fsf@boost-consulting.com>

"Cyril Bazin" <cyril.bazin at info.unicaen.fr> writes:

> Just a question :
> Is there a third method where I can define the iterator and member of the
> class Triangulation2<...> and define the inheritance between Triangulation_2
> and Delaunay using boost?

Sure, but if Triangulation_2 is an implementation detail of CGAL, I
don't recommend it.

> That would let the Python binding looks more like the C++ original
> library...

It would let the code that _does_ the binding look more like the
original C++ library, but is that valuable?

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com



From roman.yakovenko at gmail.com  Tue Jun 20 23:44:54 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Wed, 21 Jun 2006 00:44:54 +0300
Subject: [C++-sig] [ vector | list ] indexing suite
In-Reply-To: <449861C7.15757.1D9833D@s_sourceforge.nedprod.com>
References: <7465b6170606200456w53eb66ecrad80540888e176@mail.gmail.com>
	<449861C7.15757.1D9833D@s_sourceforge.nedprod.com>
Message-ID: <7465b6170606201444i7bb6c415oe64fd3d12817985@mail.gmail.com>

On 6/20/06, Niall Douglas <s_sourceforge at nedprod.com> wrote:
> You would be far better off using the new and improved indexing
> suite. It's still sitting in CVS on some experimental branch. It is a
> lot more efficient, already automatically interfaces with std::list
> and std::vector and also can be easily extended to handle any
> arbitrary kind of container.

Thanks. I found it and it really looks interesting.

When this code will be moved to main branch?

Thanks

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From joel at boost-consulting.com  Wed Jun 21 02:28:07 2006
From: joel at boost-consulting.com (Joel de Guzman)
Date: Wed, 21 Jun 2006 08:28:07 +0800
Subject: [C++-sig] [ vector | list ] indexing suite
In-Reply-To: <449861C7.15757.1D9833D@s_sourceforge.nedprod.com>
References: <449861C7.15757.1D9833D@s_sourceforge.nedprod.com>
Message-ID: <44989297.6030105@boost-consulting.com>

Niall Douglas wrote:
> On 20 Jun 2006 at 14:56, Roman Yakovenko wrote:
> 
>> Hi. While working on adding support to pyplusplus for indexing suite,
>> I took a look on
>> vector_indexing_suite.
>>
>> I think that with a small change of code it is possible to make
>> "vector_indexing_suite" class
>> to work with std::list. The change is simple: to replace operator[]
>> and iterator += ( Distance )
>> with std::advance( iterator, Distance ). I am almost sure that there
>> is no performance
>> penalties for this change.
>>
>> What do you think? Am I missing some trivial reason why this change
>> could not be done?
> 
> You would be far better off using the new and improved indexing 
> suite. It's still sitting in CVS on some experimental branch. It is a 
> lot more efficient, already automatically interfaces with std::list 
> and std::vector and also can be easily extended to handle any 
> arbitrary kind of container.
> 
> TnFOX uses the new indexing suite. Saved me hours of extra work. 
> Search the mailing list archives for Raoul Gough.

I think this is definitely the way to go. It makes perfect sense
for Raoul's work to replace the current indexing suite. Last
time I heard from Raoul, he was having some problems with VC6
porting. I wonder if VC6 is still relevant for Boost.Python
now, Dave?

Regards,
-- 
Joel de Guzman
http://www.boost-consulting.com
http://spirit.sf.net


From s_sourceforge at nedprod.com  Wed Jun 21 03:37:59 2006
From: s_sourceforge at nedprod.com (Niall Douglas)
Date: Wed, 21 Jun 2006 02:37:59 +0100
Subject: [C++-sig] Default arguments in pyplusplus
In-Reply-To: <7465b6170606192249uf696ff3u4ecd0beeca76a946@mail.gmail.com>
References: <200606200226.k5K2QjwT025305@smtp-vbr6.xs4all.nl>
Message-ID: <4498B107.17031.30F156B@s_sourceforge.nedprod.com>

On 20 Jun 2006 at 8:49, Roman Yakovenko wrote:

> > Of course, this won't compile. Any ideas?
> 
> Obviously we need to supply different default arguments for wrapper
> and for code that
> actually exposes the function. The idea is simple we need new class variable on
> argument_t class default_value_xxx ( I explain later why xxx, but you
> will have to rename it
> to something readable ). And then to generate code that in wrapper
> uses original value, but
> in boost.python code uses modified one, right?
> 
> I attached the solution to your problem. I made small refactoring + added new
> functionality ( enum_declaration ) so you will need the latest version.

Cool, thanks for that. I've committed it to the TnFOX SVN repository.

> > Thanks for the enum idea. If I can get keyword parameters working I'd
> > be very happy.
> 
> Hope I made you very, very happy :-).

Well, nearly ...

Unfortunately, because I am now including all the patch code into 
every source file, I'm getting compiler limit: internal structure 
overflow error on some of the more complex files. So I came up with a 
cunning plan:

In file_writers/multiple_files.py create_source(), I added the 
following:

        answer = []
        if self.extmodule.license:
            answer.append( self.extmodule.license.create() )
            
        if self.extmodule.precompiled_header:
            answer.append( self.extmodule.precompiled_header.create() 
)
            
        answer.append( '#define PYPP_BUILDING_'+file_name.replace( 
'.', '_').upper() )
            
        answer.append( '#include "%s%s"' % ( file_name, 
self.HEADER_EXT ) )

        # Include all 'global' include files...

Now I have a common.h which is the precompiled header and then a 
patches.cpp.h which uses a series of #ifdef's to decide which patch 
code to include based on the macro defined above. Achieves what I 
wanted originally of inserting custom code just to where it's needed.

One other thing: I've just discovered you can't do a #pragma hdrstop 
if it's enclosed by an #if. So that code of:

        answer.append( '#ifdef _MSC_VER' )
        answer.append( self.indent( '#pragma hdrstop' ) )
        answer.append( '#endif //_MSC_VER' )

... I originally suggested is useless. Sorry about that. If you just 
plain remove it, you can specify on the command where to stop 
precompiled header inclusion.

Cheers,
Niall





From s_sourceforge at nedprod.com  Wed Jun 21 04:17:37 2006
From: s_sourceforge at nedprod.com (Niall Douglas)
Date: Wed, 21 Jun 2006 03:17:37 +0100
Subject: [C++-sig] bp::arg() also requires types to be declared
Message-ID: <4498BA51.13536.3335777@s_sourceforge.nedprod.com>

It appears it's not only enums that must be defined before use with 
boost::python::arg() - also all types. This causes a fault here:

        typedef bp::class_< TnFXAppEventLoop_wrapper, bp::bases< 
FX::FXObject >, std::auto_ptr<TnFXAppEventLoop_wrapper>, 
boost::noncopyable > TnFXAppEventLoop_exposer_t;

        TnFXAppEventLoop_exposer_t TnFXAppEventLoop_exposer = 
TnFXAppEventLoop_exposer_t( "TnFXAppEventLoop", bp::init< 
bp::optional< char const *, FX::TnFXApp * > >(( 
bp::arg("threadname")="Event Loop thread", 
bp::arg("app")=FX::TnFXApp::instance( ) )) );

Here TnFXApp has not been declared to BPL before TnFXAppEventLoop. 
Problem is that TnFXApp needs TnFXAppEventLoop defined before it (in 
C++).

We have a problem therefore. I can only see one should set 
use_keywords to False on every calldef taking a typed default 
parameter. Unfortunately, this is most of them and thus renders using 
bp::arg() useless.

Ideas? How hard would it be to patch BPL to let this work usefully?

Cheers,
Niall





From roman.yakovenko at gmail.com  Wed Jun 21 07:01:32 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Wed, 21 Jun 2006 08:01:32 +0300
Subject: [C++-sig] Default arguments in pyplusplus
In-Reply-To: <4498B107.17031.30F156B@s_sourceforge.nedprod.com>
References: <200606200226.k5K2QjwT025305@smtp-vbr6.xs4all.nl>
	<7465b6170606192249uf696ff3u4ecd0beeca76a946@mail.gmail.com>
	<4498B107.17031.30F156B@s_sourceforge.nedprod.com>
Message-ID: <7465b6170606202201n221cc2f7r5563ae5a982a5bde@mail.gmail.com>

On 6/21/06, Niall Douglas <s_sourceforge at nedprod.com> wrote:
> Unfortunately, because I am now including all the patch code into
> every source file, I'm getting compiler limit: internal structure
> overflow error on some of the more complex files. So I came up with a
> cunning plan:
>
It is so complex, the is a better solution.

1. add new property to class_t - additional_file_text( code ,.... give
a better name )
    for cls in image classes:
        cls.additional_file_text = ...
2. derive from multiple_files_t and redefine split_class method.
3. pass the instance of new class to split module.

The only code that should be changed it module_builder_t.split_module. I need to
add new parameter - file_writer, that by default is None.

What do you think?

> One other thing: I've just discovered you can't do a #pragma hdrstop
> if it's enclosed by an #if. So that code of:
>
>         answer.append( '#ifdef _MSC_VER' )
>         answer.append( self.indent( '#pragma hdrstop' ) )
>         answer.append( '#endif //_MSC_VER' )
>
> ... I originally suggested is useless. Sorry about that. If you just
> plain remove it, you can specify on the command where to stop
> precompiled header inclusion.

Okay.

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From roman.yakovenko at gmail.com  Wed Jun 21 07:06:42 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Wed, 21 Jun 2006 08:06:42 +0300
Subject: [C++-sig] bp::arg() also requires types to be declared
In-Reply-To: <4498BA51.13536.3335777@s_sourceforge.nedprod.com>
References: <4498BA51.13536.3335777@s_sourceforge.nedprod.com>
Message-ID: <7465b6170606202206w3a4f9244jccbe34e8eb447476@mail.gmail.com>

On 6/21/06, Niall Douglas <s_sourceforge at nedprod.com> wrote:
> It appears it's not only enums that must be defined before use with
> boost::python::arg() - also all types. This causes a fault here:
>
>         typedef bp::class_< TnFXAppEventLoop_wrapper, bp::bases<
> FX::FXObject >, std::auto_ptr<TnFXAppEventLoop_wrapper>,
> boost::noncopyable > TnFXAppEventLoop_exposer_t;
>
>         TnFXAppEventLoop_exposer_t TnFXAppEventLoop_exposer =
> TnFXAppEventLoop_exposer_t( "TnFXAppEventLoop", bp::init<
> bp::optional< char const *, FX::TnFXApp * > >((
> bp::arg("threadname")="Event Loop thread",
> bp::arg("app")=FX::TnFXApp::instance( ) )) );
>
> Here TnFXApp has not been declared to BPL before TnFXAppEventLoop.
> Problem is that TnFXApp needs TnFXAppEventLoop defined before it (in
> C++).
>
> We have a problem therefore. I can only see one should set
> use_keywords to False on every calldef taking a typed default
> parameter. Unfortunately, this is most of them and thus renders using
> bp::arg() useless.
>
> Ideas? How hard would it be to patch BPL to let this work usefully?

Niall, I have an idea, but unfortunately I don't have time to check
it. Please take a
look on pyplusplus/module_creator/class_organizer.py file. This file contains
algorithm - classic topological sort. It is used to define the order of exported
classes. I am sure you can improve it, so we will not have this problem.

Good luck!

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From dave at boost-consulting.com  Wed Jun 21 07:02:33 2006
From: dave at boost-consulting.com (David Abrahams)
Date: Wed, 21 Jun 2006 01:02:33 -0400
Subject: [C++-sig] [ vector | list ] indexing suite
References: <449861C7.15757.1D9833D@s_sourceforge.nedprod.com>
	<44989297.6030105@boost-consulting.com>
Message-ID: <upsh3drti.fsf@boost-consulting.com>

Joel de Guzman <joel at boost-consulting.com> writes:

>> TnFOX uses the new indexing suite. Saved me hours of extra work. 
>> Search the mailing list archives for Raoul Gough.
>
> I think this is definitely the way to go. It makes perfect sense
> for Raoul's work to replace the current indexing suite. Last
> time I heard from Raoul, he was having some problems with VC6
> porting. I wonder if VC6 is still relevant for Boost.Python
> now, Dave?

Having trouble forming an opinion.  Unfortunately there are serious
embedded platforms that are still stuck with vc6.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com



From dave at boost-consulting.com  Wed Jun 21 07:04:07 2006
From: dave at boost-consulting.com (David Abrahams)
Date: Wed, 21 Jun 2006 01:04:07 -0400
Subject: [C++-sig] bp::arg() also requires types to be declared
References: <4498BA51.13536.3335777@s_sourceforge.nedprod.com>
Message-ID: <uk67bdrqw.fsf@boost-consulting.com>

"Niall Douglas" <s_sourceforge at nedprod.com> writes:

> It appears it's not only enums that must be defined before use with 
> boost::python::arg() - also all types. 

What do you mean by "used with?"  

What do you mean by "defined?"  Do you mean they need to be complete
types (i.e. not forward declared)?

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com



From roman.yakovenko at gmail.com  Wed Jun 21 07:32:56 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Wed, 21 Jun 2006 08:32:56 +0300
Subject: [C++-sig] [ vector | list ] indexing suite
In-Reply-To: <upsh3drti.fsf@boost-consulting.com>
References: <449861C7.15757.1D9833D@s_sourceforge.nedprod.com>
	<44989297.6030105@boost-consulting.com>
	<upsh3drti.fsf@boost-consulting.com>
Message-ID: <7465b6170606202232u19e89e3n8422cfb74aa9a538@mail.gmail.com>

On 6/21/06, David Abrahams <dave at boost-consulting.com> wrote:
> Joel de Guzman <joel at boost-consulting.com> writes:
>
> >> TnFOX uses the new indexing suite. Saved me hours of extra work.
> >> Search the mailing list archives for Raoul Gough.
> >
> > I think this is definitely the way to go. It makes perfect sense
> > for Raoul's work to replace the current indexing suite. Last
> > time I heard from Raoul, he was having some problems with VC6
> > porting. I wonder if VC6 is still relevant for Boost.Python
> > now, Dave?
>
> Having trouble forming an opinion.  Unfortunately there are serious
> embedded platforms that are still stuck with vc6.

In this case, is it possible to apply the patch?

Another question: can both indexing suite co-exists?
if so may be it possible to move indexing_v2 to main branch?
The main objective to this is easiest install procedure and earlier testing.

Thank you.

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From roman.yakovenko at gmail.com  Wed Jun 21 07:35:44 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Wed, 21 Jun 2006 08:35:44 +0300
Subject: [C++-sig] bp::arg() also requires types to be declared
In-Reply-To: <uk67bdrqw.fsf@boost-consulting.com>
References: <4498BA51.13536.3335777@s_sourceforge.nedprod.com>
	<uk67bdrqw.fsf@boost-consulting.com>
Message-ID: <7465b6170606202235r159d3e7dy1f75c3afd1ddf8f4@mail.gmail.com>

On 6/21/06, David Abrahams <dave at boost-consulting.com> wrote:
> "Niall Douglas" <s_sourceforge at nedprod.com> writes:
>
> > It appears it's not only enums that must be defined before use with
> > boost::python::arg() - also all types.
>
> What do you mean by "used with?"
>
> What do you mean by "defined?"  Do you mean they need to be complete
> types (i.e. not forward declared)?

In order to use boost::python::arg with default value, the type of
default value
should be already registered.

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From d.bollmann at tu-berlin.de  Wed Jun 21 07:52:23 2006
From: d.bollmann at tu-berlin.de (Dietrich Bollmann)
Date: Wed, 21 Jun 2006 14:52:23 +0900
Subject: [C++-sig] C integration question - how to get the result of an
 evaluation with PyEval_EvalCode()
Message-ID: <4498DE97.70500@tu-berlin.de>

Hi, I am working on a "python command port" for an application (blender) 
written in C and using python as scripting language. The idea is to let 
other applications interact with blender via a socket using blender 
python scripts / commands.

I got everything work - but couldn't figure out how to get the result of 
the evaluation for sending it back over the socket to the client.

I am using the following commands:

   ...
   /* compile python script / command */
   script_compiled = Py_CompileString(script, "<bcp>", Py_file_input);
   ...
   /* evaluate compiled python script / command */
   PyObject *result_obj;
   result_obj = PyEval_EvalCode(script_compiled, dict, dict);
   ...

By "result" I am thinking about something like the messages printed by 
the python shell as response when executing a command - if possible as 
string. There are some differences to the python interactive loop though:
- `script' might be one command string like `a = 123\n' or a whole 
script like `def blah(a, b):\n  c = a + b\n  print c\n\nblah(3, 4)\n\n'
- there might be empty lines inside of function or class definitions as 
the script is entered as one C string and not line by line
- I am only interested in the result of the last command (like `7' in 
the case of the second example)

A second wish would be to have something like a `get_value()' function 
which would return the value of some variable as result - like just 
entering the name of a variable in the python shell or using a print 
command like `print variable' from the shell. (In general I would like 
to let `print' print to the blender stdout though.)

Any idea?

Thanks for your help & best wishes, Dietrich


By the way, when experimenting with "results" I also tried the following 
snipped (an adapted version from the book "Programming Python" by Mark 
Lutz):

void experiment() {
   int cval;
   PyObject *pdict, *pval;

   /* make a new namespace */
   pdict = PyDict_New();
   PyDict_SetItemString(pdict, "__builtins__", PyEval_GetBuiltins());

   PyDict_SetItemString(pdict, "Y", PyInt_FromLong(2));   /* dict['Y'] = 
2   */
   PyRun_String("X = 99",  Py_file_input, pdict, pdict);  /* run 
statements  */
   PyRun_String("X = X+Y", Py_file_input, pdict, pdict);  /* same X and 
Y    */
   pval = PyDict_GetItemString(pdict, "X");               /* fetch 
dict['X'] */

   PyArg_Parse(pval, "i", &cval);                         /* convert to C */
   printf("result: %d\n", cval);                          /* result=101 */
   Py_DECREF(pdict);
}

But for some reason got the following when executing it:

   result: -1074019336

What have I done wrong?

--
Dietrich Bollmann


From rwgk at yahoo.com  Wed Jun 21 17:20:26 2006
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Wed, 21 Jun 2006 08:20:26 -0700 (PDT)
Subject: [C++-sig] bp::arg() also requires types to be declared
In-Reply-To: <7465b6170606202235r159d3e7dy1f75c3afd1ddf8f4@mail.gmail.com>
Message-ID: <20060621152026.8835.qmail@web31501.mail.mud.yahoo.com>

--- Roman Yakovenko <roman.yakovenko at gmail.com> wrote:

> On 6/21/06, David Abrahams <dave at boost-consulting.com> wrote:
> > "Niall Douglas" <s_sourceforge at nedprod.com> writes:
> >
> > > It appears it's not only enums that must be defined before use with
> > > boost::python::arg() - also all types.
> >
> > What do you mean by "used with?"
> >
> > What do you mean by "defined?"  Do you mean they need to be complete
> > types (i.e. not forward declared)?
> 
> In order to use boost::python::arg with default value, the type of
> default value
> should be already registered.

You can get around this by using BOOST_PYTHON_[MEMBER_]FUNCTION_OVERLOADS, or
simply wrap the overloads one by one.


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 


From s_sourceforge at nedprod.com  Wed Jun 21 18:01:22 2006
From: s_sourceforge at nedprod.com (Niall Douglas)
Date: Wed, 21 Jun 2006 17:01:22 +0100
Subject: [C++-sig] bp::arg() also requires types to be declared
In-Reply-To: <7465b6170606202206w3a4f9244jccbe34e8eb447476@mail.gmail.com>
References: <4498BA51.13536.3335777@s_sourceforge.nedprod.com>
Message-ID: <44997B62.6825.E7172@s_sourceforge.nedprod.com>

On 21 Jun 2006 at 8:06, Roman Yakovenko wrote:

> > We have a problem therefore. I can only see one should set
> > use_keywords to False on every calldef taking a typed default
> > parameter. Unfortunately, this is most of them and thus renders using
> > bp::arg() useless.
> >
> > Ideas? How hard would it be to patch BPL to let this work usefully?
> 
> Niall, I have an idea, but unfortunately I don't have time to check
> it. Please take a
> look on pyplusplus/module_creator/class_organizer.py file. This file contains
> algorithm - classic topological sort. It is used to define the order of exported
> classes. I am sure you can improve it, so we will not have this problem.

Herein lies a problem:

class Foo;

class Foo2
{
   friend class Foo;
   Foo2(Foo *f=0);
};

class Foo
{
   Foo(Foo2 *f=0);
   inline void somemethod(Foo2 *a) { a->foo(); }
};

Here you need to declare Foo and Foo2 to BPL. Problem is, they take 
constructor args circularly so if I try to declare default args, 
either one or the other is going to be undeclared to BPL and you get 
the fault.

If, and only if, both Foo and Foo2 have constructors with no default 
args, you can delay the constructor declaration and declare it later 
via bp::init<>. Otherwise you are trapped and can't use bp::arg() at 
all.

To fix pyplusplus to correctly order stuff according to all these 
possibilities, and even then still leaving open the potential of it 
failing to work in certain rare circumstances - well, the correct 
thing to do IMHO is to alter the implementation of bp::arg() so it 
doesn't need to have the type declared to BPL via bp::class_() first.

All this said, I am now out of time for the python bindings. I am 
going to have to cease development once again for the next few 
months. Sorry, I get windows and these windows get exhausted. I must 
now begin on a new chaotic economic modelling software I want to base 
my PhD on.

Cheers,
Niall





From s_sourceforge at nedprod.com  Wed Jun 21 18:01:22 2006
From: s_sourceforge at nedprod.com (Niall Douglas)
Date: Wed, 21 Jun 2006 17:01:22 +0100
Subject: [C++-sig] Default arguments in pyplusplus
In-Reply-To: <7465b6170606202201n221cc2f7r5563ae5a982a5bde@mail.gmail.com>
References: <4498B107.17031.30F156B@s_sourceforge.nedprod.com>
Message-ID: <44997B62.5660.E7115@s_sourceforge.nedprod.com>

On 21 Jun 2006 at 8:01, Roman Yakovenko wrote:

> On 6/21/06, Niall Douglas <s_sourceforge at nedprod.com> wrote:
> > Unfortunately, because I am now including all the patch code into
> > every source file, I'm getting compiler limit: internal structure
> > overflow error on some of the more complex files. So I came up with a
> > cunning plan:
> >
> It is so complex, the is a better solution.
> 
> 1. add new property to class_t - additional_file_text( code ,.... give
> a better name )
>     for cls in image classes:
>         cls.additional_file_text = ...
> 2. derive from multiple_files_t and redefine split_class method.
> 3. pass the instance of new class to split module.
> 
> The only code that should be changed it module_builder_t.split_module. I need to
> add new parameter - file_writer, that by default is None.
> 
> What do you think?

That looks fine too. Cleaner.

Cheers,
Niall





From lee.will at gmail.com  Wed Jun 21 19:59:28 2006
From: lee.will at gmail.com (Will Lee)
Date: Wed, 21 Jun 2006 12:59:28 -0500
Subject: [C++-sig]  Overriding operator == and = for None
Message-ID: <7f03db650606211059h1b424d20j413fc5f347d43eea@mail.gmail.com>

Is there a way I can override the operator == and = for a wrapped c++ object
to take in a Python None type?  Essentially, I'd like to define a class
where it can do:

#NullableInt is a struct that holds an int and a boolean, indicating whether
it is valid or not
c = NullableInt(3)
# NullableInt wraps a C++ std::vector<NullableInt>
cv = NullableIntVec()
cv.append(c) # ok
cv[0] = None
print cv[0]
# should print out "None", and the C++'s struct would hold a boolean that's
set to false

# Generates an invalid Int
cInvalid = NullableInt()
# Should be true, it won't remove object but setting the underlying c++
object to None)
cv[0] == cInvalid
# Should also be true if we can override ==
cv[0] == None

Does anybody know if it is posslbe in Boost.Python to have this behavior?

Will
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20060621/0ee6074e/attachment.htm>

From roman.yakovenko at gmail.com  Wed Jun 21 20:12:58 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Wed, 21 Jun 2006 21:12:58 +0300
Subject: [C++-sig] Overriding operator == and = for None
In-Reply-To: <7f03db650606211059h1b424d20j413fc5f347d43eea@mail.gmail.com>
References: <7f03db650606211059h1b424d20j413fc5f347d43eea@mail.gmail.com>
Message-ID: <7465b6170606211112y4ea67d68p1d9c2929b4c0f849@mail.gmail.com>

On 6/21/06, Will Lee <lee.will at gmail.com> wrote:
> Is there a way I can override the operator == and = for a wrapped c++ object
> to take in a Python None type?

I could be wrong, but you can not change semantic of = ( assignment operator )

As for the ==( equal operator ), you can overload == that takes
boost::python::object and then test it for being None.

I am almost sure that what you are trying to do should be rethought.
For example:

x  == None will work, but x is None will not. This can lead to bugs.

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From mail at a-beyer.de  Wed Jun 21 11:22:08 2006
From: mail at a-beyer.de (Andreas Beyer)
Date: Wed, 21 Jun 2006 11:22:08 +0200
Subject: [C++-sig] Overriding operator == and = for None
In-Reply-To: <7f03db650606211059h1b424d20j413fc5f347d43eea@mail.gmail.com>
References: <7f03db650606211059h1b424d20j413fc5f347d43eea@mail.gmail.com>
Message-ID: <44990FC0.9040408@a-beyer.de>

Hi,

I think there are two things to do.
First, NullableIntVec has to define a wrapper method for __setattr_(), 
which would have to accept python objects and either create a new empty 
NullableInt or a NullableInt initialized with the respective value.

Second, you will have to wrap __eq__(), __ne__() and/or __cmp__() of 
NullableInt in order to get the desired behaviour for ==.
I suggest implementing 'rich comparison', because that would allow you 
to do things like this:

 >>> c = NullableInt(3)
 >>> c < 4
 >>> True

See also here:
http://docs.python.org/ref/customization.html
http://www.boost.org/libs/python/doc/tutorial/doc/html/python/exposing.html#python.class_operators_special_functions


Again, these wrappers will have to accept python objects (i.e. 
boost::python::object) as arguments, otherwise you can't properly 
compare against None.

Hope that helps,
   Andreas

Will Lee wrote:
> Is there a way I can override the operator == and = for a wrapped c++ 
> object to take in a Python None type?  Essentially, I'd like to define 
> a class where it can do:
>
> #NullableInt is a struct that holds an int and a boolean, indicating 
> whether it is valid or not
> c = NullableInt(3)
> # NullableInt wraps a C++ std::vector<NullableInt>
> cv = NullableIntVec()
> cv.append(c) # ok
> cv[0] = None
> print cv[0]
> # should print out "None", and the C++'s struct would hold a boolean 
> that's set to false
>
> # Generates an invalid Int
> cInvalid = NullableInt()
> # Should be true, it won't remove object but setting the underlying 
> c++ object to None)
> cv[0] == cInvalid
> # Should also be true if we can override ==
> cv[0] == None
>
> Does anybody know if it is posslbe in Boost.Python to have this behavior?
>
> Will
> ------------------------------------------------------------------------
>
> _______________________________________________
> C++-sig mailing list
> C++-sig at python.org
> http://mail.python.org/mailman/listinfo/c++-sig
>   


-- 
Dr. Andreas Beyer
mail at a-beyer.de
www.andreas-beyer.de



From rwgk at yahoo.com  Wed Jun 21 22:49:32 2006
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Wed, 21 Jun 2006 13:49:32 -0700 (PDT)
Subject: [C++-sig] Overriding operator == and = for None
In-Reply-To: <7f03db650606211059h1b424d20j413fc5f347d43eea@mail.gmail.com>
Message-ID: <20060621204932.34184.qmail@web31508.mail.mud.yahoo.com>

I think you can do what you want based on boost::optional. You may want to look
here for an example:

  http://phenix-online.org/cctbx_sources/boost_adaptbx/
    optional_conversions.h
    optional_ext.cpp
    tst_optional.py

Note that "x is None" works correctly, not just "x == None".

The boost::optional documentation is here:

  http://www.boost.org/libs/optional/doc/optional.html

Feel free to copy optional_conversions.h if you are interested.

--- Will Lee <lee.will at gmail.com> wrote:

> Is there a way I can override the operator == and = for a wrapped c++ object
> to take in a Python None type?  Essentially, I'd like to define a class
> where it can do:
> 
> #NullableInt is a struct that holds an int and a boolean, indicating whether
> it is valid or not
> c = NullableInt(3)
> # NullableInt wraps a C++ std::vector<NullableInt>
> cv = NullableIntVec()
> cv.append(c) # ok
> cv[0] = None
> print cv[0]
> # should print out "None", and the C++'s struct would hold a boolean that's
> set to false
> 
> # Generates an invalid Int
> cInvalid = NullableInt()
> # Should be true, it won't remove object but setting the underlying c++
> object to None)
> cv[0] == cInvalid
> # Should also be true if we can override ==
> cv[0] == None
> 
> Does anybody know if it is posslbe in Boost.Python to have this behavior?
> 
> Will


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 


From abierbaum at gmail.com  Wed Jun 21 23:53:36 2006
From: abierbaum at gmail.com (Allen Bierbaum)
Date: Wed, 21 Jun 2006 16:53:36 -0500
Subject: [C++-sig] Protected destructor compile error
Message-ID: <bea9869d0606211453p191d307o2dbeb9175affdc12@mail.gmail.com>

I have run into a compiler error with boost.python and I would like to
understand what is causing it so I can try to fix the problem.  This
is coming up in several places in the binding I am working on, so I am
hopeful if I can understand what is happening with one case I can fix
it everywhere.

Background:

I am wrapping a class, Camera, the method that is causing problems is:

bool Camera::calcViewRay (Line &line, Int32 x, Int32 y, const Viewport &port);

I am wrapping it (using pyplusplus) with this def:

Camera_exposer.def("calcViewRay"
    , &::osg::Camera::calcViewRay
    , ( bp::arg("line"), bp::arg("x"), bp::arg("y"), bp::arg("port") )
    , bp::default_call_policies() );

I get this error when building with boost 1.33.1 on fedora core 4 (gcc 4.0.2):

/home/allenb/Source/boost/installed/include/boost-1_33_1/boost/python/detail/destroy.hpp:
In static member function 'static void
boost::python::detail::value_destroyer<false>::execute(const volatile
T*) [with T = osg::Viewport]':
/home/allenb/Source/boost/installed/include/boost-1_33_1/boost/python/detail/destroy.hpp:90:
  instantiated from 'void
boost::python::detail::destroy_referent_impl(void*, T& (*)()) [with T
= const osg::Viewport]'
/home/allenb/Source/boost/installed/include/boost-1_33_1/boost/python/detail/destroy.hpp:101:
  instantiated from 'void
boost::python::detail::destroy_referent(void*, T (*)()) [with T =
const osg::Viewport&]'
/home/allenb/Source/boost/installed/include/boost-1_33_1/boost/python/converter/rvalue_from_python_data.hpp:135:
  instantiated from
'boost::python::converter::rvalue_from_python_data<T>::~rvalue_from_python_data()
[with T = const osg::Viewport&]'
/home/allenb/Source/boost/installed/include/boost-1_33_1/boost/python/converter/arg_from_python.hpp:108:
  instantiated from 'PyObject*
boost::python::detail::caller_arity<5u>::impl<F, Policies,
Sig>::operator()(PyObject*, PyObject*) [with F = bool
(osg::Camera::*)(osg::Line&, osg::Int32, osg::Int32, const
osg::Viewport&), Policies = boost::python::default_call_policies, Sig
= boost::mpl::vector6<bool, osg::Camera&, osg::Line&, int, int, const
osg::Viewport&>]'
/home/allenb/Source/boost/installed/include/boost-1_33_1/boost/python/object/py_function.hpp:38:
  instantiated from 'PyObject*
boost::python::objects::caller_py_function_impl<Caller>::operator()(PyObject*,
PyObject*) [with Caller = boost::python::detail::caller<bool
(osg::Camera::*)(osg::Line&, osg::Int32, osg::Int32, const
osg::Viewport&), boost::python::default_call_policies,
boost::mpl::vector6<bool, osg::Camera&, osg::Line&, int, int, const
osg::Viewport&> >]'
src/osg_module2/Camera.pypp.cpp:459:   instantiated from here
/home/allenb/Source/OpenSG/Builds/i686-pc-linux-gnu-g++/include/OpenSG/OSGViewport.h:126:
error: 'virtual osg::Viewport::~Viewport()' is protected
/home/allenb/Source/boost/installed/include/boost-1_33_1/boost/python/detail/destroy.hpp:33:
error: within this context

It looks like there is a problem with the "const Viewport& port"
argument being passed.  For some reason the generated code seems to
want to be able to destroy instances of the Viewport class.  This
class has protected constructor and destructor as it is reference
counted and the user can only create them using a helper function
(Camera::create()).

Can anyone describe what is happening here and what prerequisites
exist for this call to work?  Is there some call policy that I should
be using to make this work?

Thanks,
Allen


From diresu at web.de  Thu Jun 22 05:37:44 2006
From: diresu at web.de (Dietrich Bollmann)
Date: Thu, 22 Jun 2006 12:37:44 +0900
Subject: [C++-sig] C integration question - how to get the result of an
 evaluation with PyEval_EvalCode()
Message-ID: <449A1088.3060703@web.de>

Hi, I am working on a "python command port" for an application (blender)
written in C and using python as scripting language. The idea is to let
other applications interact with blender via a socket using blender
python scripts / commands.

I got everything work - but couldn't figure out how to get the result of
the evaluation for sending it back over the socket to the client.

I am using the following commands:

   ...
   /* compile python script / command */
   script_compiled = Py_CompileString(script, "<bcp>", Py_file_input);
   ...
   /* evaluate compiled python script / command */
   PyObject *result_obj;
   result_obj = PyEval_EvalCode(script_compiled, dict, dict);
   ...

By "result" I am thinking about something like the messages printed by
the python shell as response when executing a command - if possible as
string. There are some differences to the python interactive loop though:
- `script' might be one command string like `a = 123\n' or a whole
script like `def blah(a, b):\n  c = a + b\n  print c\n\nblah(3, 4)\n\n'
- there might be empty lines inside of function or class definitions as
the script is entered as one C string and not line by line
- I am only interested in the result of the last command (like `7' in
the case of the second example)

A second wish would be to have something like a `get_value()' function
which would return the value of some variable as result - like just
entering the name of a variable in the python shell or using a print
command like `print variable' from the shell. (In general I would like
to let `print' print to the blender stdout though.)

Any idea?

Thanks for your help & best wishes, Dietrich


By the way, when experimenting with "results" I also tried the following
snipped (an adapted version from the book "Programming Python" by Mark
Lutz):

void experiment() {
   int cval;
   PyObject *pdict, *pval;

   /* make a new namespace */
   pdict = PyDict_New();
   PyDict_SetItemString(pdict, "__builtins__", PyEval_GetBuiltins());

   PyDict_SetItemString(pdict, "Y", PyInt_FromLong(2));   /* dict['Y'] =
2   */
   PyRun_String("X = 99",  Py_file_input, pdict, pdict);  /* run
statements  */
   PyRun_String("X = X+Y", Py_file_input, pdict, pdict);  /* same X and
Y    */
   pval = PyDict_GetItemString(pdict, "X");               /* fetch
dict['X'] */

   PyArg_Parse(pval, "i", &cval);                         /* convert to C */
   printf("result: %d\n", cval);                          /* result=101 */
   Py_DECREF(pdict);
}

But for some reason got the following when executing it:

   result: -1074019336

What have I done wrong?

--
Dietrich Bollmann




From cyril.bazin at info.unicaen.fr  Thu Jun 22 10:46:55 2006
From: cyril.bazin at info.unicaen.fr (Cyril Bazin)
Date: Thu, 22 Jun 2006 10:46:55 +0200
Subject: [C++-sig] Iterator or not iterator...
Message-ID: <a10b8d620606220146v18ccc82dm83012af349d5cbec@mail.gmail.com>

Hello,

I've got a question about the bindings of iterators.

Given a C++ code which look more or less like that:

template <class O>
class C {
    It<O> the_begin() {...}

    It<O> the_end() {...}

    It2<O> foo(it<O>) {...}

}

template <class O>
class It {
   //... define everything that an iterator must define ...
}

The C++ code to bind a version of the previous class C could be :

class MyO {
   //define some methods
}

typedef C<MyO>      MyC;
typedef It<MyO>       MyIt;

class_<MyO>("MyO")
        .def(...)
        ...;

//class _<MyIt>("MyIt")
//       .def(...)
//       ...;

class_<MyC>("MyC")
        .def("the_iter", range(&MyC::the_begin , &MyC::the_end))
        .def("foo", MyC::foo)

As expected, in Python the method MyC.the_iter will return some instances of
the class MyO.
But, the method MyC::foo expect a MyIt as argument... So, I can't do
something like that :

myC = MyC()
for bar in myC:
    assert type(bar) == MyO
    x = myC.foo(bar)
    ...

In other words, Is there a way to define MyC like that :
class_<MyC>("MyC")
        .def("the_iter", range_bis(&MyC::the_begin , &MyC::the_end))

Which could allow me to do something like:
myC = MyC()
for bar in myC:
    assert type(bar) == MyIt
    x = myC.foo(bar)
    ...

If you have any suggestion...

Thanks in advance.

Cyril
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20060622/99f415e6/attachment.htm>

From j.reid at mail.cryst.bbk.ac.uk  Thu Jun 22 12:39:15 2006
From: j.reid at mail.cryst.bbk.ac.uk (John Reid)
Date: Thu, 22 Jun 2006 11:39:15 +0100
Subject: [C++-sig] boost::python std::vector< std::string > crash
Message-ID: <e7ds1k$bmu$1@sea.gmane.org>

Hi,

I'm using boost/python/suite/indexing/vector_indexing_suite.hpp from the 
cvs_head version of boost python on MSVC 8. I'm getting a runtime 
assertion failure when appending to a vector of strings.



I define a module with a vector of strings with the following code:

#include <boost/python.hpp>
#include <boost/python/suite/indexing/vector_indexing_suite.hpp>

#include <vector>

using namespace boost;
using namespace boost::python;
using namespace std;

typedef vector< string > string_vec;

BOOST_PYTHON_MODULE( _bptest )
{
	/**
	A vector of strings.
	*/
	class_< string_vec >( "StringVec" )
		.def( vector_indexing_suite< string_vec >() )
		;
}



I use this python code:

import sys

print 'Importing bptest'
_path_to_module = 
'../bin/Bio/Biopsy/_bptest.pyd/vc-8_0/debug/threading-multi'
sys.path.append( _path_to_module )
import _bptest

print 'Creating string vec'
strings = _bptest.StringVec()

print 'Appending string'
strings.append( 'a' )




and it crashes in "strings.append( 'a' )"



There seems to be very little in the documenation about how std::strings 
are handled. They seem to work fine when used outside containers. Also 
containers of user-defined classes (as in the vector indexing example, 
struct X) and containers of doubles, ints, etc.. seem to work fine.

What am I missing here?



Eventually I want to use a C++ method that accepts a
const std::vector< std::string > &
as an argument. The FAQ it suggests there are two methods to do this: 
vector_indexing and also using custom rvalue converters. Unfortunately 
this documentation seems rather out of date. The instructions to log in 
to cvs to see the example code don't appear to work. I did manage to 
find the code but it is not clear how it could help me.


Any help would be much appreciated.

Thanks,
John.



P.S. here is the stack trace:
 > 
_biopsy.pyd!boost::python::detail::value_destroyer<0>::execute<std::basic_string<char,std::char_traits<char>,std::allocator<char> 
 > >(const 
std::basic_string<char,std::char_traits<char>,std::allocator<char> > * 
p="")  Line 34	C++
 
_biopsy.pyd!boost::python::detail::destroy_referent_impl<std::basic_string<char,std::char_traits<char>,std::allocator<char> 
 > const >(void * p=0x0021fbc8, const 
std::basic_string<char,std::char_traits<char>,std::allocator<char> > & 
(void)* __formal=0x00000000)  Line 95 + 0x9 bytes	C++
 
_biopsy.pyd!boost::python::detail::destroy_referent<std::basic_string<char,std::char_traits<char>,std::allocator<char> 
 > const &>(void * p=0x0021fbc8, const 
std::basic_string<char,std::char_traits<char>,std::allocator<char> > & 
(void)* __formal=0x00000000)  Line 101 + 0xb bytes	C++
 
_biopsy.pyd!boost::python::converter::rvalue_from_python_data<std::basic_string<char,std::char_traits<char>,std::allocator<char> 
 > const 
&>::~rvalue_from_python_data<std::basic_string<char,std::char_traits<char>,std::allocator<char> 
 > const &>()  Line 135 + 0xe bytes	C++
 
_biopsy.pyd!boost::python::converter::arg_rvalue_from_python<std::basic_string<char,std::char_traits<char>,std::allocator<char> 
 > const 
&>::~arg_rvalue_from_python<std::basic_string<char,std::char_traits<char>,std::allocator<char> 
 > const &>()  + 0xf bytes	C++
 
_biopsy.pyd!boost::python::arg_from_python<std::basic_string<char,std::char_traits<char>,std::allocator<char> 
 > const 
&>::~arg_from_python<std::basic_string<char,std::char_traits<char>,std::allocator<char> 
 > const &>()  + 0xf bytes	C++
  	_biopsy.pyd!boost::python::detail::caller_arity<2>::impl<void 
(__cdecl*)(std::vector<std::basic_string<char,std::char_traits<char>,std::allocator<char> 
 >,std::allocator<std::basic_string<char,std::char_traits<char>,std::allocator<char> 
 > > > 
&,std::basic_string<char,std::char_traits<char>,std::allocator<char> > 
const 
&),boost::python::default_call_policies,boost::mpl::vector3<void,std::vector<std::basic_string<char,std::char_traits<char>,std::allocator<char> 
 >,std::allocator<std::basic_string<char,std::char_traits<char>,std::allocator<char> 
 > > > 
&,std::basic_string<char,std::char_traits<char>,std::allocator<char> > 
const &> >::operator()(_object * args_=0x009d58a0, _object * 
__formal=0x00000000)  Line 201 + 0x2a bytes	C++
 
_biopsy.pyd!boost::python::objects::caller_py_function_impl<boost::python::detail::caller<void 
(__cdecl*)(std::vector<std::basic_string<char,std::char_traits<char>,std::allocator<char> 
 >,std::allocator<std::basic_string<char,std::char_traits<char>,std::allocator<char> 
 > > > 
&,std::basic_string<char,std::char_traits<char>,std::allocator<char> > 
const 
&),boost::python::default_call_policies,boost::mpl::vector3<void,std::vector<std::basic_string<char,std::char_traits<char>,std::allocator<char> 
 >,std::allocator<std::basic_string<char,std::char_traits<char>,std::allocator<char> 
 > > > 
&,std::basic_string<char,std::char_traits<char>,std::allocator<char> > 
const &> > >::operator()(_object * args=0x009d58a0, _object * 
kw=0x00000000)  Line 39	C++




From dave at boost-consulting.com  Thu Jun 22 14:20:13 2006
From: dave at boost-consulting.com (David Abrahams)
Date: Thu, 22 Jun 2006 08:20:13 -0400
Subject: [C++-sig] [ vector | list ] indexing suite
References: <449861C7.15757.1D9833D@s_sourceforge.nedprod.com>
	<44989297.6030105@boost-consulting.com>
	<upsh3drti.fsf@boost-consulting.com>
	<7465b6170606202232u19e89e3n8422cfb74aa9a538@mail.gmail.com>
Message-ID: <uac85crgi.fsf@boost-consulting.com>

"Roman Yakovenko" <roman.yakovenko at gmail.com> writes:

> On 6/21/06, David Abrahams <dave at boost-consulting.com> wrote:
>> Joel de Guzman <joel at boost-consulting.com> writes:
>>
>> >> TnFOX uses the new indexing suite. Saved me hours of extra work.
>> >> Search the mailing list archives for Raoul Gough.
>> >
>> > I think this is definitely the way to go. It makes perfect sense
>> > for Raoul's work to replace the current indexing suite. Last
>> > time I heard from Raoul, he was having some problems with VC6
>> > porting. I wonder if VC6 is still relevant for Boost.Python
>> > now, Dave?
>>
>> Having trouble forming an opinion.  Unfortunately there are serious
>> embedded platforms that are still stuck with vc6.
>
> In this case, is it possible to apply the patch?
>
> Another question: can both indexing suite co-exists?
> if so may be it possible to move indexing_v2 to main branch?
> The main objective to this is easiest install procedure and earlier testing.

FYI, I'm on vacation right now, so I won't be making any heavy
decisions :)

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com



From dave at boost-consulting.com  Thu Jun 22 14:27:33 2006
From: dave at boost-consulting.com (David Abrahams)
Date: Thu, 22 Jun 2006 08:27:33 -0400
Subject: [C++-sig] Protected destructor compile error
References: <bea9869d0606211453p191d307o2dbeb9175affdc12@mail.gmail.com>
Message-ID: <u4pydcr4a.fsf@boost-consulting.com>

"Allen Bierbaum" <abierbaum at gmail.com> writes:

> It looks like there is a problem with the "const Viewport& port"
> argument being passed.  For some reason the generated code seems to
> want to be able to destroy instances of the Viewport class.  This
> class has protected constructor and destructor as it is reference
> counted and the user can only create them using a helper function
> (Camera::create()).
>
> Can anyone describe what is happening here 

Boost.Python assumes that a const reference argument means that the
function can accept "rvalues", e.g. objects implicitly converted from
other python types.  The generated converter thus has storage for a
temporary Viewport object which potentially needs to be destroyed
after the wrapped function is called.

> and what prerequisites
> exist for this call to work?  Is there some call policy that I should
> be using to make this work?

Nope, but a thin wrapper that exposes Viewport& instead of Viewport
const& will help you.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com



From abierbaum at gmail.com  Thu Jun 22 15:28:41 2006
From: abierbaum at gmail.com (Allen Bierbaum)
Date: Thu, 22 Jun 2006 08:28:41 -0500
Subject: [C++-sig] Protected destructor compile error
In-Reply-To: <u4pydcr4a.fsf@boost-consulting.com>
References: <bea9869d0606211453p191d307o2dbeb9175affdc12@mail.gmail.com>
	<u4pydcr4a.fsf@boost-consulting.com>
Message-ID: <bea9869d0606220628t25a3634eq59138324b0fddf8f@mail.gmail.com>

On 6/22/06, David Abrahams <dave at boost-consulting.com> wrote:
> "Allen Bierbaum" <abierbaum at gmail.com> writes:
>
> > It looks like there is a problem with the "const Viewport& port"
> > argument being passed.  For some reason the generated code seems to
> > want to be able to destroy instances of the Viewport class.  This
> > class has protected constructor and destructor as it is reference
> > counted and the user can only create them using a helper function
> > (Camera::create()).
> >
> > Can anyone describe what is happening here
>
> Boost.Python assumes that a const reference argument means that the
> function can accept "rvalues", e.g. objects implicitly converted from
> other python types.  The generated converter thus has storage for a
> temporary Viewport object which potentially needs to be destroyed
> after the wrapped function is called.

Ok.  This makes sense although I still don't fully understand what is
happening behind the scenes but from this description I should be able
to look at the code to find out.

>
> > and what prerequisites
> > exist for this call to work?  Is there some call policy that I should
> > be using to make this work?
>
> Nope, but a thin wrapper that exposes Viewport& instead of Viewport
> const& will help you.

This makes sense as well.  Unfortunately there could be hundreds of
places in the code where I would need to do this so it will be
difficult but I will do it.

One additional question:  Is there any metafunction (something like
pointee<>) that could be specialized to tell boost.python that when it
sees a const& of this type it should not assume that it can make a
temporary?  In effect some way to tell it to choose to use the path
that the thin wrapper would use instead.

If there isn't a metafunction like this, would it be difficult to
implement something like this?  (if not, I am willing to work on it if
you give me some direction on where to look and what to look for)  To
help with cases like mine, it would be nice to make the default
version use type traits to detect if the type has a public destructor
(and anything else it needs to make the temporary).

Thanks again for the quick response.

-Allen

>
> --
> Dave Abrahams
> Boost Consulting
> www.boost-consulting.com
>
> _______________________________________________
> C++-sig mailing list
> C++-sig at python.org
> http://mail.python.org/mailman/listinfo/c++-sig
>


From joel at boost-consulting.com  Thu Jun 22 16:06:09 2006
From: joel at boost-consulting.com (Joel de Guzman)
Date: Thu, 22 Jun 2006 22:06:09 +0800
Subject: [C++-sig] boost::python std::vector< std::string > crash
In-Reply-To: <e7ds1k$bmu$1@sea.gmane.org>
References: <e7ds1k$bmu$1@sea.gmane.org>
Message-ID: <449AA3D1.1080805@boost-consulting.com>

John Reid wrote:
> Hi,
> 
> I'm using boost/python/suite/indexing/vector_indexing_suite.hpp from the 
> cvs_head version of boost python on MSVC 8. I'm getting a runtime 
> assertion failure when appending to a vector of strings.
> 
[snips]
> 
> There seems to be very little in the documenation about how std::strings 
> are handled. They seem to work fine when used outside containers. Also 
> containers of user-defined classes (as in the vector indexing example, 
> struct X) and containers of doubles, ints, etc.. seem to work fine.
> 
> What am I missing here?

That's odd. I can't reproduce your results. To be sure, I added
your test in libs/python/test/vector_indexing_suite.cpp/py.
I'm getting no crash. Please check out the test suite and run it.

Regards,
-- 
Joel de Guzman
http://www.boost-consulting.com
http://spirit.sf.net


From j.reid at mail.cryst.bbk.ac.uk  Thu Jun 22 17:59:51 2006
From: j.reid at mail.cryst.bbk.ac.uk (John Reid)
Date: Thu, 22 Jun 2006 16:59:51 +0100
Subject: [C++-sig] boost::python std::vector< std::string > crash
In-Reply-To: <449AA3D1.1080805@boost-consulting.com>
References: <e7ds1k$bmu$1@sea.gmane.org>
	<449AA3D1.1080805@boost-consulting.com>
Message-ID: <e7eeqp$gt3$1@sea.gmane.org>

Joel de Guzman wrote:
> 
> [snips]
> 
> That's odd. I can't reproduce your results. To be sure, I added
> your test in libs/python/test/vector_indexing_suite.cpp/py.
> I'm getting no crash. Please check out the test suite and run it.
> 
> Regards,


The test suite crashes in the same place as my code. I guess there's 
something wrong in my build set up.

I'm on XP with

Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 14.00.50727.42 
for 80x86

and

Python 2.4.3

I've tried various set ups to build python extensions, perhaps I bungled 
something up somewhere. I'm going to try doing a complete refresh from 
cvs and running the python test again unless anyone has something better 
to suggest.

Thanks,
John.



From j.reid at mail.cryst.bbk.ac.uk  Thu Jun 22 19:12:29 2006
From: j.reid at mail.cryst.bbk.ac.uk (John Reid)
Date: Thu, 22 Jun 2006 18:12:29 +0100
Subject: [C++-sig] boost::python std::vector< std::string > crash
In-Reply-To: <e7eeqp$gt3$1@sea.gmane.org>
References: <e7ds1k$bmu$1@sea.gmane.org>	<449AA3D1.1080805@boost-consulting.com>
	<e7eeqp$gt3$1@sea.gmane.org>
Message-ID: <e7ej2v$17e$1@sea.gmane.org>



John Reid wrote:
> Joel de Guzman wrote:
> 
>>[snips]
>>
>>That's odd. I can't reproduce your results. To be sure, I added
>>your test in libs/python/test/vector_indexing_suite.cpp/py.
>>I'm getting no crash. Please check out the test suite and run it.
>>
>>Regards,
> 
> 

I've investigated further. I wasn't aware of the correct way to run the 
python tests. Now I've recompiled a clean version of boost cvs_head 
everything works ok if I run the tests using 'bjam -sPYTHON_TEST_ARGS= 
test' in the 'boost\libs\python\test' directory.

What might be relevant is that if I run 'bjam' with no arguments, I can 
generate 'vector_indexing_suite_ext.pyd' in the 
'boost\bin\boost\libs\python\test\vector_indexing_suite_ext.pyd\vc-8_0\debug\threading-multi'
directory. When I manually set PYTHONPATH to this directory and run 
'python vector_indexing_suite.py' I get the crash as before. Is this 
expected, relevant and/or interesting? I would have thought this would 
just have run that particular test.

Any help appreciated as always. Thanks,
John.



From roman.yakovenko at gmail.com  Thu Jun 22 21:12:32 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Thu, 22 Jun 2006 22:12:32 +0300
Subject: [C++-sig] Protected destructor compile error
In-Reply-To: <u4pydcr4a.fsf@boost-consulting.com>
References: <bea9869d0606211453p191d307o2dbeb9175affdc12@mail.gmail.com>
	<u4pydcr4a.fsf@boost-consulting.com>
Message-ID: <7465b6170606221212i5d8a5f80mec3591e0b4ed6bf5@mail.gmail.com>

On 6/22/06, David Abrahams <dave at boost-consulting.com> wrote:
> Boost.Python assumes that a const reference argument means that the
> function can accept "rvalues", e.g. objects implicitly converted from
> other python types.

I am sure I am missing something trivial.

Why Boost.Python does such assumption? In my opinion Boost.Python, in this case
does not have to make any assumption.

If user want to enjoy from conversion he should register it, right? He
does this using:

1. implicitly_convertible
2. register_ptr_to_python
3. class_::bases

If none of those has been used, Boost.Python can safely assume that
the class could not be built from other python types.

Would you mind to explain, please?

Thank you.

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From roman.yakovenko at gmail.com  Thu Jun 22 21:19:37 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Thu, 22 Jun 2006 22:19:37 +0300
Subject: [C++-sig] [ vector | list ] indexing suite
In-Reply-To: <uac85crgi.fsf@boost-consulting.com>
References: <449861C7.15757.1D9833D@s_sourceforge.nedprod.com>
	<44989297.6030105@boost-consulting.com>
	<upsh3drti.fsf@boost-consulting.com>
	<7465b6170606202232u19e89e3n8422cfb74aa9a538@mail.gmail.com>
	<uac85crgi.fsf@boost-consulting.com>
Message-ID: <7465b6170606221219k59d1d97eh61ae2730cca23ac8@mail.gmail.com>

On 6/22/06, David Abrahams <dave at boost-consulting.com> wrote:
> FYI, I'm on vacation right now, so I won't be making any heavy
> decisions :)

When I am on vacation the only heavy decisions I have to make are:
1. where and what to eat
2. to go to the pool or to the sea

:-)

Enjoy your vacation!

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From hpj at urpla.net  Thu Jun 22 22:30:07 2006
From: hpj at urpla.net (Hans-Peter Jansen)
Date: Thu, 22 Jun 2006 22:30:07 +0200
Subject: [C++-sig] boost/python 1.33.1 breaks aliasing rules
Message-ID: <200606222230.07791.hpj@urpla.net>

Hi Philipp,

since you seem the originator of that boost patch:

http://mail.python.org/pipermail/c++-sig/2005-December/009895.html

you might want to know, that this broke boost 1.33.1 compilation on SuSE 
9.3:

    "c++"  -O2 -g -march=i586 -mcpu=i686 -fmessage-length=0  -c -Wall 
-ftemplate-depth-255  -DNDEBUG -DNDEBUG -DBOOST_PYTHON_DYNAMIC_LIB 
-DBOOST_PYTHON_SOURCE  -O2 -finline-functions -Wno-inline -fPIC   
-I"bin/boost/libs/python/build"   -I "/usr/include/python2.4" -I 
"/usr/src/packages/BUILD/boost_1_33_1"  -o 
"bin/boost/libs/python/build/libboost_python.so/gcc/release/shared-linkable-true/long.o"  
"/usr/src/packages/BUILD/boost_1_33_1/libs/
python/build/../src/long.cpp" 
    

/usr/src/packages/BUILD/boost_1_33_1/libs/python/src/long.cpp: In 
constructor `boost::python::detail::long_base::long_base()':
/usr/src/packages/BUILD/boost_1_33_1/libs/python/src/long.cpp:27: error: 
`new_reference' specified as declarator-id
/usr/src/packages/BUILD/boost_1_33_1/libs/python/src/long.cpp:27: error: 
multiple declarations `int' and `boost::python::api::object'
/usr/src/packages/BUILD/boost_1_33_1/libs/python/src/long.cpp:27: error: 
invalid conversion from `PyObject*' to `int'
/usr/src/packages/BUILD/boost_1_33_1/libs/python/src/long.cpp:27: 
warning: unused variable `int boost::python::detail::new_reference'
/usr/src/packages/BUILD/boost_1_33_1/libs/python/src/long.cpp:27: 
confused by earlier errors, bailing out
...failed gcc-C++-action 
bin/boost/libs/python/build/libboost_python.so/gcc/release/shared-linkable-true/long.o...

which led to a broken packman package  boost-1.33.1-17.pm.1.i586.rpm for 
that release (at least):
# rpm -V boost
missing    /usr/lib/libboost_python-gcc-1_33.so.1.33.0
missing    /usr/lib/libboost_python-gcc-1_33_1.so.1.33.1

It's a pity, that rpm doesn't check for dangling symlinks, otherwise 
this failure would have been caught. Why the boost build system bjam 
doesn't propagate errors properly escapes me at the moment..

Essentially reverting your patch to long_base::long_base() fixed it for 
me, see attached patch for reference. 

Manfred, you might want to apply it to your boost build in order to 
avoid unspecific "boost doesn't work with python", 
"libboost_python*.so* missing" complains..

Cheers,
Pete
-------------- next part --------------
A non-text attachment was scrubbed...
Name: boost-fix_python_build.patch
Type: text/x-diff
Size: 652 bytes
Desc: not available
URL: <http://mail.python.org/pipermail/cplusplus-sig/attachments/20060622/69107419/attachment.patch>

From rwgk at yahoo.com  Fri Jun 23 01:22:09 2006
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Thu, 22 Jun 2006 16:22:09 -0700 (PDT)
Subject: [C++-sig] Protected destructor compile error
In-Reply-To: <7465b6170606221212i5d8a5f80mec3591e0b4ed6bf5@mail.gmail.com>
Message-ID: <20060622232209.90938.qmail@web31514.mail.mud.yahoo.com>

--- Roman Yakovenko <roman.yakovenko at gmail.com> wrote:

> On 6/22/06, David Abrahams <dave at boost-consulting.com> wrote:
> > Boost.Python assumes that a const reference argument means that the
> > function can accept "rvalues", e.g. objects implicitly converted from
> > other python types.
> 
> I am sure I am missing something trivial.
> 
> Why Boost.Python does such assumption? In my opinion Boost.Python, in this
> case
> does not have to make any assumption.
> 
> If user want to enjoy from conversion he should register it, right? He
> does this using:
> 
> 1. implicitly_convertible
> 2. register_ptr_to_python
> 3. class_::bases
> 
> If none of those has been used, Boost.Python can safely assume that
> the class could not be built from other python types.

Boost.Python also supports custom rvalue converters. See the FAQ for examples.
You'd need two different mechanisms for types that allow custom from_python
conversions (involving the construction and destruction of temporaries) and
those that don't. I am guessing David considered that extra effort not
worthwhile or impractical.


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 


From dave at boost-consulting.com  Fri Jun 23 05:12:50 2006
From: dave at boost-consulting.com (David Abrahams)
Date: Thu, 22 Jun 2006 23:12:50 -0400
Subject: [C++-sig] Protected destructor compile error
References: <7465b6170606221212i5d8a5f80mec3591e0b4ed6bf5@mail.gmail.com>
	<20060622232209.90938.qmail@web31514.mail.mud.yahoo.com>
Message-ID: <uk678a7kd.fsf@boost-consulting.com>

"Ralf W. Grosse-Kunstleve" <rwgk at yahoo.com> writes:

> --- Roman Yakovenko <roman.yakovenko at gmail.com> wrote:
>
>> On 6/22/06, David Abrahams <dave at boost-consulting.com> wrote:
>> > Boost.Python assumes that a const reference argument means that
>> > the function can accept "rvalues", e.g. objects implicitly
>> > converted from other python types.
>> 
>> I am sure I am missing something trivial.
>> 
>> Why Boost.Python does such assumption? 

Because that's the way it works in C++.  The aim of Boost.Python is to
make the Python interface reflect the C++ interface.  Probably
"assumes" was the wrong word.  I should have said "is designed so."

>> In my opinion Boost.Python, in this case does not have to make any
>> assumption.
>> 
>> If user want to enjoy from conversion he should register it, right?
>> He does this using:
>> 
>> 1. implicitly_convertible
>> 2. register_ptr_to_python
>> 3. class_::bases

bases<...> has nothing to do with it; they don't enable rvalue
conversions.  Nor does register_ptr_to_python.  We're talking about
from-python conversions here.

>> If none of those has been used, Boost.Python can safely assume that
>> the class could not be built from other python types.
>
> Boost.Python also supports custom rvalue converters. See the FAQ for
> examples.  

And there are built-in rvalue conversions, e.g., Python str to
std::string, Python float to C++ float, etc.  You definitely don't
want the temporary string or float binding to a non-const reference.

> You'd need two different mechanisms for types that allow
> custom from_python conversions (involving the construction and
> destruction of temporaries) and those that don't. I am guessing
> David considered that extra effort not worthwhile or impractical.

Not really; I actually never considered treating types nonuniformly in
this way, and I don't think it would be tenable to do so based on
whether custom from_python conversions are allowed.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com



From dave at boost-consulting.com  Fri Jun 23 05:36:33 2006
From: dave at boost-consulting.com (David Abrahams)
Date: Thu, 22 Jun 2006 23:36:33 -0400
Subject: [C++-sig] Protected destructor compile error
References: <bea9869d0606211453p191d307o2dbeb9175affdc12@mail.gmail.com>
	<u4pydcr4a.fsf@boost-consulting.com>
	<bea9869d0606220628t25a3634eq59138324b0fddf8f@mail.gmail.com>
Message-ID: <uejxga6gu.fsf@boost-consulting.com>

"Allen Bierbaum" <abierbaum at gmail.com> writes:

> One additional question: Is there any metafunction (something like
> pointee<>) that could be specialized to tell boost.python that when
> it sees a const& of this type it should not assume that it can make
> a temporary?  In effect some way to tell it to choose to use the
> path that the thin wrapper would use instead.
>
> If there isn't a metafunction like this, would it be difficult to
> nimplement something like this?

I think the answer to both questions is no.  It sounds like a
reasonable thing to do, though.

> (if not, I am willing to work on it if you give me some direction on
> where to look and what to look for) To help with cases like mine, it
> would be nice to make the default version use type traits to detect
> if the type has a public destructor (and anything else it needs to
> make the temporary).

It's not possible to detect that, unfortunately.  And _making_ the
temporary is a completely separate issue from destroying it.  The part
of the converter that makes the temporary is looked up dynamically
based on the source Python type and in this case of course, nothing is
ever found.  The fact that the destruction code is instantiated
unconditionally by the static part of the is really a sort of
optimization.

The simplest way to make this work for everyone would be to make the
dynamic part of the converter responsible for destroying the
temporary.  That would make your case work without the need for any
traits.  However, there's a tradeoff; there would be an extra function
pointer indirection, at least.  In the case of converting to a C++
type with a trivial destructor, that no-op call through a function
pointer would replace an "inline no-op" (i.e., nothing at all).  Other
minor costs include storage for that function pointer.

Also there's the issue that this would cause a hard incompatibility
with extension modules built against an earlier Boost.Python (their
dynamic converters don't supply that destruction code).

Did I mention I'm not making any heavy decisions while I'm on
vacation?  Why don't y'all discuss the tradeoffs here and decide
whether it's worth it.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com



From hans_meine at gmx.net  Fri Jun 23 17:35:00 2006
From: hans_meine at gmx.net (Hans Meine)
Date: Fri, 23 Jun 2006 17:35:00 +0200
Subject: [C++-sig] bp::arg() also requires types to be declared
In-Reply-To: <44997B62.6825.E7172@s_sourceforge.nedprod.com>
References: <4498BA51.13536.3335777@s_sourceforge.nedprod.com>
	<44997B62.6825.E7172@s_sourceforge.nedprod.com>
Message-ID: <200606231735.02769.hans_meine@gmx.net>

On Wednesday 21 June 2006 18:01, Niall Douglas wrote:
> If, and only if, both Foo and Foo2 have constructors with no default
> args, you can delay the constructor declaration and declare it later
> via bp::init<>. Otherwise you are trapped and can't use bp::arg() at
> all.

That's not the whole truth.  AFAICS, you can use no_init to delay constructor 
declaration in any case, create all class_<...> instantiations and then later 
add the methods.  Maybe that helps?

Ciao, /  /
     /--/
    /  / ANS


From roman.yakovenko at gmail.com  Fri Jun 23 20:04:31 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Fri, 23 Jun 2006 21:04:31 +0300
Subject: [C++-sig] bp::arg() also requires types to be declared
In-Reply-To: <200606231735.02769.hans_meine@gmx.net>
References: <4498BA51.13536.3335777@s_sourceforge.nedprod.com>
	<44997B62.6825.E7172@s_sourceforge.nedprod.com>
	<200606231735.02769.hans_meine@gmx.net>
Message-ID: <7465b6170606231104g14eb042fy3f57e99e322209d4@mail.gmail.com>

On 6/23/06, Hans Meine <hans_meine at gmx.net> wrote:
> On Wednesday 21 June 2006 18:01, Niall Douglas wrote:
> > If, and only if, both Foo and Foo2 have constructors with no default
> > args, you can delay the constructor declaration and declare it later
> > via bp::init<>. Otherwise you are trapped and can't use bp::arg() at
> > all.
>
> That's not the whole truth.  AFAICS, you can use no_init to delay constructor
> declaration in any case, create all class_<...> instantiations and then later
> add the methods.  Maybe that helps?

Thanks for idea. I am aware of this work around. Boost.Python contains
functionality
that helps to use such approach - def_visitor.
I can change pyplusplus to generate code, that implements this idea.
But right now it is too much work, that should be done + generated
code will be less readable
( IMHO ).

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From roman.yakovenko at gmail.com  Fri Jun 23 20:26:44 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Fri, 23 Jun 2006 21:26:44 +0300
Subject: [C++-sig] rvalue custom converters ( was Protected destructor
	compile error )
Message-ID: <7465b6170606231126o7d2e6543x7b577b16fa4a19a0@mail.gmail.com>

On 6/23/06, Ralf W. Grosse-Kunstleve <rwgk at yahoo.com> wrote:
> Boost.Python also supports custom rvalue converters. See the FAQ for examples.
> You'd need two different mechanisms for types that allow custom from_python
> conversions (involving the construction and destruction of temporaries) and
> those that don't.

It seems that I can not find documentation about custom rvalue converters.
How does this work? What happen if I register custom rvalue converter,
for example for vector< MyClassX > and also class_< vector< MyClassX >
> using
indexing suite.

Can you point me to the documentation?

Thank you!

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From abierbaum at gmail.com  Fri Jun 23 20:29:28 2006
From: abierbaum at gmail.com (Allen Bierbaum)
Date: Fri, 23 Jun 2006 13:29:28 -0500
Subject: [C++-sig] Protected destructor compile error
In-Reply-To: <uejxga6gu.fsf@boost-consulting.com>
References: <bea9869d0606211453p191d307o2dbeb9175affdc12@mail.gmail.com>
	<u4pydcr4a.fsf@boost-consulting.com>
	<bea9869d0606220628t25a3634eq59138324b0fddf8f@mail.gmail.com>
	<uejxga6gu.fsf@boost-consulting.com>
Message-ID: <bea9869d0606231129j4ac13806pe35bbeffb489d2e0@mail.gmail.com>

On 6/22/06, David Abrahams <dave at boost-consulting.com> wrote:
> "Allen Bierbaum" <abierbaum at gmail.com> writes:
>
> > One additional question: Is there any metafunction (something like
> > pointee<>) that could be specialized to tell boost.python that when
> > it sees a const& of this type it should not assume that it can make
> > a temporary?  In effect some way to tell it to choose to use the
> > path that the thin wrapper would use instead.
> >
> > If there isn't a metafunction like this, would it be difficult to
> > nimplement something like this?
>
> I think the answer to both questions is no.  It sounds like a
> reasonable thing to do, though.
>
> > (if not, I am willing to work on it if you give me some direction on
> > where to look and what to look for) To help with cases like mine, it
> > would be nice to make the default version use type traits to detect
> > if the type has a public destructor (and anything else it needs to
> > make the temporary).
>
> It's not possible to detect that, unfortunately.

It would not be possible to do in C++, but it would be possible to
have a code generator like pyplusplus detect this case and
automatically set some metafunction/flag/whatever to tell the system
that it can't destruct the type so it should not allow creation of a
temporary.

Now that I think about my case it more detail I am starting to wonder
if there is a problem with the API I a binding.
For example the reason this method:

bool Camera::calcViewRay (Line &line, Int32 x, Int32 y, const Viewport &port);


is not working is because Viewport objects are only supposed to be
handled using ViewportPtrs (custom ref-counted shared pointer).  Since
they are not handled this way in this case it causes these problems.

This API would cause problems even in pure C++ code if someone passed
in an object of another type that is implicitly convertible to a
Viewport.  If Viewport had a constructor that took an std::string for
example and they tried to pass that in C++ would attempt to create a
temporary behind the scenes and have the same problem at compile time.
 So I guess this is a case of valid (but possibly questionable) C++
that boost.python fails to wrap by default.  (I should note that in my
case the API does not allow implicit conversion to Viewport or any of
the other types that have protected destructors.  They are all created
only by calling a static creation method that returns a smart ptr.)

It is worth mentioning that the other methods where this is a problem
in the current library I am binding are things like
operator==(const&), operator<(const&), and operator!=(const&).

> And _making_ the
> temporary is a completely separate issue from destroying it.  The part
> of the converter that makes the temporary is looked up dynamically
> based on the source Python type and in this case of course, nothing is
> ever found.  The fact that the destruction code is instantiated
> unconditionally by the static part of the is really a sort of
> optimization.
>
> The simplest way to make this work for everyone would be to make the
> dynamic part of the converter responsible for destroying the
> temporary.  That would make your case work without the need for any
> traits.  However, there's a tradeoff; there would be an extra function
> pointer indirection, at least.  In the case of converting to a C++
> type with a trivial destructor, that no-op call through a function
> pointer would replace an "inline no-op" (i.e., nothing at all).  Other
> minor costs include storage for that function pointer.
>
> Also there's the issue that this would cause a hard incompatibility
> with extension modules built against an earlier Boost.Python (their
> dynamic converters don't supply that destruction code).
>
> Did I mention I'm not making any heavy decisions while I'm on
> vacation?  Why don't y'all discuss the tradeoffs here and decide
> whether it's worth it.

Has anyone else ever ran into this problem?  If not, then maybe it is
not worth the effort and performance cost to try to fix generically.
I could always just try to make some smart wrapper code with
pyplusplus that automatically detects this case and creates a thin
wrapper for me.  (any ideas here Roman?)

David: Thanks for replying while on vacation.  I wouldn't even be
reading my e-mail if I was on vacation. :)

-Allen

> --
> Dave Abrahams
> Boost Consulting
> www.boost-consulting.com
>
> _______________________________________________
> C++-sig mailing list
> C++-sig at python.org
> http://mail.python.org/mailman/listinfo/c++-sig
>


From roman.yakovenko at gmail.com  Fri Jun 23 20:56:42 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Fri, 23 Jun 2006 21:56:42 +0300
Subject: [C++-sig] Protected destructor compile error
In-Reply-To: <bea9869d0606231129j4ac13806pe35bbeffb489d2e0@mail.gmail.com>
References: <bea9869d0606211453p191d307o2dbeb9175affdc12@mail.gmail.com>
	<u4pydcr4a.fsf@boost-consulting.com>
	<bea9869d0606220628t25a3634eq59138324b0fddf8f@mail.gmail.com>
	<uejxga6gu.fsf@boost-consulting.com>
	<bea9869d0606231129j4ac13806pe35bbeffb489d2e0@mail.gmail.com>
Message-ID: <7465b6170606231156u79114a6dlf0dc06c6380ad093@mail.gmail.com>

On 6/23/06, Allen Bierbaum <abierbaum at gmail.com> wrote:
> I could always just try to make some smart wrapper code with
> pyplusplus that automatically detects this case and creates a thin
> wrapper for me.  (any ideas here Roman?)

Actually yes. You have 2 ways to go:
1. Contribute to pyplusplus:
    1. Add code to pyplusplus.decl_wrappers.calldef_t.has_wrapper.
This code should
        "understand", that function takes an argument const reference to
        non-destructable object.
    2. Add code that generates wrapper code to
code_creators/calldef.py relevant
        classes.
    This is not should be very difficult. If you need more
information, please ask on
    pyplusplus mailing list.

2. Easy and quick.
    Build custom small code generator, that will generate function
wrapper, for every
    function, that takes as argument "const Viewport&". After this, exclude all
    original functions and replace them with new functions. This
should work and it
    should take less then an hour to implement it.

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From rwgk at yahoo.com  Fri Jun 23 21:54:24 2006
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Fri, 23 Jun 2006 12:54:24 -0700 (PDT)
Subject: [C++-sig] rvalue custom converters ( was Protected destructor
	compile error )
In-Reply-To: <7465b6170606231126o7d2e6543x7b577b16fa4a19a0@mail.gmail.com>
Message-ID: <20060623195424.73729.qmail@web31509.mail.mud.yahoo.com>

--- Roman Yakovenko <roman.yakovenko at gmail.com> wrote:

> On 6/23/06, Ralf W. Grosse-Kunstleve <rwgk at yahoo.com> wrote:
> > Boost.Python also supports custom rvalue converters. See the FAQ for
> examples.
> > You'd need two different mechanisms for types that allow custom from_python
> > conversions (involving the construction and destruction of temporaries) and
> > those that don't.
> 
> It seems that I can not find documentation about custom rvalue converters.
> How does this work? What happen if I register custom rvalue converter,
> for example for vector< MyClassX > and also class_< vector< MyClassX >
> > using
> indexing suite.
> 
> Can you point me to the documentation?

The info in the FAQ is all there is. A couple of times David mentioned that the
mechanics could be simplified (and then formally documented), but it never
happened. -- It doesn't seem all that important to me. I find the examples
fully sufficient.


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 


From roman.yakovenko at gmail.com  Fri Jun 23 22:06:46 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Fri, 23 Jun 2006 23:06:46 +0300
Subject: [C++-sig] rvalue custom converters ( was Protected destructor
	compile error )
In-Reply-To: <20060623195424.73729.qmail@web31509.mail.mud.yahoo.com>
References: <7465b6170606231126o7d2e6543x7b577b16fa4a19a0@mail.gmail.com>
	<20060623195424.73729.qmail@web31509.mail.mud.yahoo.com>
Message-ID: <7465b6170606231306r642c299esec28696c58bc5804@mail.gmail.com>

On 6/23/06, Ralf W. Grosse-Kunstleve <rwgk at yahoo.com> wrote:
> > How does this work? What happen if I register custom rvalue converter,
> > for example for vector< MyClassX > and also class_< vector< MyClassX >
> > > using
> > indexing suite.
> >
> > Can you point me to the documentation?
>
> The info in the FAQ is all there is. A couple of times David mentioned that the
> mechanics could be simplified (and then formally documented), but it never
> happened. -- It doesn't seem all that important to me. I find the examples
> fully sufficient.

I will study the examples. I and Lakin Wecker work on Python bindings for Ogre.
That library contains few classes like Matrix3 or Matrix4( Vector3, Vector4 ).
Those classes have constructors from regular C++ arrays. It could be nice if
we can create Python bindings, where user could pass Python list or
tuple instead
of [Matrix|Vector][3|4] as argument to functions.

If I understand right, custom rvalue converters is the way to go, right?
The only question I have is how it will work in presents of registration of
[Matrix|Vector][3|4] classes?

Thank you for your help.

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From Bruce_Sherwood at ncsu.edu  Sat Jun 24 02:23:24 2006
From: Bruce_Sherwood at ncsu.edu (Bruce Sherwood)
Date: Fri, 23 Jun 2006 20:23:24 -0400
Subject: [C++-sig] VPython uses Boost
Message-ID: <449C85FC.30202@ncsu.edu>

VPython is an extension for Python to allow easy, "pythonic" 3D. It is 
used in education for various purposes, including teaching physics and 
programming, but it has also been used by research scientists to 
visualize systems or data in 3D. See vpython.org. VPython uses the Boost 
python and thread libraries.

Bruce Sherwood



From rwgk at yahoo.com  Sat Jun 24 07:02:59 2006
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Fri, 23 Jun 2006 22:02:59 -0700 (PDT)
Subject: [C++-sig] rvalue custom converters ( was Protected destructor
	compile error )
In-Reply-To: <7465b6170606231306r642c299esec28696c58bc5804@mail.gmail.com>
Message-ID: <20060624050259.97442.qmail@web31510.mail.mud.yahoo.com>

--- Roman Yakovenko <roman.yakovenko at gmail.com> wrote:
> I will study the examples. I and Lakin Wecker work on Python bindings for
> Ogre.
> That library contains few classes like Matrix3 or Matrix4( Vector3, Vector4
> ).
> Those classes have constructors from regular C++ arrays. It could be nice if
> we can create Python bindings, where user could pass Python list or
> tuple instead
> of [Matrix|Vector][3|4] as argument to functions.
> 
> If I understand right, custom rvalue converters is the way to go, right?
> The only question I have is how it will work in presents of registration of
> [Matrix|Vector][3|4] classes?

Situation 1 a:

  class_<matrix>(...)
    .def(init<matrix const&>()) // wrap the copy constructor
  ;

  Also define custom from_python converters converting Python sequences (e.g.
list, tuple, iter) to matrix. This way any function expecting a matrix const&
will also accept a Python sequence, including the copy constructor above.

Situation 1 b:

  If you don't want all function to convert sequences on the fly (e.g. because
you are afraid your users don't understand the impact on runtime performance),
don't define the custom from_python converters. Instead of wrapping the copy
constructor, use make_constructor to inject a constructor converting Python
sequences.

Situation 2:

  Define both custom to_python and from_python conversions, don't use class_.

Examples:

http://cctbx.cvs.sourceforge.net/cctbx/scitbx/include/scitbx/boost_python/container_conversions.h?view=markup

Which enables, e.g.:
tuple_mapping_fixed_size<matrix>(); // situation 2 above

http://cctbx.cvs.sourceforge.net/cctbx/scitbx/include/scitbx/stl/vector_wrapper.h?revision=1.4&view=markup

This also uses container_conversions.h, but only to define from_python
conversions. The copy constructor is wrapped. This is situation 1 a above.

http://cctbx.cvs.sourceforge.net/cctbx/scitbx/array_family/boost_python/flex_double.cpp?revision=1.31&view=markup

Look for from_list_or_tuple_of_lists_or_tuples() which is used with
make_constructor. This is situation 1 b above.


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 


From rwgk at yahoo.com  Sat Jun 24 07:44:08 2006
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Fri, 23 Jun 2006 22:44:08 -0700 (PDT)
Subject: [C++-sig] boost/python 1.33.1 breaks aliasing rules
Message-ID: <20060624054408.52339.qmail@web31507.mail.mud.yahoo.com>

Coincidentally, on Python-Dev Scott David Daniels just suggested this trick to
get rid of the "will break strict-aliasing rules" warnings:

>     PyObject_CallFunction((PyObject*) (void *) &PyRange_Type,
>                           "lll", start, start+len*step, step)

I.e. the trick is to insert an intermediate (void *) cast. It works with g++
4.1.0 at least. Could this be a general solution?


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 


From roman.yakovenko at gmail.com  Sat Jun 24 12:41:05 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Sat, 24 Jun 2006 13:41:05 +0300
Subject: [C++-sig] rvalue custom converters ( was Protected destructor
	compile error )
In-Reply-To: <20060624050259.97442.qmail@web31510.mail.mud.yahoo.com>
References: <7465b6170606231306r642c299esec28696c58bc5804@mail.gmail.com>
	<20060624050259.97442.qmail@web31510.mail.mud.yahoo.com>
Message-ID: <7465b6170606240341w28f3aed2l5c242ae4e53f31f2@mail.gmail.com>

On 6/24/06, Ralf W. Grosse-Kunstleve <rwgk at yahoo.com> wrote:
> ...

Thank you very much!

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From dave at boost-consulting.com  Sat Jun 24 15:21:56 2006
From: dave at boost-consulting.com (David Abrahams)
Date: Sat, 24 Jun 2006 09:21:56 -0400
Subject: [C++-sig] boost/python 1.33.1 breaks aliasing rules
References: <20060624054408.52339.qmail@web31507.mail.mud.yahoo.com>
Message-ID: <uslluy9hn.fsf@boost-consulting.com>

"Ralf W. Grosse-Kunstleve" <rwgk at yahoo.com> writes:

> Coincidentally, on Python-Dev Scott David Daniels just suggested this trick to
> get rid of the "will break strict-aliasing rules" warnings:
>
>>     PyObject_CallFunction((PyObject*) (void *) &PyRange_Type,
>>                           "lll", start, start+len*step, step)
>
> I.e. the trick is to insert an intermediate (void *) cast. It works with g++
> 4.1.0 at least. Could this be a general solution?

There's no such thing as a general solution to a warning (they're
implementation specific), and these "strict aliasing rules" are a C99
thing, and C++ is based on C89 (or C90, I forget which), and the code
above does the so it's hard to say what should be done.

Maybe reading
http://www.cellperformance.com/mike_acton/2006/06/understanding_strict_aliasing.html
would help.


-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com



From abierbaum at gmail.com  Sat Jun 24 16:10:47 2006
From: abierbaum at gmail.com (Allen Bierbaum)
Date: Sat, 24 Jun 2006 09:10:47 -0500
Subject: [C++-sig] Protected destructor compile error
In-Reply-To: <7465b6170606231156u79114a6dlf0dc06c6380ad093@mail.gmail.com>
References: <bea9869d0606211453p191d307o2dbeb9175affdc12@mail.gmail.com>
	<u4pydcr4a.fsf@boost-consulting.com>
	<bea9869d0606220628t25a3634eq59138324b0fddf8f@mail.gmail.com>
	<uejxga6gu.fsf@boost-consulting.com>
	<bea9869d0606231129j4ac13806pe35bbeffb489d2e0@mail.gmail.com>
	<7465b6170606231156u79114a6dlf0dc06c6380ad093@mail.gmail.com>
Message-ID: <bea9869d0606240710j2843dfcbxe7f42f78366b24a3@mail.gmail.com>

On 6/23/06, Roman Yakovenko <roman.yakovenko at gmail.com> wrote:
> On 6/23/06, Allen Bierbaum <abierbaum at gmail.com> wrote:
> > I could always just try to make some smart wrapper code with
> > pyplusplus that automatically detects this case and creates a thin
> > wrapper for me.  (any ideas here Roman?)
>
> Actually yes. You have 2 ways to go:
> 1. Contribute to pyplusplus:
>     1. Add code to pyplusplus.decl_wrappers.calldef_t.has_wrapper.
> This code should
>         "understand", that function takes an argument const reference to
>         non-destructable object.
>     2. Add code that generates wrapper code to
> code_creators/calldef.py relevant
>         classes.
>     This is not should be very difficult. If you need more
> information, please ask on
>     pyplusplus mailing list.

Sounds interesting but I don't know that code nearly as well as you.  :)

> 2. Easy and quick.
>     Build custom small code generator, that will generate function
> wrapper, for every
>     function, that takes as argument "const Viewport&". After this, exclude all
>     original functions and replace them with new functions. This
> should work and it
>     should take less then an hour to implement it.

This one sounds more approachable to me right now.  The question I
have though is how to add the function wrapper code to the generated
code.  Is this something I can do to the class decl wrapper?  What
method do I use to add it?  (if you provide pseudo code I can fill in
all the details and debug it)

Thanks for the quick reply.

-Allen

>
> --
> Roman Yakovenko
> C++ Python language binding
> http://www.language-binding.net/
> _______________________________________________
> C++-sig mailing list
> C++-sig at python.org
> http://mail.python.org/mailman/listinfo/c++-sig
>


From ngoodspeed at solidworks.com  Sat Jun 24 16:26:34 2006
From: ngoodspeed at solidworks.com (Nat Goodspeed)
Date: Sat, 24 Jun 2006 10:26:34 -0400
Subject: [C++-sig] Publish a big C++ vector to Python
Message-ID: <ED67712627032740B15C91998D43FB4C02CE6BBA@corp-mail4.solidworks.swk>

I'm interested by the recent discussion about the new indexing suite:

http://mail.python.org/pipermail/c++-sig/2006-June/010830.html

I will soon want to be able to pass very large volumes of numeric data
from C++ into a Python module. From the Python point of view, we can
treat it as read-only.

The C++ code internally uses std::vector<Vector3>, where each Vector3
object contains -- surprise -- double c[3].

We'd like to be able to export it to various different file formats. So
I expect the Python code to have to make at least one pass over the
data.

But I want to avoid requiring the Python code to make an additional pass
solely for the purpose of copying to another memory structure. If we
must do that, we should do it on the C++ side beforehand. (Yes, the
memory penalty is of some concern as well.)

What approach would you recommend?

I imagine a few different possibilities, none of which I've researched
yet:

1. Use the existing indexing suite to publish std::vector<Vector3>, if
it would suffice for our purposes.

2. Use the new indexing suite to publish std::vector<Vector3>. Perhaps
fortunately, we're still using Boost 1.31. But it sounds as if there are
other hurdles to jump?

3. Make a preliminary C++ pass to construct a Python array object
containing doubles, three array entries per Vector3. (I don't yet know
how to construct a Python 'array' of appropriate size in C++.)

4. Make a preliminary C++ pass to construct a NumPy array containing
doubles, as above. (I don't yet know how to do that either.)

5. Abstract away the whole problem by publishing a little proxy class
with access methods. (Could that possibly be as fast, within an order of
magnitude? For our purposes, "fast enough" means that memory access time
is still swamped by disk-write time.)

6. Other suggestions?

We'll be adding this to a large existing hand-coded boost::python
module, so -- although I'm intrigued -- converting to pyplusplus doesn't
seem like the short path.

Thanks for your suggestions!


From roman.yakovenko at gmail.com  Sat Jun 24 20:43:26 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Sat, 24 Jun 2006 21:43:26 +0300
Subject: [C++-sig] Protected destructor compile error
In-Reply-To: <bea9869d0606240710j2843dfcbxe7f42f78366b24a3@mail.gmail.com>
References: <bea9869d0606211453p191d307o2dbeb9175affdc12@mail.gmail.com>
	<u4pydcr4a.fsf@boost-consulting.com>
	<bea9869d0606220628t25a3634eq59138324b0fddf8f@mail.gmail.com>
	<uejxga6gu.fsf@boost-consulting.com>
	<bea9869d0606231129j4ac13806pe35bbeffb489d2e0@mail.gmail.com>
	<7465b6170606231156u79114a6dlf0dc06c6380ad093@mail.gmail.com>
	<bea9869d0606240710j2843dfcbxe7f42f78366b24a3@mail.gmail.com>
Message-ID: <7465b6170606241143l564a098av91273c1d05483aa2@mail.gmail.com>

On 6/24/06, Allen Bierbaum <abierbaum at gmail.com> wrote:
> On 6/23/06, Roman Yakovenko <roman.yakovenko at gmail.com> wrote:
> > On 6/23/06, Allen Bierbaum <abierbaum at gmail.com> wrote:
> > > I could always just try to make some smart wrapper code with
> > > pyplusplus that automatically detects this case and creates a thin
> > > wrapper for me.  (any ideas here Roman?)
> >
> > Actually yes. You have 2 ways to go:
> > 1. Contribute to pyplusplus:
> >     1. Add code to pyplusplus.decl_wrappers.calldef_t.has_wrapper.
> > This code should
> >         "understand", that function takes an argument const reference to
> >         non-destructable object.
> >     2. Add code that generates wrapper code to
> > code_creators/calldef.py relevant
> >         classes.
> >     This is not should be very difficult. If you need more
> > information, please ask on
> >     pyplusplus mailing list.
>
> Sounds interesting but I don't know that code nearly as well as you.  :)

I will guide you :-).

> > 2. Easy and quick.
> >     Build custom small code generator, that will generate function
> > wrapper, for every
> >     function, that takes as argument "const Viewport&". After this, exclude all
> >     original functions and replace them with new functions. This
> > should work and it
> >     should take less then an hour to implement it.
>
> This one sounds more approachable to me right now.  The question I
> have though is how to add the function wrapper code to the generated
> code.  Is this something I can do to the class decl wrapper?  What
> method do I use to add it?  (if you provide pseudo code I can fill in
> all the details and debug it)

struct X{
    void do_smth( const Viewport& );
};

wrapper_code_example = """
static void do_smth( X& x, Viewport& v){
    x.do_smth( v );
};
"""

mb = module_builder_t( ... )
for cls in mb.classes():
    for f in cls.member_functions( function=contains const Viewport& as arg )
        f.exclude()
        #Next code you will have to replace with something that
actually deals with f
        cls.add_wrapper_code ( wrapper_code_example )
        cls.add_code( 'def( "do_smth", &X_wrapper::do_smth )' )

cls.wrapper_alias will give you the name of the generated class wrapper name

code_creators/calldef.py - calldef_wrapper_t.args_declaration will help you
to get idea how to generate function wrapper code

There is another approach. In this approach you don't generate code,  but modify
code creators tree. I can explain you if you are interested.

> Thanks for the quick reply.

My pleasure.



-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From roman.yakovenko at gmail.com  Sat Jun 24 21:01:51 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Sat, 24 Jun 2006 22:01:51 +0300
Subject: [C++-sig] Publish a big C++ vector to Python
In-Reply-To: <ED67712627032740B15C91998D43FB4C02CE6BBA@corp-mail4.solidworks.swk>
References: <ED67712627032740B15C91998D43FB4C02CE6BBA@corp-mail4.solidworks.swk>
Message-ID: <7465b6170606241201i607e718flc58f7f4acbb2d381@mail.gmail.com>

On 6/24/06, Nat Goodspeed <ngoodspeed at solidworks.com> wrote:
> We'll be adding this to a large existing hand-coded boost::python
> module, so -- although I'm intrigued -- converting to pyplusplus doesn't
> seem like the short path.

You never know:

http://www.ogre3d.org/phpBB2addons/viewtopic.php?t=1478&sid=c7b74a8f74f265dc204f30446a64dc0d

20 - 30 hours from start to working Python bindings. Almost zero knowledge
in Boost.Python and sure zero knowledge in pyplusplus.

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From gleizesd at gmail.com  Mon Jun 26 09:58:53 2006
From: gleizesd at gmail.com (Damien Gleizes)
Date: Mon, 26 Jun 2006 07:58:53 +0000 (UTC)
Subject: [C++-sig] boost.python automatic policies
Message-ID: <loom.20060626T095803-859@post.gmane.org>

Hello,

I am currently using boost.Python for binding C++ code. In fact, I'am parsing a
sort of container wich stores all classes I need to bind, including their
methods and members (methods signatures, arguments types). While parsing these
classes, I generates CPP files containing the binding code (wich uses
boost.Python). The problem I'am confronted to concerns the use of policies. Is
there a way to know wich policies to use without knowing what the function
exactly do? Without policies the code does not compile, so I wonder if it is
possible to know wich policies to use just with the decoration of arguments and
return value (types, reference, pointer, const etc). 

Thanks for your suggestions!



From roman.yakovenko at gmail.com  Mon Jun 26 11:20:01 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Mon, 26 Jun 2006 12:20:01 +0300
Subject: [C++-sig] boost.python automatic policies
In-Reply-To: <loom.20060626T095803-859@post.gmane.org>
References: <loom.20060626T095803-859@post.gmane.org>
Message-ID: <7465b6170606260220s4f2b1a60hd2db3bd73d1a4a24@mail.gmail.com>

On 6/26/06, Damien Gleizes <gleizesd at gmail.com> wrote:
> Hello,
>
> I am currently using boost.Python for binding C++ code. In fact, I'am parsing a
> sort of container wich stores all classes I need to bind, including their
> methods and members (methods signatures, arguments types). While parsing these
> classes, I generates CPP files containing the binding code (wich uses
> boost.Python). The problem I'am confronted to concerns the use of policies. Is
> there a way to know wich policies to use without knowing what the function
> exactly do? Without policies the code does not compile, so I wonder if it is
> possible to know wich policies to use just with the decoration of arguments and
> return value (types, reference, pointer, const etc).

The short answer is no. The long answer is here:
http://www.boost.org/libs/python/doc/tutorial/doc/html/python/functions.html#python.call_policies

You can describe, what your function is doing, so it will be possible to suggest
best call policies

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From markus.schoepflin at comsoft.de  Mon Jun 26 11:29:40 2006
From: markus.schoepflin at comsoft.de (=?ISO-8859-1?Q?Markus_Sch=F6pflin?=)
Date: Mon, 26 Jun 2006 11:29:40 +0200
Subject: [C++-sig] [pyplusplus] Which version of gccxml should I use?
Message-ID: <e7o9e4$58q$1@sea.gmane.org>

Hello,

I am wondering which version of gccxml is recommended for use with 
pyplusplus? Use of gccxml-0.6.0 is discouraged by the gccxml homepage, so 
are there any known-to-be-good CVS snapshot tar balls available?

Thanks, Markus



From gleizesd at gmail.com  Mon Jun 26 11:45:04 2006
From: gleizesd at gmail.com (Damien Gleizes)
Date: Mon, 26 Jun 2006 09:45:04 +0000 (UTC)
Subject: [C++-sig] boost.python automatic policies
References: <loom.20060626T095803-859@post.gmane.org>
	<7465b6170606260220s4f2b1a60hd2db3bd73d1a4a24@mail.gmail.com>
Message-ID: <loom.20060626T114339-408@post.gmane.org>


Yes it's confirm what I was thinking, I will try to do it differently. Thank you
for your answer, I am now sure that what I wanted to do was in fact impossible!






From roman.yakovenko at gmail.com  Mon Jun 26 15:59:58 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Mon, 26 Jun 2006 16:59:58 +0300
Subject: [C++-sig] [pyplusplus] Which version of gccxml should I use?
In-Reply-To: <e7o9e4$58q$1@sea.gmane.org>
References: <e7o9e4$58q$1@sea.gmane.org>
Message-ID: <7465b6170606260659u72855de7xbb37a6b0dfaca6b4@mail.gmail.com>

On 6/26/06, Markus Sch?pflin <markus.schoepflin at comsoft.de> wrote:
> Hello,
>
> I am wondering which version of gccxml is recommended for use with
> pyplusplus? Use of gccxml-0.6.0 is discouraged by the gccxml homepage, so
> are there any known-to-be-good CVS snapshot tar balls available?
>

pygccxml works with latest/current version of gcc xml.

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From markus.schoepflin at comsoft.de  Mon Jun 26 16:14:50 2006
From: markus.schoepflin at comsoft.de (=?ISO-8859-1?Q?Markus_Sch=F6pflin?=)
Date: Mon, 26 Jun 2006 16:14:50 +0200
Subject: [C++-sig] [pyplusplus] Which version of gccxml should I use?
In-Reply-To: <7465b6170606260659u72855de7xbb37a6b0dfaca6b4@mail.gmail.com>
References: <e7o9e4$58q$1@sea.gmane.org>
	<7465b6170606260659u72855de7xbb37a6b0dfaca6b4@mail.gmail.com>
Message-ID: <e7oq4r$vmd$1@sea.gmane.org>

Roman Yakovenko wrote:
> On 6/26/06, Markus Sch?pflin <markus.schoepflin at comsoft.de> wrote:

>> I am wondering which version of gccxml is recommended for use with
>> pyplusplus? Use of gccxml-0.6.0 is discouraged by the gccxml homepage, so
>> are there any known-to-be-good CVS snapshot tar balls available?
>>
> 
> pygccxml works with latest/current version of gcc xml.
> 

I am sorry to insist, but would this be 0.6.0 or latest CVS?

Markus



From roman.yakovenko at gmail.com  Mon Jun 26 19:36:45 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Mon, 26 Jun 2006 20:36:45 +0300
Subject: [C++-sig] [pyplusplus] Which version of gccxml should I use?
In-Reply-To: <e7oq4r$vmd$1@sea.gmane.org>
References: <e7o9e4$58q$1@sea.gmane.org>
	<7465b6170606260659u72855de7xbb37a6b0dfaca6b4@mail.gmail.com>
	<e7oq4r$vmd$1@sea.gmane.org>
Message-ID: <7465b6170606261036u17bc4e0fk4c9c7156ce88f289@mail.gmail.com>

On 6/26/06, Markus Sch?pflin <markus.schoepflin at comsoft.de> wrote:
> Roman Yakovenko wrote:
> > On 6/26/06, Markus Sch?pflin <markus.schoepflin at comsoft.de> wrote:
>
> >> I am wondering which version of gccxml is recommended for use with
> >> pyplusplus? Use of gccxml-0.6.0 is discouraged by the gccxml homepage, so
> >> are there any known-to-be-good CVS snapshot tar balls available?
> >>
> >
> > pygccxml works with latest/current version of gcc xml.
> >
>
> I am sorry to insist, but would this be 0.6.0 or latest CVS?

Sorry, I was not clear - latest CVS version. I am subsribed to GCCXML
commit mailing
list, so pygccxml ( Subversion ) will always work with latest CVS
version of GCCXML.


-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From kleistereimer at gmx.de  Tue Jun 27 04:18:02 2006
From: kleistereimer at gmx.de (Kleistereimer)
Date: Tue, 27 Jun 2006 02:18:02 +0000
Subject: [C++-sig] extending a c++ object by overriding it's virtual
	functions..
Message-ID: <44A0955A.8060008@gmx.de>

hi!

i've a question about virtual functions and boost::python.
i use it with a boost c++ extended python which is embeeded in c++.

i have a virtual function that's going to be overridden in Python and
called polymorphically from C++. (so i'm not wrapping c++ for use from
python)

for instance i want to allow users of my software to override the
virtual void Login::OnLog(std::string) function using python.


that's what 'wrapper' and 'get_override' is used for, right?
if i would use it in python only, overriding works without these stuff too.

so i duplicated the example at
http://www.boost.org/libs/python/doc/tutorial/doc/html/python/exposing.html#python.virtual_functions_with_default_implementations

and it works fine inside python as advertised:
>>> base.f()
0
>>> derived.f()
42

but for this i wont need 'wrapper' and 'get_override'. This works for
any boost c++ extension.
now for the interesting stuff: (ak the problem)

after running the script i do:

	Base *b = new BaseWrap;
	int result = b->f();
	delete b;

now i would expect the result to be 42, but i got 0.
====================================================

so how to put 'wrapper' and 'get_override' to any use?


my sollution is to obtain the PyObject* of 'derived'
and to force it manualy into m_ptr of 'wrapper' using
'detail::initialize_wrapper'
then 'get_override' worked as expected and i got 42.

so i need to hardcode the name of the instantiated class ('derived')
into the BaseWrap class to be able to initialize the wrapper with it if
someone instantiated a c++ object and used f()... that's ugly


but my understanding of 'virtual function that's going to be overridden
in Python' is:

write a c++ class and a corresponding derived class in python, and then
use one or more instantiations of the c++ class without creating and
maintaining python instances etc.


or did i understood something completely wrong?

regards
kl


From markus.schoepflin at comsoft.de  Tue Jun 27 09:16:06 2006
From: markus.schoepflin at comsoft.de (=?ISO-8859-1?Q?Markus_Sch=F6pflin?=)
Date: Tue, 27 Jun 2006 09:16:06 +0200
Subject: [C++-sig] [pyplusplus] Which version of gccxml should I use?
In-Reply-To: <7465b6170606261036u17bc4e0fk4c9c7156ce88f289@mail.gmail.com>
References: <e7o9e4$58q$1@sea.gmane.org>	<7465b6170606260659u72855de7xbb37a6b0dfaca6b4@mail.gmail.com>	<e7oq4r$vmd$1@sea.gmane.org>
	<7465b6170606261036u17bc4e0fk4c9c7156ce88f289@mail.gmail.com>
Message-ID: <e7qlvm$vqk$2@sea.gmane.org>

Roman Yakovenko wrote:
> On 6/26/06, Markus Sch?pflin <markus.schoepflin at comsoft.de> wrote:
...
>>> pygccxml works with latest/current version of gcc xml.
>>>
>> I am sorry to insist, but would this be 0.6.0 or latest CVS?
> 
> Sorry, I was not clear - latest CVS version. I am subsribed to GCCXML
> commit mailing
> list, so pygccxml ( Subversion ) will always work with latest CVS
> version of GCCXML.

Thank you. Maybe you could add a little note to that effect to the 
documentation of pygccxml.

And do you know of any tarballs of this being available?

Markus



From roman.yakovenko at gmail.com  Tue Jun 27 09:25:25 2006
From: roman.yakovenko at gmail.com (Roman Yakovenko)
Date: Tue, 27 Jun 2006 10:25:25 +0300
Subject: [C++-sig] [pyplusplus] Which version of gccxml should I use?
In-Reply-To: <e7qlvm$vqk$2@sea.gmane.org>
References: <e7o9e4$58q$1@sea.gmane.org>
	<7465b6170606260659u72855de7xbb37a6b0dfaca6b4@mail.gmail.com>
	<e7oq4r$vmd$1@sea.gmane.org>
	<7465b6170606261036u17bc4e0fk4c9c7156ce88f289@mail.gmail.com>
	<e7qlvm$vqk$2@sea.gmane.org>
Message-ID: <7465b6170606270025o2a30510claee3bc5915a48a49@mail.gmail.com>

On 6/27/06, Markus Sch?pflin <markus.schoepflin at comsoft.de> wrote:
> Thank you. Maybe you could add a little note to that effect to the
> documentation of pygccxml.

I will do that.

> And do you know of any tarballs of this being available?
>

No, but if you follow instructions published here(
http://gccxml.org/HTML/Download.html )
and here( http://gccxml.org/HTML/Install.html ) you will be able to
setup it pretty quickly.

-- 
Roman Yakovenko
C++ Python language binding
http://www.language-binding.net/


From markus.schoepflin at comsoft.de  Tue Jun 27 09:35:31 2006
From: markus.schoepflin at comsoft.de (=?ISO-8859-1?Q?Markus_Sch=F6pflin?=)
Date: Tue, 27 Jun 2006 09:35:31 +0200
Subject: [C++-sig] [pyplusplus] Which version of gccxml should I use?
In-Reply-To: <7465b6170606270025o2a30510claee3bc5915a48a49@mail.gmail.com>
References: <e7o9e4$58q$1@sea.gmane.org>	<7465b6170606260659u72855de7xbb37a6b0dfaca6b4@mail.gmail.com>	<e7oq4r$vmd$1@sea.gmane.org>	<7465b6170606261036u17bc4e0fk4c9c7156ce88f289@mail.gmail.com>	<e7qlvm$vqk$2@sea.gmane.org>
	<7465b6170606270025o2a30510claee3bc5915a48a49@mail.gmail.com>
Message-ID: <e7qn43$4ij$1@sea.gmane.org>

Roman Yakovenko wrote:

> On 6/27/06, Markus Sch?pflin <markus.schoepflin at comsoft.de> wrote:
>> Thank you. Maybe you could add a little note to that effect to the
>> documentation of pygccxml.
> 
> I will do that.
> 
>> And do you know of any tarballs of this being available?
> 
> No, but if you follow instructions published here(
> http://gccxml.org/HTML/Download.html )
> and here( http://gccxml.org/HTML/Install.html ) you will be able to
> setup it pretty quickly.

Yes, I did read these pages. The only problem is that I'm firewalled, 
therefore I was asking about the tarball. I will have to figure something 
out, then...

Thanks for your help,
Markus



From Naceur.Meskini at sophia.inria.fr  Tue Jun 27 11:56:29 2006
From: Naceur.Meskini at sophia.inria.fr (Naceur Meskini)
Date: Tue, 27 Jun 2006 11:56:29 +0200
Subject: [C++-sig] KeyboardInterrupt with boost?!
In-Reply-To: <e7oq4r$vmd$1@sea.gmane.org>
References: <e7o9e4$58q$1@sea.gmane.org>	<7465b6170606260659u72855de7xbb37a6b0dfaca6b4@mail.gmail.com>
	<e7oq4r$vmd$1@sea.gmane.org>
Message-ID: <44A100CD.3020204@sophia.inria.fr>

Hi every body,

I have one question about how boost-python deal with Keyboard interrupting
I have exposed a collection of functions and I would like to be able to 
interupt the execution of some of them in the python side.

Thank you in advance.
Naceur.
INRIA Sophia Antipolis


From s_sourceforge at nedprod.com  Tue Jun 27 15:15:14 2006
From: s_sourceforge at nedprod.com (Niall Douglas)
Date: Tue, 27 Jun 2006 14:15:14 +0100
Subject: [C++-sig] [pyplusplus] Which version of gccxml should I use?
In-Reply-To: <7465b6170606261036u17bc4e0fk4c9c7156ce88f289@mail.gmail.com>
References: <e7oq4r$vmd$1@sea.gmane.org>
Message-ID: <44A13D72.21808.4A2C86@s_sourceforge.nedprod.com>

On 26 Jun 2006 at 20:36, Roman Yakovenko wrote:

> > >> I am wondering which version of gccxml is recommended for use with
> > >> pyplusplus? Use of gccxml-0.6.0 is discouraged by the gccxml homepage, so
> > >> are there any known-to-be-good CVS snapshot tar balls available?
> > >>
> > >
> > > pygccxml works with latest/current version of gcc xml.
> > >
> >
> > I am sorry to insist, but would this be 0.6.0 or latest CVS?
> 
> Sorry, I was not clear - latest CVS version. I am subsribed to GCCXML
> commit mailing
> list, so pygccxml ( Subversion ) will always work with latest CVS
> version of GCCXML.

It should be noted however that 0.6.0 works just fine for the 
majority of code. I'm still using it with TnFOX's python bindings. 
There are some bugs in the generated XML, but they can usually be 
ignored or if not, it's very easy to tell pyplusplus to fix the 
output code.

Cheers,
Niall





From rwgk at yahoo.com  Tue Jun 27 18:31:50 2006
From: rwgk at yahoo.com (Ralf W. Grosse-Kunstleve)
Date: Tue, 27 Jun 2006 09:31:50 -0700 (PDT)
Subject: [C++-sig] KeyboardInterrupt with boost?!
In-Reply-To: <44A100CD.3020204@sophia.inria.fr>
Message-ID: <20060627163150.19467.qmail@web31506.mail.mud.yahoo.com>

--- Naceur Meskini <Naceur.Meskini at sophia.inria.fr> wrote:

> I have one question about how boost-python deal with Keyboard interrupting
> I have exposed a collection of functions and I would like to be able to 
> interupt the execution of some of them in the python side.

I am not aware of an elegant way to handle this situation. If you find out
please let us know. It is not a Boost.Python specific problem btw. All C/C++
extensions have the same problem.

My solution is a very simple one: Almost all of our (very many) C++ functions
typically return after a few seconds at the most, most in fractions of a
second. Then Python gets a chance to handle the KeybordInterrupt. You don't
have to do anything special. I.e. simply leave that outermost loop in Python if
you can.


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 


From dave at boost-consulting.com  Tue Jun 27 19:42:46 2006
From: dave at boost-consulting.com (David Abrahams)
Date: Tue, 27 Jun 2006 13:42:46 -0400
Subject: [C++-sig] extending a c++ object by overriding it's virtual
	functions..
References: <44A0955A.8060008@gmx.de>
Message-ID: <uirmmsdex.fsf@boost-consulting.com>

Kleistereimer <kleistereimer at gmx.de> writes:

> hi!
>
> i've a question about virtual functions and boost::python.
> i use it with a boost c++ extended python which is embeeded in c++.
>
> i have a virtual function that's going to be overridden in Python and
> called polymorphically from C++. (so i'm not wrapping c++ for use from
> python)

Huh?  It sounds like that's exactly what you're doing.

> for instance i want to allow users of my software to override the
> virtual void Login::OnLog(std::string) function using python.
>
>
> that's what 'wrapper' and 'get_override' is used for, right?

Right.

> if i would use it in python only, overriding works without these
> stuff too.
>
> so i duplicated the example at
> http://www.boost.org/libs/python/doc/tutorial/doc/html/python/exposing.html#python.virtual_functions_with_default_implementations
>
> and it works fine inside python as advertised:
>>>> base.f()
> 0
>>>> derived.f()
> 42
>
> but for this i wont need 'wrapper' and 'get_override'. This works for
> any boost c++ extension.
> now for the interesting stuff: (ak the problem)
>
> after running the script i do:
>
> 	Base *b = new BaseWrap;

BaseWrap is purely for the use of the Python class Called Base, which
is generated by class_<BaseWrap, ...>.  You shouldn't be using it
directly.

> 	int result = b->f();
> 	delete b;
>
> now i would expect the result to be 42, but i got 0.
> ====================================================

You have no right to expect anything in particular from that.  Of
course that doesn't build a Python derived object.  Just imagine you
had several Python classes derived from the Python class called Base.
Which one would it choose?

> so how to put 'wrapper' and 'get_override' to any use?

What you really seem to be asking is how do you construct an instance
of derived (a class defined in Python) from C++.  

> my sollution is to obtain the PyObject* of 'derived'
> and to force it manualy into m_ptr of 'wrapper' using
> 'detail::initialize_wrapper'

Stuff in detail namespaces is undocumented and not for user
consumption.  You shouldn't be using it directly.

> then 'get_override' worked as expected and i got 42.

 // substitute the module name where "derived" is defined for "__main__"
 python::object module = python::import("__main__");  
 python::object derived = module.attr("derived");
 python::object b = derived( /* arguments to __init__ here */ );
 int result = b.f();

The import function is not documented for 1.33.1 but you can see it at
http://boost-consulting.com/boost/libs/python/doc/v2/reference.html#embedding
(otherwise you have to use the Python 'C' API to do something
similar).

[Joel, any chance of getting
http://www.boost.org/libs/python/doc/tutorial/doc/html/python/embedding.html
updated with the facilities described in
http://boost-consulting.com/boost/libs/python/doc/v2/reference.html#embedding?]


HTH,

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com



From joel at boost-consulting.com  Wed Jun 28 11:12:26 2006
From: joel at boost-consulting.com (Joel de Guzman)
Date: Wed, 28 Jun 2006 17:12:26 +0800
Subject: [C++-sig] extending a c++ object by overriding it's
	virtual	functions..
In-Reply-To: <uirmmsdex.fsf@boost-consulting.com>
References: <44A0955A.8060008@gmx.de> <uirmmsdex.fsf@boost-consulting.com>
Message-ID: <44A247FA.9070404@boost-consulting.com>

David Abrahams wrote:
> 
> [Joel, any chance of getting
> http://www.boost.org/libs/python/doc/tutorial/doc/html/python/embedding.html
> updated with the facilities described in
> http://boost-consulting.com/boost/libs/python/doc/v2/reference.html#embedding?]

I added this in my TODO list. I'm neck deep with tasks to do
though, so I'm not sure when I can get to this. I'd appreciate
it if someone can do it ahead of me.

Regards,
-- 
Joel de Guzman
http://www.boost-consulting.com
http://spirit.sf.net


From kleistereimer at gmx.de  Wed Jun 28 11:43:57 2006
From: kleistereimer at gmx.de (Kleistereimer)
Date: Wed, 28 Jun 2006 09:43:57 +0000
Subject: [C++-sig] extending a c++ object by overriding it's virtual
	functions..
Message-ID: <44A24F5D.5010404@gmx.de>

hi dave,

thankyou for your explanations and sourcecode.

i have some more questions about it thought.



first here is the sourcecode again:


//Python
>>> class Derived(Base):
...     def f(self):
...         return 42
>>> base = Base()
>>> derived = Derived()
>>> base.f()
0
>>> derived.f()
42

//C++ code from your answer
>1: // substitute the module name where "derived" is defined for "__main__"
>2: python::object module = python::import("__main__");
>3: python::object derived = module.attr("derived");
>4: python::object b = derived( /* arguments to __init__ here */ );
>5: int result = b.f();




1)
the mentioned include file <boost/python/import.hpp> is not in boost
1.33.1. where can i get it? is it in cvs? does it suffice if i get this
.hpp file only, or do i need the complete cvs version? for now i used
the c api.. instead of this line.

2)
what does the above code exactly? i think it instantiates a python
'Derived' class to object 'derived', so line 3 has to be:
 3 python::object derived = module.attr("Derived"); //capital 'D'
and at line 5 'f' is called on this instance. this gives me  a compiler
error:
"f is no element of ..::api::object". how to solve this?
i guess it involves something like:
 python::call_method<int>(b.ptr(),"f")
but this fails in arg_to_python.hpp:181
"cannot_convert_raw_PyObject<T*>::to_python_use_handle_instead();"
how to do it corectly?

3)
so this scheme is not 'overwriting c++ virtual functions with python
code' directly, instead it's 'deriving a python class from a c++ class,
overwriting  virtual functions, and then making an instance of this
python class and access it from c++'
right?


this means if i have some heavily used c++ class which needs only some
minor functions tweakable from python, i should better do it like so:


class py_MyClass:
	def Log(self):
		...		

class MyClass
{
	MyClass() {m_PyObject = Instantiate("py_MyClass");} //using line 3 and 4
	~MyClass() {delete m_PyObject;}
	...
	...
	void Log(std::string) {python::call_method<void>(m_PyObject, "Log"); }
};

because this calls into python only if one of the 'pythonized' functions
is needed (which is seldom) but not if other c++ functions are used
(which is extremely often)


did i understood it right?


regards
kl


From dave at boost-consulting.com  Wed Jun 28 14:40:59 2006
From: dave at boost-consulting.com (David Abrahams)
Date: Wed, 28 Jun 2006 08:40:59 -0400
Subject: [C++-sig] extending a c++ object by overriding it's virtual
	functions..
References: <44A24F5D.5010404@gmx.de>
Message-ID: <uy7vho3l0.fsf@boost-consulting.com>

Kleistereimer <kleistereimer at gmx.de> writes:

> hi dave,
>
> thankyou for your explanations and sourcecode.
>
> i have some more questions about it thought.
>
>
>
> first here is the sourcecode again:
>
>
> //Python
>>>> class Derived(Base):
> ...     def f(self):
> ...         return 42
>>>> base = Base()
>>>> derived = Derived()
>>>> base.f()
> 0
>>>> derived.f()
> 42
>
> //C++ code from your answer
>>1: // substitute the module name where "derived" is defined for "__main__"
>>2: python::object module = python::import("__main__");
>>3: python::object derived = module.attr("derived");
>>4: python::object b = derived( /* arguments to __init__ here */ );
>>5: int result = b.f();

Whoops:

   int result = b.attrib("f")();


> 1)
> the mentioned include file <boost/python/import.hpp> is not in boost
> 1.33.1. where can i get it? is it in cvs? 

Yes.  Get the RC_1_34_0 branch.


> does it suffice if i get this
> .hpp file only, or do i need the complete cvs version? 

I can't guarantee it works without the complete CVS

> for now i used the c api.. instead of this line.
>
> 2)
> what does the above code exactly? i think it instantiates a python
> 'Derived' class to object 'derived', 

No, creates a reference to the Python class object called "derived."  

> so line 3 has to be:
>  3 python::object derived = module.attr("Derived"); //capital 'D'

Your post mixes case ["so i need to hardcode the name of the
instantiated class ('derived')"] so I used lowercase, but if you
define it as Derived, obviously, capitalize.

> and at line 5 'f' is called on this instance. this gives me  a compiler
> error:
> "f is no element of ..::api::object". how to solve this?
> i guess it involves something like:
>  python::call_method<int>(b.ptr(),"f")
> but this fails in arg_to_python.hpp:181
> "cannot_convert_raw_PyObject<T*>::to_python_use_handle_instead();"
> how to do it corectly?

   int result = b.attrib("f")();

> 3)
> so this scheme is not 'overwriting c++ virtual functions with python

I assume you mean "overriding"

> code' directly, instead it's 'deriving a python class from a c++ class,
> overwriting  virtual functions, 

One can only override member functions by deriving in C++, so I don't
see where your expectations for a more "direct" overriding could come
from.

> and then making an instance of this python class and access it from
> c++' right?

That's what you seemed to be wanting to do.

> this means if i have some heavily used c++ class which needs only some
> minor functions tweakable from python, i should better do it like so:
>
>
> class py_MyClass:
> 	def Log(self):
> 		...		
>
> class MyClass
> {
> 	MyClass() {m_PyObject = Instantiate("py_MyClass");} //using line 3 and 4
> 	~MyClass() {delete m_PyObject;}
> 	...
> 	...
> 	void Log(std::string) {python::call_method<void>(m_PyObject, "Log"); }
> };
>
> because this calls into python only if one of the 'pythonized' functions
> is needed (which is seldom) but not if other c++ functions are used
> (which is extremely often)
>
>
> did i understood it right?

Not really.  If you want to use derived polymorphically from C++, you
can also do this:

    python::object d = derived( /*args*/ );
    Base& b = python::extract<Base&>( d );
    b.f();

Now only the functions you've made overridable in Python (e.g. f) will
use any part of the Python API at all, and those that aren't actually
overridden will only do a fast lookup, find nothing, and continue with
C++ code, without interpreting any bytecode.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com



From austin.bingham at gmail.com  Thu Jun 29 23:31:01 2006
From: austin.bingham at gmail.com (Austin Bingham)
Date: Thu, 29 Jun 2006 16:31:01 -0500
Subject: [C++-sig] Another project using boost.python
Message-ID: <bf1abd10606291431k17f70eb3n6491b312e002f47e@mail.gmail.com>

The boost.python website asks for submissions of projects using
boost.python, so here's another one. pyxiph
(http://sourceforge.net/projects/pyxiph) is a set of bindings to
several of the xiph.org libraries (currently ao, ogg, and vorbis).
Compared to python extensions I've written in the past using the
standard C API, this was a breeze! I find the boost.python approach
easier to reason with, easier to modify, much easier to get correct,
and easier to explain. Count me as one happy customer.
-- 
Austin Bingham

"If I were creating the world I wouldn't mess about
with butterflies and daffodils. I would have started
with lasers, eight o'clock, Day One!"
Evil


From j.reid at mail.cryst.bbk.ac.uk  Fri Jun 30 15:13:33 2006
From: j.reid at mail.cryst.bbk.ac.uk (John Reid)
Date: Fri, 30 Jun 2006 14:13:33 +0100
Subject: [C++-sig] boost::python std::vector< std::string > crash
In-Reply-To: <e7ej2v$17e$1@sea.gmane.org>
References: <e7ds1k$bmu$1@sea.gmane.org>	<449AA3D1.1080805@boost-consulting.com>	<e7eeqp$gt3$1@sea.gmane.org>
	<e7ej2v$17e$1@sea.gmane.org>
Message-ID: <e83834$jps$1@sea.gmane.org>

Ok this seems to be a crash in debug mode when I am picking up the 
release version of boost_python.dll in my path.

How to make boost build v1 install boost_python.dll next to my 
extension? Or do I need to manually copy the correct (debug/release) 
version every time I want to debug or run the release version?

I've tried adding

		<dll>@boost/libs/build/boost_python

to my stage target to no avail.

I have followed the 'Creating Packages' guidelines in the 'General 
Techniques' documentation but could not find any documentation on how to 
set up a build environment where I can alternate between debug and 
release builds easily.

John.



John Reid wrote:
> 
> John Reid wrote:
> 
>>Joel de Guzman wrote:
>>
>>
>>>[snips]
>>>
>>>That's odd. I can't reproduce your results. To be sure, I added
>>>your test in libs/python/test/vector_indexing_suite.cpp/py.
>>>I'm getting no crash. Please check out the test suite and run it.
>>>
>>>Regards,
>>
>>
> 
> I've investigated further. I wasn't aware of the correct way to run the 
> python tests. Now I've recompiled a clean version of boost cvs_head 
> everything works ok if I run the tests using 'bjam -sPYTHON_TEST_ARGS= 
> test' in the 'boost\libs\python\test' directory.
> 
> What might be relevant is that if I run 'bjam' with no arguments, I can 
> generate 'vector_indexing_suite_ext.pyd' in the 
> 'boost\bin\boost\libs\python\test\vector_indexing_suite_ext.pyd\vc-8_0\debug\threading-multi'
> directory. When I manually set PYTHONPATH to this directory and run 
> 'python vector_indexing_suite.py' I get the crash as before. Is this 
> expected, relevant and/or interesting? I would have thought this would 
> just have run that particular test.
> 
> Any help appreciated as always. Thanks,
> John.